PATENTSCOPE will be unavailable a few hours for maintenance reason on Tuesday 19.11.2019 at 4:00 PM CET
Search International and National Patent Collections
Some content of this application is unavailable at the moment.
If this situation persists, please contact us atFeedback&Contact
1. (WO2017137089) USER EQUIPMENT PROFILING FOR NETWORK ADMINISTRATION
Note: Text based on automatic Optical Character Recognition processes. Please use the PDF version for legal matters

USER EQUIPMENT PROFILING FOR NETWORK ADMINISTRATION

TECHNICAL FIELD

The present invention relates to a user profiling system for a communication network and a user profiling method for a communication network.

The present invention also relates to a computer-readable storage medium storing program code, the program code comprising instructions for carrying out such a method.

BACKGROUND

The number of wireless and mobile devices is expected to increase considerably. Along with it, a huge increase of mobile traffic will also take place. The deployment of 5G cellular net-works targets to support this vast number of devices, while at the same time the existing 3GPP specifications will keep on supporting legacy cellular access networks (e.g., GSM, HSPA, LTE, LTE-A) as well as alternative radio access technologies (e.g., WiFi). In this environment, the end users will have access to a diverse set of services (ranging from high definition video and audio, web browsing, games, to keep alive messages etc.). The new services (e.g., augmented reality, cloud services, car to car communication, etc.) lead to a paradigm shift of network access mechanisms since the users want connectivity everywhere (including indoor and outdoor environments, environments with ultra-dense or limited infrastructure, or where the environment is extremely crowded, etc.) with various mobilities (e.g., high speed trains, random indoor mobility, moving crowd, etc.).

In such a demanding environment, the use of user network information is imperative for enabling an efficient use of the network resources. However, the acquisition and distribution of user network information may cause an additional load for the communication network.

The combination of user network information elements enables more accurate decision making. However, such combination is a demanding task in terms of: Accessing, processing, and combining in real time many information elements as well as identifying the user preferences in relation to specific parameters.

Specifically, accessing and processing in real time many information elements, it is a demanding task for the network. Additionally, combining these information elements increases the number of variables, which consequently increases the problem state space, a problem known in the research area as the curse of dimensionality. The term curse of dimensionality refers to the fact that in the absence of simplifying assumptions the sample size needed to estimate a function of several variables to a given degree of accuracy in order to get a reasonably low variance estimate grows exponentially with the number of variables.

Context aware mechanisms for predicting the future behaviors of users exploit current infor-mation (i.e., current contextual information) so as to predict the user behavior including the user mobility, the service that the user will access and/or the duration of the service access. However, in existing systems, the predictions are often not accurate, or they cannot make predictions for longer time periods.

SUMMARY OF THE INVENTION

The objective of the present invention is to provide a user profiling system and a user profiling method, wherein the user profiling system and the user profiling method overcome one or more of the above-mentioned problems of the prior art.

A first aspect of the invention provides a user profiling system for a communication network, comprising:

an information acquisition unit configured to acquire user network information for a plurality of users of the communication network,

- a profile generation unit configured to generate a plurality of user profiles for the plurality of users based on the acquired user network information,

a profile determination unit configured to determine, based on a current characteristic of a specific user, from the plurality of user profiles one or more candidate user profiles for the specific user, and

- a configuration unit that is configured to configure a network component of the communication network based on an active user profile of the one or more candidate user profiles.

The user profiling system of the first aspect can operate in two phases: In a first phase ("learning phase"), it can learn typical behaviors from user network information. The user network information in the following is not limited to network information, but can include all information available about a user, their user equipment and their interaction with the network. Typically, the user network information includes historic information, such as information about user-related network events in the past.

In a second phase, the system can apply the learned knowledge by determining one or more candidate user profiles for a user and an active user profile from the one or more candidate profiles. This user profile may be associated with a typical behavior of the user, including typical kinds of network interaction of the user, e.g. a typical mobility level and/or a typical amount of data transfer.

The profile determination unit can be configured to determine a plurality of candidate user profiles. The user profiles may comprise or be associated with input parameters or input parameter ranges that indicate when which one of the user profiles should be an active user profile. For example, a user profile may indicate a time range when it can be an active user profile.

In other embodiments, the profile determination unit can be configured to always determine only one candidate user profile for a user. In this case, the candidate user profile is the active user profile. In other words, the profile determination unit can be configured to directly determine the active user profile. This may be preferable if frequent communications between the profile determination unit and other units such as the configuration unit are not an issue.

The configuration unit can configure a network component by directly setting one or more parameters of the network component. The configuration unit can also be part of the network component that it is configuring. In other embodiments, the configuration unit can be a distribution unit that is configured to distribute the active user profile or an indication of the active user profile to components in the communication network. The configuration unit can be implemented e.g. in base station, mobility management entity or even in a user equipment.

People tend to have specific behaviors, in terms of properties including, but not limited to specific accessed services, duration of accessed service, mobility, etc. These specific behav- iors are related to characteristics such as location, date, time, etc. For example, some people tend to access the internet using their user equipments, or tend to have long calls. It is beneficial for the operator if the latter has such knowledge, for managing the communication network. All this knowledge can be used for determining the user profiles.

Based on historic user network information about all users of a communication network, an overall list of user profiles can be built. For a specific user, a list of "enabled" user profiles can be determined that are suitable for this user. Subsequently, based e.g. on characteristics, such as an approximate location of the specific user and the date/time a list of "candidate" user profiles for this user can be determined. For example, if the user is at a shopping mall in the late afternoon, based on past information, the system may predict that the user is only making few and short phone calls whereas when the same user is at home he behaves as a heavy data user watching high definition videos for hours. However, the user behavior will typically still depend on further information such as a battery level of remaining credits on a prepaid account. Thus, based on further real time information such as battery level or remaining credits, the active user profile may be chosen.

The user behavior may change based on a large number of information, including but not limited to the user equipment type, the battery level of the user equipment, the charging status of the user account (e.g., remaining credits, etc.), the user overall user income, the user educational education level, etc. Due to the complexity and high dimensionality of this information, reducing the information to a selection from a set of predefined user profiles significantly reduces computational complexity.

The profiling system of the first aspect provides mechanisms that solve the abovementioned challenges and enable an efficient network management and control by automatically building user profiles which are associated with a predicted user behavior. The proposed mechanism can take advantage of existing data mining mechanisms for identifying the most impacting parameters in the user behavior. Supervised and/or unsupervised learning mechanisms can be used for the identification of the user profiles. These profiles are being used for optimizing mobility management operations such as handover and/or other control operations such as call admission control, cell selection and reselection, etc.

A user profile can be associated with a predicted user behavior in the sense that it comprises a list of parameters that indicate a predicted user behavior. This can include: expected services to be accessed, the access rate, access duration, user mobility. These parameters can indicate different behavior based on the values of additional variables, including but not limited to the location/day/time as well as battery level and charging status.

