WO2016029936A1

WO2016029936A1 - Technique for analytical identification of spatio-temporal relationships in a mobile communications network

Info

Publication number: WO2016029936A1
Application number: PCT/EP2014/068077
Authority: WO
Inventors: László Toka; Péter Hága; Péter MÁTRAY
Original assignee: Telefonaktiebolaget L M Ericsson (Publ)
Priority date: 2014-08-26
Filing date: 2014-08-26
Publication date: 2016-03-03

Abstract

A technique for identification of spatio-temporal relationships in a mobile communications network comprising a plurality of subscribers with associated mobile terminals is disclosed. In a method aspect, the method comprises building, per each subscriber, a mobility profile by tracking the mobile terminals. Further, there is filtering, from the mobility profiles, at least one first irregular location visited by a first subscriber and at least one second irregular location visited by a second subscriber. Then, there is determining a match between a first subscriber and a second subscriber when the first and second subscriber visit one common irregular location at the same time, and identifying, based on the number of matches, a spatio-temporal relationship between the first and second subscribers.

Description

Technique for analytical identification of spatio-temporal relationships in a mobile communications network

Technical Field

The present disclosure generally relates to analytical identification of spatio-temporal relationships in a mobile communications network, in particular to analytical identification of spatio-temporal relationships in a mobile communications network comprising a plurality of subscribers with associated mobile terminals. The techniques of the present disclosure may be embodied in methods and/or apparatuses.

Background

Many mobile network operators strive to estimate their customers' value individually. These estimates of customer value are useful in a wide range of applications from user experience management to intelligent service differentiation in situations when networking resources become scarce. The outcome of customer value analysis usually supports business decision making, the optimization of resource allocation, and in turn helps to increase profitability.

There has been an approach to use customer importance score in the general area of Customer Relationship Management (CRM) to help the prioritization and allocation of resources. Parameters are applied, like the length of relationship, the breadth of relationship, the revenue earned from the customer in the last fiscal year and the future expected revenue.

There has been another approach involving a framework to estimate the customer value of mobile network operator's subscribers. A neural network based data analysis method and retention-based interactions with the existing customers based on a churn hazard function are used.

Finally, there has been another approach in which a mobile network operator can determine customer importance by social influence, i.e. the ability of the customer to drive the opinion of their social connections. The system extracts online social interactions from network data traffic and estimates customer importance based on social significance. There have been certain problems with existing solutions not hitherto realized, as will be detailed below.

Hitherto, the calculation of customer value took into account the influence of the subscribers on their fellow subscribers. In order to estimate a subscriber's influence, it had been suggested to rely on the analysis of Call Detail Records (CDRs) and the packet traces e.g. of online social networking applications. Nevertheless, those important social connections lacking using mobile voice or data channels for communication between the two subscribers are never discovered or are underestimated.

Summary

Accordingly, there is a need for an implementation of a scheme that avoids one or more of the problems discussed above, or other related problems. In particular, a solution is still needed to discover these "offline" social connections (as a non-limiting example of the spatio-temporal relationships discussed herein below).

In a first aspect, there is provided a method for analytical identification of spatio- temporal relationships in a mobile communications network comprising a plurality of subscribers with associated mobile terminals, the method comprising building, per each subscriber, a mobility profile by tracking the mobile terminals; filtering, from the mobility profiles, at least one first irregular location visited by a first subscriber and at least one second irregular location visited by a second subscriber; determining a match between a first subscriber and a second subscriber when the first and second subscriber visit one common irregular location at the same time; and identifying, based on the number of matches, a spatio-temporal relationship between the first and second subscribers. In this way, an analytical framework for the mobile network operator to determine connections between customers that are hidden, i.e., cannot be found by traditional analysis of CDRs or DPI methods, is provided.

In a first refinement of the first aspect, the building step may comprise repeatedly collecting time-stamped location information for each mobile terminal. In this case, the collecting step may comprises querying one of a Mobile Switching Center, MSC, node and a Visitor Location Register, VLR, node in a second generation, 2G, network; a Serving GPRS Support Node, SGSN, and Radio Network Controller, RNC, in a third generation, 3G, network; and a Mobility Management Entity, MME, in a Long Term Evolution, LTE, network or fourth generation, 4G, network. Furthermore, when the tracked mobile terminal of a subscriber is in a ready state, the location information may comprise geographical coordinates of a cell of the respective mobile terminal; alternatively, when the tracked mobile terminal of a subscriber is in one of a stand-by state and an idle state, the collecting step may further comprise tracking the respective mobile terminal on at least one of the location level; the Routing level; and the Tracked Area level. In this way, spatio-temporal matches can be determined with greater reliability.