The association between user profiles and expected user behaviors can be implemented e.g. using a table that associates different user profiles with different behavior parameters. The behavior parameters can be individual values (e.g. an expected movement speed), value rang-es (e.g. from an expected minimum movement speed to an expected maximum movement speed) and probability distributions (e.g. a probability distribution that indicates which movement speeds are how likely).

In a first implementation of the user profiling system according to the first aspect, the current characteristic of the specific user includes at least one of the following: a time, a location, a reception level, a battery level, and a credit level of the specific user.

These current characteristics allow the system to accurately determine a current user profile. For example, a low battery level is typically associated with less network usage. A user may be less likely to start streaming a YouTube video if the battery level already indicates that the remaining battery life is not sufficient for finishing the video playback. The reception level and/or the battery level can be determined at the user equipment and then transmitted to the network.

The user network information may include all kinds of information that are exchanged among network entities, even heterogeneous ones so as to solve challenging networking problems such as management and control of the network resources. The broad definition of network user information implies that all information types may be used for the optimization of the network management and control. This includes, but is not limited to radio information (e.g., RSS, RSRP, RSRQ, backhaul link capacity and quality, etc.), mobility information (e.g., user speed, number of handovers, etc.), and power/energy information (e.g., battery consumption rate, battery level, etc.). From this point, the term "user network information" may refer to all information types that may be used for decision making, in addition to radio measurements such as Received Signal Strength - RSS, Signal to Interference Ratio.

A simple example of modelling user behavior depending on user network information could be the following:

• Joe is stationary at the Office every weekday from 9:00 - 18:00, he performs long voice calls, and he does not access internet through his cell phone.

• Joe at his House every weekday after work is stationary, accesses web applications via his cell phone using WiFi, and he does not make phone calls.

• Joe in city center every Saturday from 10:00 - 16:00 is highly mobile, he performs short voice calls, and he does NOT access the Internet through his cell phone.

Such knowledge can enable the network to make decisions regarding the placement to the user in a specific radio access technology (RAT) or to a specific layer (e.g., macro cell or micro cell) and the interference cancelation methods to be used (e.g., CoMP or interference coordination, etc.).

In a second implementation of the user profiling system according to the first aspect, the profile determination unit and/or the network component is configured to select the active user profile from the one or more candidate user profiles of the user based on a further current characteristic of the specific user.

The one or more candidate user profiles can be transmitted e.g. to the network component which then uses an internal algorithm to determine the active user profile based on the values of additional variables such as day/time, location, battery level and charging (in terms of monetary credits) status. To this end, the list of user profiles, in particular the candidate user profiles, can comprise for each of the candidate user profiles an indication, e.g. a parameter range that indicates when this user profile can be the active user profile.

For example, the further current characteristic can be one that changes more rapidly. Thus, it can be preferable to make this decision based on the further current characteristic locally at the network component based on the further current characteristic that is available at the network component. This has the advantage that a signalling effort within the network is reduced because the one or more candidate user profiles need to be transmitted only once.

In a third implementation of the user profiling system according to the first aspect, the user profiling system further comprises a distribution unit configured to distribute a subset of the plurality of user profiles in the communication network, wherein in particular the distribution unit is comprised in a dedicated unit, in particular a home subscriber server and/or a home location register.

As outlined above, the configuration unit can be the distribution unit, i.e., the configuration unit configures network components by passing user profile information to the network components. In other embodiments, the distribution unit distribute active user profiles (possibly including associated user behavior information for each user profile) and the configuration unit configures a network component based on the distributed active user profiles.

For example, the distribution unit may be configured to distribute enabled user profiles of users that are known to be present in a certain geographical region to network components, in particular base stations, in that geographical region. This has the advantage that a reduced number of user profiles can be kept near to where the network components may require access to the user profiles.

The distribution unit can also be configured to distribute the subset of the plurality of user profiles, e.g. the active user profile for a specific user, to a user equipment of the user.

In a fourth implementation of the user profiling system according to the first aspect, the information acquisition unit is configured to acquire the user network information from at least one of the following: a user equipment, a base station, a mobility server, a charging gateway, a data router, and a network database of the communication network.

This has the advantage that the user network information can be acquired from different kinds of components of the communication network, thus providing the profiling system with more information sources.

In a fifth implementation of the user profiling system according to the first aspect, the profile generation unit is further configured to determine, based on the acquired user network information, one or more enabled user profiles that are enabled for the specific user and the profile determination unit is configured to determine the one or more candidate user profiles from the one or more enabled user profiles of the specific user.

In other embodiments, there is another unit, separate from the profile generation unit, that determines the enabled user profiles for the specific user.

This has the advantage that a (potentially large) number of user profiles that are available in the system as a whole can be reduced to a smaller number of user profiles that are potentially relevant for a certain user. The profile generation unit and/or the profile determination unit may have access to or may comprise a database. In particular, the database may be located at one or more central components of the communication network.

In a sixth implementation of the user profiling system according to the first aspect, the user network information comprises at least one of the following:

- a network measurement, in particular at least one of the following: a received signal strength, a RSRP/RSRQ, a backhaul link capacity, a backhaul link quality, a packet loss, a connection delay, and an interface information of a user equipment, a mobility information, in particular a user movement speed and/or a number of handovers of a user equipment,

- a service measurement information, in particular at least one of the following: an accessed service type, an accessed service duration, an accessed service characteristic, a packet size, a packet transmission interval, a packet reception interval, an uplink bit rate, a downlink bit rate, a jitter, a packet loss, and a packet error rate of a base station or a user equipment,

- a social information of the user, in particular at least one of the following: an age, an employment, a profession, an education level, an income, and a gender of a user, a user contract information, in particular a contract ID and/or a contract expiration date of a user,

a credit level information, in particular a credit model and/or an available credit level of a user,

a list of one or more user-chargeable events, and

a user equipment information, in particular at least one of the following: an available battery level, a maximum battery charging level, a device central processing unit description, a memory size, an operating system identifier, a screen size, a screen resolu- tion, a power information, a current CPU usage level, a current memory usage level, an information about one or more protocols supported by the user equipment, and an information about one or more physical interfaces of the user equipment.

Including one or more of this information in the user network information allows the user profiling system to create user profiles that accurately characterise different user behaviors. One or more of the above user network information may be acquired locally at the user equipment and transmitted to the communication network so that it is available at components of the user profiling system that are distributed in the communication network.

In a seventh implementation of the user profiling system according to the first aspect, the profile generation unit is configured to generate the plurality of user profiles by performing supervised and/or unsupervised learning on the acquired user network information,

in particular wherein the profile generation unit is configured to determine one or more rele-vant features of the acquired user network information by analysing at least one of an Information Gain, a X2 statistic, and a Mutual Information of the acquired user network information.

Determining relevant features using one or more of the above techniques has the advantage that the large number of dimensions of the acquired network user information is reduced, which makes the information easier to process and transmit within the communication network.

The profile generation unit can be configured to generate the plurality of user profiles using at least one of decision trees, Support Vector Machines, and clustering. These have been shown to be particularly efficient learning methods.

Unsupervised and supervised learning can also be combined. For example, the system can first perform unsupervised learning by clustering the users into a number of user groups. Sub-sequently, using techniques from supervised learning, certain enabled user profiles can be associated with predicted user behaviors.

In an eighth implementation of the user profiling system according to the first aspect, the network component is configured to make a control function, in particular a radio resource man- agement action and/or a handover decision based on the active user profile, wherein in particular the network component is configured to decide how to execute the control function based on the active user profile.

The radio resource management action may comprise allocation of resources, scheduling and/or interference management.

As outlined above, the active user profile can be associated with one or more parameters about an expected user behavior. A network component can be configured to make a control function based on these one or more parameters. This has the advantage that the network operation can be adjusted based on an expected user behavior.

In a ninth implementation of the user profiling system according to the first aspect, the user profiling system further comprises a user equipment that is configured to select a cell and/or a radio access technology while it is in an idle mode.

For this purpose, the user profiling system can be configured to transmit an active user profile to a user equipment and the user equipment can be configured to select the cell and/or the radio access technology based on the transmitted active user profile.

The network can inform a user equipment about an active user profile and the user equipment can then make a cell/RAT decision, e.g. cell selection, based on the active user profile. For example, the active user profile may be passed to the user equipment during a previous attachment to the network.

In a tenth implementation of the user profiling system according to the first aspect, the user profile is associated with a predicted behavior of a user, in particular a predicted behavior related to at least one of the following: a predicted mobility level, a predicted movement direction, a predicted type of accesses services, and a predicted service access duration.

Determining user profiles such that they are associated with a predicted behavior of the user has the advantage that network control decisions can be based on the associated predicted behavior.

A second aspect of the invention refers to a user profiling method for a communication network, the method comprising:

acquiring user network information for a plurality of users of the communication network,

generating a plurality of user profiles for the plurality of users based on the acquired user network information,

determining, based on a current characteristic of a specific user, from the plurality of user profiles one or more candidate user profiles, and

making one or more network control function decisions related to a user equipment of the specific user based on an active user profile of the one or more candidate user profiles.

The profiling methods according to the second aspect of the invention can be performed by the user profiling system according to the first aspect of the invention. Further features or implementations of the profiling method according to the second aspect of the invention can perform the functionality of the user profiling system according to the first aspect of the invention and its different implementation forms.

The method of the second aspect can comprise two phases of operation, the offline one and the online one. In the offline phase the required inputs are being gathered and processed, so as to extract the behavioral profiles. The profiles are being distributed to the network components and used in online manner by combining online information with the profiles for more accurate prediction. The predicted behavior can then be used for network management and control operations. Afterwards, the overall network performance can be evaluated and fine-tuned according to the effectiveness in the network performance.

The offline phase can comprise one or more of the following functions:

• Gathering user network information, including e.g. a user personal context and history, and storing it in a logically centralized entity.

• Using this collected user network information for identifying the most relevant parameters.

• Extraction of the user profiles, per several variable parameters including but not limited to location, time, subscription, battery level, charging status, mobility pattern, etc.

Formation of a list of user profiles for each user for each one of the previous parameters; and

Storing the list of user profiles.

Storing the list of user profiles can involve writing a table that associates users with user profiles that are enabled for these users. It can further involve writing a table that associates user profiles with expected behaviors of these user profiles.

The online phase can include one or more the following two functionalities:

· The distribution of the user profiles to different networking entities. The networking entities include but are not limited to user equipments, Access Stratum and Non Access Stratum control entities, databases etc.

• The integration of the user profiles in the network control and management function and the combination with online information, so as to enable more accurate prediction of the user behavior.

In a first implementation of the user profiling method of the second aspect, the method further comprises:

evaluating a performance of the communication network, and

- adapting one or more of the plurality of user profiles based on the evaluated performance of the communication network.

This has the advantage that the user profiles can be adapted such that the performance of the communication network is further improved.

In a second implementation of the user profiling method of the second aspect, evaluating the performance of the communication network comprises evaluating whether a network decision that is based on active user profiles have been beneficial to the performance of the communication network.

This has the advantage that e.g. particular profile-based decisions that turn out to be detrimental to the performance of the communication network can be eliminated.

A third aspect of the invention refers to a computer-readable storage medium storing program code, the program code comprising instructions for carrying out the profiling method of the second aspect or one of the implementations.

BRIEF DESCRIPTION OF THE DRAWINGS

To illustrate the technical features of embodiments of the present invention more clearly, the accompanying drawings provided for describing the embodiments are introduced briefly in the following. The accompanying drawings in the following description are merely some embodiments of the present invention, but modifications on these embodiments are possible without departing from the scope of the present invention as defined in the claims.

FIG. 1 is a block diagram illustrating a user profiling system in accordance with an embodiment of the present invention;

FIG. 2 is a flow chart illustrating a user profiling method in accordance with a further embodiment of the present invention;

FIG. 3 is a further flow chart illustrating phases and steps of a method in accordance with a further embodiment of the present invention;

FIG. 4 is a diagram illustrating an exemplary implementation of the process of acquiring and using the list of user profiles from the Profiling Engine in accordance with a further embodiment of the present invention;

FIG. 5 is a diagram illustrating an exemplary implementation of the process of acquiring and using the list of user profiles from a networking component from the Profiling Engine in accordance with a further embodiment of the present invention;

FIG. 6 is a diagram illustrating an exemplary implementation of the process of acquiring and using the list of user profiles from a networking component from the Profiling Engine in accordance with a further embodiment of the present invention;

is a diagram illustrating an exemplary implementation of the process of acquiring the list of user profiles from eNB from the Profiling Engine in an

LTE/LTE-A network in accordance with a further embodiment of the present invention;

is a diagram illustrating an exemplary implementation of the process of handover using the list of user profiles in an LTE/LTE-A network in accordance with a further embodiment of the present invention;

is a diagram illustrating an exemplary implementation of the process of acquiring the list of user profiles from the Profiling Engine in an LTE/LTE-A network for Cell Selection/Reselection or Call Admission Control processes in accordance with a further embodiment of the present invention; and

is a diagram illustrating an exemplary implementation of the process of Call Admission Control using the list of user profiles in an LTE/LTE-A network in accordance with a further embodiment of the present invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

FIG. 1 shows a user profiling system 100 for a communication network in accordance with an embodiment of the present invention. The user profiling system 100 comprises an information acquisition unit 110, a profile generation unit 120, a profile determination unit 130 and a configuration unit 140. The units 110-140 may be distributed over different locations of the communication network or one or more of the units 110-140 may be comprised in a profiling engine which may be located in one device, e.g. a component of the communication network.