In a second refinement of the first aspect, the filtering step may further comprise categorizing the visited locations into categories. In this case, the categories may comprise two or more of characteristic locations; transit locations; and the irregular locations. In addition, the categorizing may be performed based on at least one of a total amount of time spent at a location during an observation period; and an average duration of visits at the location during the observation period. Still further, an irregular location may be determined if: and wherein δ, is a predefined upper bound parameter for subscriber i; ε, is a predefined lower bound on the average duration at the irregular location; t' is the total amount of time of subscriber i spent at a location I during an observation period; and d,' is the average duration of visits of subscriber i at the location I during the observation period. In the latter case, the observation period may be selected based on a frequency of occurrences of the irregular locations, and wherein the frequency of occurrences is based on δ,.

Accordingly, the occurrence of the irregular locations visited by the subscribers can be determined analytically. In other words, the irregular locations are distinguished from the mobility profiles with a clear separation from the characteristic and transit locations.

In a third refinement (based on the second refinement) of the first aspect, the determining step may further comprise correlating the visits of a subscriber i and a subscriber j according to the following measure

wherein Aj'(t) represents a spatio-temporal presence function of subscriber i at the irregular location I, A_j'(t)represents a spatio-temporal presence function of subscriber j at the irregular location I, and Ay'OO represents the overlapping spatio-temporal presence of subscribers i, j. In that case, the spatio-temporal presence functions may be defined in the following way Aj'(t) = 1 or A_j'(t) = 1 during the time when subscriber i or j is at location I, and 0 otherwise. In addition, the identifying step further comprises thresholding the correlation measure such that: c_i(j>c*, wherein the threshold c* is calculated with a probability theory approach in order to rule out the effects of randomness. In this way, the size of the operator's network, the number of subscribers and their mobility profiles can be taken into account.

In a fourth refinement (involving the second and third aspects) of the first aspect, the categorizing may be performed based on at least one of a route taken to an irregular location during the observation period; and a movement speed on the route during the observation period. In this way, the determination of the spatio-temporal relationship can be confirmed.

In a second aspect, there is provided method for analytical identification of temporal relationships in a mobile communications network comprising a plurality of

subscribers with associated mobile terminals, the method comprising building, per each subscriber, an application traffic usage profile from application usages via the mobile terminals; determining, from the application traffic usage profiles, a match between a first subscriber and a second subscriber when the first and second subscriber use the same application at the same time; and identifying, based on the number of matches, a temporal relationship between the first and second

subscribers. Also in this way, an analytical framework for the mobile network operator to determine connections between customers that are hidden, i.e., cannot be found by traditional analysis of CDRs or DPI methods, is provided.

In a first refinement of the second aspect, the application traffic usage profile for a particular subscriber may comprise time windows during which the particular subscriber is active in an over-the-top, OTT, mobile application in which the particular subscriber interacts with at least one other subscriber; and the OTT mobile application is such that the particular subscriber interacts with the at least one other subscriber in real-time. In addition, the application may be one of a Voice over Internet Protocol, VoIP, application; and an online multiplayer gaming application. In the latter case, the VoIP application may be one of Skype™; Viber™; Whatsapp™; and Line™.

In a third aspect, there is provided a computer program product comprising program code portions for performing the method of the first aspect when the computer program product is executed on one or more computing devices. The computer program product may be stored on a computer readable recording medium. In a fourth aspect, there is provided an apparatus for analytical identification of spatio-temporal relationships in a mobile communications network comprising a plurality of subscribers with associated mobile terminals, the apparatus comprising a processor configured to build, per each subscriber, a mobility profile by tracking the mobile terminals; filter, from the mobility profiles, at least one first irregular location visited by a first subscriber and at least one second irregular location visited by a second subscriber; determine a match between a first subscriber and a second subscriber when the first and second subscriber visit one common irregular location at the same time; and identify, based on the number of matches, a spatio-temporal relationship between the first and second subscribers.

In a fifth aspect, there is provided an apparatus for analytical identification of spatio- temporal relationships in a mobile communications network comprising a plurality of subscribers with associated mobile terminals, the apparatus comprising a processor configured to build, per each subscriber, an application traffic usage profile from application usages via the mobile terminals; determine, from the application traffic usage profiles, a match between a first subscriber and a second subscriber when the first and second subscriber use the same application at the same time; and identify, based on the number of matches, a spatio-temporal relationship between the first and second subscribers.

Still further, it is to be noted that the method aspects may also be embodied on the apparatus of the fourth and fifth aspects comprising at least one processor and/or appropriate means for carrying out any one of the method steps.