The information acquisition unit 110 is configured to acquire user network information for a plurality of users of the communication network. For example, the information acquisition unit may be configured to derive the user network information from a logging unit (not shown in FIG. 1) that is configured to log user network information in a log file or a log database.

The profile generation unit 120 is configured to generate a plurality of user profiles for the plurality of users based on the acquired user network information. In particular, the profile generation unit 120 may be configured to use machine learning techniques to generate the plurality of user profiles.

The profile determination unit 130 is configured to determine, based on a current characteristic of a specific user, from the plurality of user profiles one or more candidate user profiles for the specific user.

The configuration unit 140 is configured to configure a network component of the communication network based on an active user profile of the one or more candidate user profiles. For example, the configuration unit 140 can be localized in a processor of a network component that it is configuring.

FIG. 2 shows a user profiling method 200 for a communication network in accordance with an embodiment of the present invention.

The method comprises a first step 210 of acquiring user network information for a plurality of users of the communication network. The first step 210 may be carried out repeatedly, e.g., in regular intervals, in order to retrieve newer user network information.

The method comprises a second step 220 of generating a plurality of user profiles for the plurality of users based on the acquired user network information. The second step may be based on machine learning methods. The second step 220 may be carried out repeatedly, e.g. each time when new user network information is available.

The method comprises a third step 230 of determining, based on a current characteristic of a specific user, from the plurality of user profiles one or more candidate user profiles. In embodiments, only one candidate user profile is determined, which is the active user profile.

The method comprises a fourth step 240 of making one or more network control function decisions related to a user equipment of the specific user based on an active user profile of the one or more candidate user profiles.

One exemplary use of the list of user profiles extracted from the Profiling Engine could be its application in Handover in an LTE/LTE-A network. Some of the following examples describe how the user behavioral prediction in terms of accessed services, access service duration, user mobility, etc. could be applied in the LTE/LTE-A network. The same mechanisms could be applied in other types of cellular networks such as GSM, UMTS, etc. or interworking networks of more than one type e.g., interworking GSM, UMTS, LTE/LTE-A, and WiFi networks, etc.

FIG. 3 shows a flow chart of the phases and steps of a further profiling method in accordance with a further embodiment of the present invention.

The method shown in FIG. 3 comprises an offline phase 302 and an online phase 304.

The offline phase involves gathering of user network information for generating the list of profiles. This significantly reduces the overhead in the network. Compared to context aware mechanisms available in the prior art that directly predict the user behavior, the proposed method does not require online interactions among several network components, thus making the operation of the proposed system more efficient. This characteristic makes the scheme scalable. Also, the grouping of the users based on user profiles has complexity gains by reducing the problem space, since it enables the network to predict the user behavior through user groups (corresponding to users with the same user profile) and not individually, which reduces significantly the problem space and thus handles the curse of dimensionality problem.

In detail, the offline phase 302 comprises first, second, third, fourth, fifth and eighth steps 310, 320, 330, 340, 350 and 380 as illustrated in FIG. 3.

In the first step 310, user network information, including personal context and user history data are acquired.

Step 310 can involve information gathering in a logically centralized entity which is responsi-ble for the processing and the analysis of the data. The logically centralized entity can be referred to as Profiling Engine. The information that it is collected from the Profiling Engine per user equipment may include, but is not limited to the following categories:

• Network measurements, comprising received signal strength, RSRP/RSRQ, backhaul link capacity and quality, packet loss, delays, and interface information.

• Mobility Information, comprising user speed and number of handovers.

• Service measurements, comprising accessed service type, accessed service duration, and accessed service characteristics (i.e., packet size, packet transmission interval, packet reception interval, uplink and downlink bit rate, acceptable jitter, acceptable packet loss, acceptable packet error rate, etc.)

• Social information, comprising age, employment/profession, education, income, and gender.

• User contract information comprising Contract Id, Signature and Expiration dates.

• Charging Information, comprising charging model, and available credits.

• User Equipment description information, comprising available battery, maximum battery charging, device central processing unit description, memory, operating system, screen size, screen resolution, power/energy information (e.g., battery consumption rate, battery level, current CPU, current memory, information about protocols supported by the user equipment (e.g., Type of protocol, Required memory, Required CPU), and data describing the physical interfaces offered by a device (e.g., uplink rate, downlink rate, round trip delay, errors, packets sent, packets received, etc.).

The contextual information can be captured in vectors comprising the above information types. Preferably, each vector is uniquely characterized by the combination of the unique user identifier (e.g., IMSI) and the extensive time information; other information fields may be used for the unique characterization of each observation such as the location, the battery, the charging status, etc. Each unique information vector can be referred to as "observation".

The measurements may be represented by natural numbers, real positive numbers or event nominal values (e.g., for representing location, time, charging status, social information, etc.).

In the second step 320 the most relevant parameters are identified. This can be performed e.g. using automated feature selection.

Once all the aforementioned information fields have been collected for all users, in step 320 the most relevant parameters are identified by the Profiling Engine, so as to exclude information fields that do not contain relevant information so as to reduce the number of variables involved in the analysis. This process is called dimensionality reduction and aims at simplifying the available dataset, by transformations that embed data in a space of significantly lower dimensions. In the proposed scheme dimensionality reduction is achieved using feature selection schemes; the mechanisms that could be used for this process include but are not limited to Information Gain, X2 statistic, and Mutual Information. Such schemes attempt to quantify the influence of a feature to the description of a class and retain only the features which appear to contribute more knowledge to the classification task. All methods of this family are essentially heuristic and operate based on statistical information available in the dataset.

Given a large set of observations X= UXi, with Xi being described by n distinct variables vj, the effect of vj on the classification of Xi into class CXi can be quantified using at least one of the following:

• Information Gain, X2 statistic, and Mutual Information.

• Information Gain computes the expected entropy reduction by the categorization of an instance Xi into class CXi due to the value of variable vj.

• X2 statistic attempts to quantify the independence of a class G and a variable vj .

• Mutual information is similar to X2 and depicts the dependence of a class G by a variable v.

A prerequisite of these methods is that classification information is available in advance for all or at least some indicative observations of the dataset. For the feature extraction selection, even though the previous mechanisms are proposed for use, other schemes available in the state of the art can also be used (e.g., Chi square). Dimensionality reduction may be achieved by feature extraction schemes as well which aim at constructing combinations of the variables that reduce the dimensions of the dataset while still describing the data with sufficient accuracy (e.g., multidimensional scaling - MDS, Linear Discriminant Analysis, etc.).

In a third step 330, user profiles (per location, subscription, time, battery level, charging level, mobility pattern, etc.) are created using data mining (using supervised and/or unsupervised learning). To this end, after concluding on the importance of each parameter, the reduced dataset can be used for the extraction of the creation of the profiles by the Profiling Engine.

• In supervised learning the focus in on the identification of a learning model from a large set of data and its subsequent application on new, unobserved data which come from the same generation procedure as the original one. In general, all supervised learning algorithms are assumed to receive as input a large set of observation appropriately labelled

(i.e. Training Set). The labels correspond to class membership information. For the extraction of the profiles, in several schemes can be used including but not limited to Decision Trees, Bayesian Classification, Support Vector Machines, etc.

• In unsupervised learning the focus is on the identification of similarities among a large set of observations. This procedure is usually referred as clustering. Formally, clustering is the process of grouping a set of physical or abstract objects into classes of similar objects. Within a cluster, objects tend to be similar to one another and dissimilar with objects of other clusters. For the extraction of the profiles, in several schemes can be used including but not limited to partitioning methods, hierarchical methods, density-based methods, etc.

Other machine learning techniques can be used for the profiles extraction such as the semi- supervised learning, etc.

In the fourth step 340, a list of profiles per user is created. In other words, for each user a list of user profiles that are enabled for this user is created.

The Profiling Engine can use the reduced dataset of step 320 with the previously mentioned techniques and similarities among the observations can be identified. Thus, groups of obser- vations can be extracted containing several information fields, including the location, time, accessed service (or accessed service type), duration of the accessed service, device type (e.g., high end smartphone, etc.), device status (e.g., battery level, etc.), charging level (i.e., amount of remaining credits). In the above mentioned exemplary case, a group of observations will then indicate previous specific behavior (i.e., accessed service and duration) of users under specific conditions (e.g., location, time, device type, device status, charging level, etc.) (Table 1). Since the users tend to have the same (or similar) behavior under specific conditions this will enable the accurate prediction of the user behavior if the conditions are the same or similar (see step 340 in FIG. 3). Table 1 presents a simple example where all different behaviors from all users are categorized in four different exemplary user profiles.

Table 1 : Example of the overall user profiles



Since this profiling scheme contains all possible combinations, all potential behaviors are being captured, and all users belong to at least one profile for the situation combinations that they have been in the past. Thus, user A when he is in a specific timeslot (e.g., 9:00-18:00 in Mondays) will have a certain behavioral profile, when he is located in a specific area (e.g., in the Stadium), under certain preconditions (e.g., based on battery status and/or charging status), but when he is in this area in a different timeslot or under other preconditions then his behavior is being captured by other behavioral profile related to the precondition values. This implies that each user' s behavior is being captured by a list of profiles, which capture the user behavior under all the potential combinations of location, time, and preconditions. As an example, in Table 2 we capture that user A will behave as indicated by profiles A or C depending on a number of parameters (e.g., Location, Day, Time, Battery and Charging status).

Thus, the Profiling Engine can provide an accurate prediction of the user behavior, based on his personal past behavior when he is under specific of combinations of location, time, and preconditions. Table 2 provides indicatively a list of potential user profiles of user A, for different locations and preconditions, (e.g. battery status and/or credit charging status). Note that in order to select the precise profile of a user, real time information may be needed to be collected. For example in Table 2, User A will behave as indicated by profile B instead of profile A, since all pre-conditions (e.g., certain location, day, time, and/or charging status) are the same, except from the battery level which has to be collected as information in real-time by a networking device. Similarly, each user will have a behavioral profile for every combination of the locations, time, and preconditions. This list of profiles is being stored locally either in the Profiling Engine, or in any other logically centralized entity that it is accessible from the other network elements (Step 350 in FIG. 3). From this point, for ease of presentation, we assume that this logically centralized entity resides together with the Profiling Engine and the profiling engine is responsible for the distribution of the profiles to the network devices.

However, other logically centralized entities could undertake the role of storing and distributing such information (e.g., Home subscriber Server in LTE/LTE-A network, Home Location Register in GSM, etc.).

Table 2: Two dimensional exemplification of the enabled user profiles for user A


As shown in Tables 1 and 2, the user profiles may be associated with parameter ranges that indicate when a specific user profile should be the active user profile. Preferably, these parameter ranges are non- overlapping such that for each input parameter (e.g. time of the day) only one user profile can be the active user profile.

In the fifth step 350, the user profiles are stored in a subscriber server. This may involve storing a table similar to Table 2 (with entries for a plurality of users). In other embodiments, more than one table may be used. For example, a first table can associate input information such as location, day, time and battery status with an associated profile, and a second table can associate a profile with an associated predicted behavior.

The first to fifth step 310-350 are carried out initially, before the system can be used for optimizing network performance. They can be repeated when further user network information is available, for example in regular intervals, such as once per day, week or month. Preferably, the collection can occur during hours of low network usage, e.g. during 3am to 4am.

The online phase 304 comprises a sixth and a seventh step 360, 370.

In the sixth step 360, the profiles are distributed to one or more network components. This can occur "on demand". For example, a network component can request enabled user profiles for

a specific user when the specific user is within reach of the network component. Thus, step 360 can be carried out repeatedly.

Once the profiles are built and stored they have to be distributed to the network components. When a network device (e.g., base station, user equipment, or mobility server) requires accurate prediction of the behavior of a specific user then this device will ask for (via sending the respective message) this from the Profiling Engine and latter will provide the corresponding list of profiles for the location and the time under consideration. Then, the network device will contact the user equipment and/or other network devices for gathering real-time infor-mation regarding the preconditions for the user behavior. These preconditions include but are not limited to battery level and charging status.

The task of collecting real-time information (e.g., battery level, and/or charging status) for accurate user behavior prediction from the user equipments and networking devices could take place at the networking device that plans to use the profile or the Profiling Engine itself, or both, depending on which device has direct access to such information.

In the seventh step 370, profile information can be integrated into radio resource management functions. For example, decisions about handovers can be based on an expected behavior that is associated with a certain user profile.

In the eighth step 380, an overall network performance is evaluated. For example, it can be evaluated whether decisions that were based on integrating profile information in the seventh step 370 were actually beneficial to network performance. The profiles can then be adjusted accordingly. For example, there may be users whose behavior is "erratic" in the sense that it turns out that it is not reliably possible to predict their behavior. In these cases, it may be preferable not to make network decisions based on an associated user profile. These users may be assigned a "null profile", which merely indicates that no profile-based decisions should be performed.

The eighth step 380, which may be considered part of the offline phase, can be carried out periodically, e.g. in regular time intervals.

The methods illustrated in FIGs 3 and 4 may be carried out by a profiling engine as shown in FIGs 5 to 10.

FIG. 4 shows an exemplary implementation of the process of acquiring and using the list of user profiles from a networking component from the profiling engine.

The network system shown in FIG. 4 comprises a UE A 402, a networking component A 404, a profiling engine 406 and a networking component B 408.

In this example and the following examples, the profiling engine may comprise the information acquisition unit, the profile generation unit, the profile determination unit and/or the configuration unit. For example, these four units can be realized on a processor of the profiling engine.

The network system of FIG. 4 provides an example of the process of acquiring the user profile from a networking component from the Profiling Engine (see step 360 in FIG. 3). When a networking component (for this example it is Networking Component A) requires accurate behavior prediction for a user equipment (for this example UE A), in a first step 410, it asks from the Profiling Engine to provide the list of user profiles for User Equipment A, for loca-tion X, and time Y.

In a second step 420, the profiling engine identifies a list of enabled user profiles for user equipment A (corresponding to a user A) based on some characteristics (e.g. time, location, etc.).

In this exemplary implementation, the Networking Component A stores enabled user profiles of UE A (preferably including associated predicted behavior information) and when it wants to use it, it contacts other networking components for accessing to dynamic real-time information (see steps 440, 450, 460, 470, and 480).

In a third step 430 the profiling engine provides a list of enabled profiles for user equipment A, for location X, and time Y.

In a fourth step 440 the networking component A requests a real-time dynamic information for user equipment A.

In a fifth step 450 the networking component B provides a real-time dynamic information for user equipment A.

In a sixth step 460 the networking component A requests for UE real-time information (e.g. battery level).

In a seventh step 470 the UE A provides UE real-time information (e.g. battery level).

In an eighth step 480 the networking component A stores and uses a profile list.

In this exemplary implementation Networking Component A communicates with Networking Component B for accessing real time information related to UE A. Networking component A may communicate with UE A for accessing UE-related real time information which is available only in the UE A (e.g., battery level, etc.). Then the networking component A may use the profile (Step 490 in FIG. 4, see also step 370 in FIG. 3). The processing of obtaining, storing, and using the user profiles may or may not be interconnected in terms of time and thus the Networking Component A may obtain the list of enabled user profiles of UE A, store it, and use it in a totally unrelated timeframe.

FIG. 5 illustrates an exemplary implementation of the process of acquiring and using the list of user profiles from a networking component from the Profiling Engine.

The system shown in FIG. 5 comprises a UEA 502, a networking component A 504, a profiling engine 506 and a networking component B 508.

In a first step 510, the networking component A requests for a UE real-time information (e.g. battery level).

In a second step 520, the UE A provides UE real-time information (e.g. battery level).

In a third step 530, the networking component A requests a list of profiles for user equipment A, for location X, and time Y, and for UE real-time information (e.g. battery level).

In a fourth step 540, the networking component B provides real-time dynamic information for user equipment A (e.g. charging status).

In a fifth step 550, the profiling engine requests real-time dynamic information for user equipment A (e.g. charging status).

In a sixth step 560, the profiling engine identifies a list of profiles for user equipment A based on characteristics (e.g. time, location, etc.) and real-time information (e.g. charging status and/or battery level).

In a seventh step 570, the profiling engine provides a user profile, in particular the active user profile, to the networking component A 504.

In an eight step 580, the networking component A 504 stores and uses the UE profile, e.g. to make a network function decision based on the active user profile.

FIG. 6 shows an exemplary implementation of the process of acquiring and using the list of user profiles from a networking component from the profiling engine.

The system shown in FIG. 6 comprises a UEA 602, a networking component A 604, a profiling engine 606 and a networking component B 608.

In a first step 610, the networking component A requests a list or profiles for user equipment A, for location X, and time Y.

In a second step 620, the profiling engine requests real-time dynamic information for user equipment A.

In a third step 630, the networking component B provides real-time dynamic information for user equipment A.

In a fourth step 640, the profiling engine identifies a list of possible profiles for user equipment A based on characteristics (e.g. time, location, etc.) and real-time information (e.g. charging status).

In a fifth step 650, the profiling engine provides the networking component 604 with a list of candidate user profiles for UE A.

In a sixth step 660, the networking component A requests for a UE real-time information (e.g. battery level).

In a seventh step 670, the UE A provides UE real-time information (e.g. battery level) to the networking component A 604.

In an eighth step 680 the networking component A stores and uses a list of profiles for UE A 602. In particular, the networking component A 604 may determine an active user profile from its list of candidate user profiles based on the real-time information provided from the UE A 602.

When the user equipment A (UE A) performs its initial connection to the network (e.g., via sending an attach request) to an eNB, the eNB communicates with the Mobility Management Entity (MME). The MME communicates with the Profiling Engine to obtain the list of the UE profiles. In the described exemplary implementation the Profiling Engine maintains the list of profiles for all the UEs in the network; however it could be any type of logically centralized entity that could provide the list of user profile (e.g., the HSS) . The MME then retrieves the list of profiles that fits to the characteristics of the attach request (e.g., location, time, etc.) and provides it to the eNB that has requested it. Then, the eNB stores the list of profiles locally. This process is summarized in FIG. 7.

FIG. 7 shows an exemplary implementation of the process of acquiring the list of user profiles from eNB from the Profiling Engine in an LTE/LTE-A network.

The system shown in FIG. 7 comprises a UEA 702, an eNB 704, a profiling engine HSS 706 and an MME 708.

In a first step 710, the UE A attaches a request from the UE A.

In a second step 720, the eNB forwards the attach request from the UE A. In response, in a third step 730, the MME requests a list of enabled profiles for UE A from the profiling en-gine/HSS 706.

In a fifth step 750, the MME 708 evaluates and stores the list of enabled profiles for UE A based on some current characteristics (e.g. time, location, etc.). Based on this current characteristic, the MME 708 can also determine a set of candidate user profiles for UE A.

In a sixth step 760, the profiling engine/HSS 706 provides a list of profiles for UE A. This can be the complete list of enabled user profiles for UE A, or it can be the (reduced) set of candidate user profiles that are selected based on the current characteristics. This has the advantage that the eNB 704 is provided with a smaller list of relevant user profiles. The eNB 704 can then use further current characteristics to determine an active user profile from the list of candidate user profiles.

FIG. 8 shows an exemplary implementation of a handover process using a list of user profiles in an LTE/LTE-A network.

The network system shown in FIG. 8 comprises a UE A 802, an eNB 804 and an online charging server 809.

When the UE A 802 moves to another eNB and performs a handover, with its handover re-quest it provides real time information related to its status (e.g., battery level, etc.). This real time information can be collected before the execution of a control function like a handover. To this end, the UE A 802 or other devices may periodically send the real time information. In other embodiments, the real time information may be sent asynchronously, e.g. whenever there is a significant deviation compared to the previously reported value.

In detail, the procedure can be carried out as follows:

In a first step 810, the UE A 802 submits a handover request to the eNB 804 that it is currently connected to. In an exemplary embodiment, the handover request may comprise some realtime information from the UE A 802, e.g. a battery level of the UE A 802.

In a second step 820, the eNB 804 requests the charging status of the UE A 802. Charging status here refers not to a battery charging level, but rather to a credit level e.g. of a prepaid account. In an LTE network, the charging status can be requested from the Online Charging Server of the Policy and Charging Control. In other networks, the charging status may be provided from other nodes of the network.

In other embodiments, instead of communicating with an Online Charging Server the eNB 804 may communicate with other network components for acquiring other real-time information.

In a third step 830, the online charging server 809 provides the charging status of the UE A 802 to the eNB 804. The eNB 804 uses it, as well as any other real time information coming from the UE A 804 and/or other network components to determine, in fourth step 840, an active user profile of the UE A that is specific to conditions/constraints of time, location, battery level and/or charging, of the UE A 804. Choosing the active user profile can be based on a list of candidate user profiles for UE A 802 that has previously been stored at the eNB 804.

The information about the active user profile enables the eNB 804 to predict whether the UE A 802 is likely to move, or what type of service requests it will perform, the duration of each service access, etc. This information, combined with measurements for the link quality among the UE A 802 and candidate eNBs for handover, will enable the eNB 804 where the UE A 802 resides to decide to which of the neighboring eNBs the UE A 802 should be handed over to.

More specifically, using the active user profile information the eNB 804 can decide to which layer UE A 802 should be handed over. Layer here means macro-cell, micro-cell, small-cell, etc. Other layers could be included in such layered structure as well, depending on the functionalities of the base station. Metrics that can be used to characterize the quality of the link between a UE and an eNB include but are not limited to the Received Signal Strength (RSS), Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), etc. Similarly, the eNB 804 or any other device in the network that has access to such information, could decide to handover the UE A 802 to any other types of interworking networks, including GSM, UMTS, and WiFi networks, based on the predicted behavior that is associated with the active user profile.

FIG. 9 shows an exemplary implementation of the process of acquiring the list of user profiles from UE from the Profiling Engine in an LTE/LTE-A network for a Cell Selection/-Reselection and/or a Call Admission Control processes.

The system shown in FIG. 9 comprises a UEA 902, an eNB 904, a profiling engine 906, an MME 908 and an online charging server 909.

Other exemplary uses of the multi-dimensional user profiles extracted from the Profiling Engine could be its application in Cell Selection/Reselection or Call Admission Control processes in an LTE/LTE-A network. The same mechanisms could be applied in other types of cellu-lar networks such as GSM, UMTS, etc. or interworking networks of more than one type e.g., interworking GSM, UMTS, LTE/LTE-A, and WiFi networks, etc.

When the UE A 902 performs a Tracking Area Update process it receives from the MME 908 (through the eNB 904) a list of candidate user profiles for the respective location, day, etc. Specifically, when the MME 908 receives a TAU request from the UE A 902, as in the case of the handover, it will retrieve the list of user profiles from the Profiling Engine 906. Additionally, as shown in FIG. 9, the MME 908 may interact with the Online Charging Server 909 (Online Charging Server of the Policy and Charging Control) to retrieve the charging status of the UE A 902. Then, MME 908 with the TAU response will also provide to the UE A 902 his list of profiles for the respective location. When the UE A 902 attempts to perform a Cell Selection/Reselection process or a Call Admission request then it will attempt to connect to the eNB or the cell which is most suitable according to the predicted behavior associated with its active user profile.

For example, if the predicated behavior associated with its active user profile is that the UE A 902 will be moving with high speed, it will camp in the case of the Cell Selection/Reselection, or attempt to connect in the case of the Call Admission to a macro BS. Similarly if the UE A 902 is predicted to be static or slowly moving then it will attempt to connect to the most suitable micro-BS, small-cell. Similarly to the handover case, the prediction of the user behavior is combined with metrics that are used to characterize the quality of the link between the UE A 902 and the eNB 904 and include but are not limited to the Received Signal Strength (RSS), Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), etc. Similarly, the UE A for the Cell Selection/Reselection or Call Admission Con-trol processes may obtain the list of user profiles profiles periodically, or on a demand basis linked to other processes (e.g., PLMN selection, attach request, etc.).

For the Call Admission Control process, when the UE A performs an attach request the eNB automatically obtains the list of enabled user profiles or the (reduced) set of candidate user profiles for UE A and stores it locally as illustrated in FIG. 9. Based on the determined active user profile the eNB 904 can redirect the UE A 902 to other eNBs 904 more suitable for his predicted behavior.

In detail, the method can be carried out as follows:

In a first step 910, the UE A 902 sends a TAU request to eNB 902.

In a second step 920, the eNB 904 forwards the TAU request from the UE A 902 to the MME 908.

In a third step 930, the MME 908 requests a list of enabled user profiles for the UE A 902 from the profiling engine 906.

In a fourth step 940, the profiling engine 906 provides a list of enabled user profiles for the UE A 902. For this purpose, the profiling engine 906 may request a list of enabled user profiles for UE A 902 from a database.

In a fifth step 950, the MME 908 requests the charging status of the UE A 909 from the online charging server 960.

In a sixth step 960, the online charging server 909 provides the charging status of the UE A 902 to the MME 908.

In a seventh step 970, the MME 908 evaluates and stores a list of candidate user profiles for UE A 902 based on the dynamic characteristics (e.g. location etc.). In particular, the MME 908 may choose from the list of enabled user profiles for UE A, those user profiles as candidate user profiles that match the dynamic characteristics. Furthermore, the MME 908 may choose an active user profile from the candidate user profiles. In other embodiments, the MME 908 may directly choose an active user profile from the enabled user profiles.

In an eighth step 980, the MME 908 sends a TAU response for the UE A to the eNB 904, wherein the TAU response is based on the active user profile. The TAU response may also comprise an indication of the active user profile and/or information about the predicted behavior associated with the active user profile.

In a ninth step 990, the eNB 904 forwards the TAU response for the UE A 902 to the UE A 902.

FIG. 10 shows an exemplary implementation of the process of Call Admission Control using the list of user profiles in an LTE/LTE-A network.

The system shown in FIG. 10 comprises a UE A 1002, an eNB 1004, a profiling engine 1006, an MME 1008 and an online charging server 1009.

In a first step 1010, the UE A 1002 sends an attach request to the eNB 1010.

In a second step 1020, the eNB 1004 forwards the attach request from the UE A 1002 to the MME 1008.

In a third step 1030, the MME 1008 requests a list of user profiles for UE A 1002.

In a fourth step 1040, the profiling engine 1006 provides the list of user profile for UE A 1002 to the MME 1008. For example, the list of user profiles can be a list of enabled user profiles that are enabled for UE A 1002. Alternatively, the MME 1008 can send with the request one or more current characteristics of the UE A 1002, and the profiling engine 1006 can respond to the request with a list of candidate user profile that correspond to the one or more current characteristics.

In a fifth step 1050, the MME 1008 evaluates the list of user profiles for UE A 1002 based on dynamic characteristics (e.g. location, etc.). For example, the MME 1008 can determine from the list of enabled user profiles that were provided by the profiling engine 1006 a list of can-didate user profiles, based on the dynamic characteristics. Alternatively, the MME 1008 can determine an active user profile based on the dynamic characteristics.

In a sixth step 1060, the MME 1008 provides a list of candidate user profiles for UE A 1002 to the eNB 1004. Alternatively, as indicated above, if the MME 1008 has already determined an active user profile, it may provide an indication of the active user profile to the eNB 1004.

In a seventh step 1070, the eNB 1008 requests the charging status of the UE A 1002 from the Online Charging Server 1009.

In an eighth step 1080, the Online Charging Server 1009 provides the charging status of the UE A 1002 to the eNB 1004.

If it is already determined which user profile is the active user profile, there is no need to request the charging status. Therefore, the seventh and eighth step 1070, 1080 may be skipped if the MME 1008 has provided only one candidate user profile, i.e., the active user profile.

In a ninth step 1090, the eNB 1004 identifies and uses an active user profile for the call admission control. For example, the eNB 1004 can select an active user profile from the list of candidate user profiles that was provided by the MME 1008.

In a tenth step 1100, the eNB 1004 sends an attach response for UE A 1002 to the UE A 1002. The attach response can comprise the selected active user profile.

In the systems shown in FIGs. 4 to 10, the profiling engines 406, 506, 606, 706, 806, 906, 1006 act as profile determination units. The profiling engines 406, 506, 606, 706, 806, 906, 1006 may furthermore also comprise functionalities of a profile generation unit and/or an information acquisition unit. Alternatively, a profile generation unit and/or an information acquisition unit may be provided separately from the profiling engines 406, 506, 606, 706, 806, 906, 1006.

The networking components A 404, 504, 604, the networking components B 408, 508, 608, the user equipments 402, 502, 602, 702, 802, 902, 1002, the eNBs 704, 804, 904, 1004, and the MMEs 708, 908 and 1008 can be configured to configure themselves or other components in the network based on an active user profile. Thus, they act as configuration units.

As outlined above, embodiments of the invention relate to methods of using a UE profile to assist the network administration, including but not limited to handover optimization, cell selection and re-selection, call admission control, and using big data analytics. The methods can be used, for example, in massive mobile broadband scenarios, in machine type communication scenarios etc.

Embodiments of the profiling method may comprise:

• Mechanisms for gathering the required information from various network points (such as user equipments, base stations, mobility servers, databases etc.).

• Methods for analyzing the inputs and identifying the most relevant/representative parameters from the all of the gathered ones for the extraction of the user profiles.

• Schemes to extract the user profiles of the users based on their past behavior. These schemes may be based in supervised or unsupervised learning schemes, which will identify the user behavior in relation to several parameters including but not limited to date, time, location, device status and charging status.

• Schemes to distribute the extracted user profiles to network components including but not limited to databases, Access Stratum and Non Access Stratum control entities, and end devices, so as to be used for network administration.

Some benefits expected from some of the profiling systems and profiling methods introduced above include accurate user behavior prediction by taking into consideration previous user behaviors. The user behavioral profile can be built by using the user previous behaviors, thus it is accurate under specific preconditions. Additionally, by combining the user behavioral profile with real time information the user behavioral prediction becomes very accurate. For example, the proposed mechanisms infer that when the user is located in location A, he is highly mobile, and if his cell phone is fully charged then he will watch streaming videos, whereas when the same user is in location A, but his cell phone is low on battery he will only perform short calls (his mobility is not affected). This enables the network to associate the user with the respective more suitable eNB or even access network. This significantly benefits the system throughput, since the predicted user requirements are being handled by the most suitable network/technology.

Furthermore, in some embodiments of the invention a significant decrease in the number of handovers is expected, since the users are associated to the network according to their predicted mobility and thus low moving users may be associated to small-cells and high moving users to macro-BS. The proper placement of the users in the network, and the reduction of the number of the handovers additionally reduces the experienced latency from the user as well.

Further embodiments of the present invention may include:

1. A method to generate user profiles, that comprises one or more systems that capture and analyze user activity in the network, and distribute them into network components.

2. The outcome of method 1 is combined by the corresponding network components with real time information related to the current status of a UE and used by control functions to optimize the performance of the network.

3. The method of 1, further comprising collecting information from the corresponding points. This includes, but is not limited to:

i. Information from several network components including but not limited to channel conditions, charging status information, serving access node, mobility information, etc.

ii. Service specific information from the UEs including but not limited to accessed services, duration of sessions, date, time, etc.

iii. Device status information including but not limited to battery level, cpu load, memory load, etc.

iv. Information related to the user, including but not limited to age, income, contract type, user equipment type (e.g., smartphone, low end cell phone, etc.), user equipment information (e.g., radio access technology interfaces, screen size, CPU, memory, etc.), etc.

4. The method of 1, further comprising the analysis of the collected data and the identification of the most important/relevant parameters by using one mechanism of the following, but not limited to Information Gain, Frequency -based feature selection, Chi- square, etc.

5. The method of 1, further comprising the use of the output from method 4 and after using data mining techniques including but not limited to supervised and unsupervised learning produces user profiles based on characteristics including but not limited to:

i. Date,

ii. Time,

iii. Location,

iv. Device status (e.g., battery level), and

v. Charging status.

6. The method of 1, further comprising as output the predicted behavior of a user in terms of but not limited to, mobility level, access to type of services (e.g., voice, data), service access duration.

7. The method 1, further comprising the grouping of the produced profiles of method 5 on a per user basis.

8. The distribution of the produced grouped profiles of method 6 to network components including but not limited to databases, AS and non AS control entities and the end devices.

9. The real time acquisition by network components of UE related information includes but is not limited to battery level and charging status.

10. The selection of a specific user profile from the grouped profiles of step 6 based on the acquired real time information of method 8.

11. The combination of the information of method 8 with the produced user profiles by network control functions including but not limited to call admission control, handover, and broadcasting of cell (re)selection information to improve the performance of these activities.

The foregoing descriptions are only implementation manners of the present invention, the scope of the present invention is not limited to this. Any variations or replacements can be easily made through person skilled in the art. Therefore, the protection scope of the present invention should be subject to the protection scope of the attached claims.