Brief Description of the Drawings

The embodiments of the technique presented herein are described herein below with reference to the accompanying drawings, in which:

Fig. 1 shows a principle of communications network and components involved in which embodiments of the present disclosure can be performed;

Fig. 2 shows components comprised in an exemplary device embodiment realized in the form of an apparatus (which may reside e.g. in a data collector, a profiler and an analyzer); - -

Fig. 3A shows a first method embodiment which also reflects the interaction between the components of the apparatus embodiment; and

Fig. 3B shows a second method embodiment which also reflects the interaction between the components of the apparatus embodiment.

Detailed Description

In the following description, for purposes of explanation and not limitation, specific details are set forth (such as particular signalling steps) in order to provide a thorough understanding of the technique presented herein. It will be apparent to one skilled in the art that the present technique may be practiced in other embodiments that depart from these specific details. For example, the embodiments will primarily be described in the context of 3^rd generation (3G) or 4^th generation/long term evolution (4G/LTE); however, this does not rule out the use of the present technique in connection with (future) technologies consistent with 3G or 4G/LTE, be it a wire- bound communications network or a wireless communications network.

Moreover, those skilled in the art will appreciate that the services, functions and steps explained herein may be implemented using software functioning in conjunction with a programmed microprocessor, or using an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a field programmable gate array (FPGA) or general purpose computer. It will also be appreciated that while the following embodiments are described in the context of methods and devices, the technique presented herein may also be embodied in a computer program product as well as in a system comprising a computer processor and a memory coupled to the processor, wherein the memory is encoded with one or more programs that execute the services, functions and steps disclosed herein.

Without loss of generality, in the proposed solution, to this end the proposed system can build individual mobility profiles for each and every subscriber. Based on the mobility profiles, the proposed method may distil the irregular locations where subscribers appear. Then the method looks for spatio-temporal matches between subscribers when they are at the same location, irregular to both of them, at the same time. The more of such matches are found for the same subscribers, the higher the confidence is that a spatio-temporal relationship (e.g. a social connection) exists between the two subscribers. - -

Without loss of generality, alternatively or simultaneously, the proposed system can also build individual application traffic usage profiles for each and every subscriber, i.e., the time windows during which the user is active in an over-the-top (OTT) mobile application in which the user interacts with other user(s). These OTT applications might include but not limited to VoIP applications, e.g., Skype™, Viber™,

WhatsApp™, Line™, and online multiplayer gaming applications. Based on these application usage profiles, the method looks for temporal matches between subscribers when they are using the same application at the same time. The more of such matches are found for the same subscriber pair, the higher the confidence is that a spatio-temporal relationship (e.g. a social connection) exists between the two subscribers.

Fig. 1 shows a principle of communications network 200 and components (data collector 2001, profiler 2002 and analyzer 2003) involved in which embodiments of the present disclosure can be performed. The communications network 200 may comprise a cellular network 201 comprising cells, wherein one or more mobile terminals (MTs) 202 may camp on a cell. It is to be noted that Fig. 1 shows the data collector 2001, the profiler 2002 and the analyzer 2003 are situated in some form of controlling entity (such as a Radio Network Controller (RNC)) or backbone; however, this does not preclude that that the data collector 2001, the profiler 2002 and/or the analyzer 2003 are situated in some other suitable entity, such as a Base Station Controller (BSC) (not shown).

Fig. 2 shows components comprised in an exemplary device embodiment realized in the form of the data collector 2001, the profiler 2002 and/or the analyzer 2003. As shown in Fig. 2, the data collector 2001 comprises a core functionality (e.g., one or more of a Central Processing Unit (CPU), dedicated circuitry and/or a software module) 20011, an optional memory (and/or database) 20012, an optional transmitter 20013 and an optional receiver 20014. Moreover, the data collector 2001 comprises a builder 20015, an optional collector 20016, an optional querier 20017 and an optional tracker 20018.

Moreover, the profiler 2002 comprises a core functionality (e.g., one or more of a Central Processing Unit (CPU), dedicated circuitry and/or a software module) 20021, an optional memory (and/or database) 20022, an optional transmitter 20023 and an optional receiver 20024. Moreover, the profiler 2002 comprises a filter 20025 and an optional categorizer 20026. - -

Finally, the analyzer 2003 comprises a core functionality (e.g., one or more of a Central Processing Unit (CPU), dedicated circuitry and/or a software module) 20031, an optional memory (and/or database) 20032, an optional transmitter 20033 and an optional receiver 20034. Moreover, the analyzer 2003 comprises a determiner 20035, and identifier 20036, an optional correlator 20037 and an optional thresholder 20038.

In the following paragraphs, assume that x = 1, 2 or 3 (the data collector 2001, the profiler 2002 or the analyzer 2003). As partly indicated by the dashed extensions of the functional blocks of the CPUs 200x1, the builder 20015, the collector 20016, the querier 20017 and the tracker 20018 (of the data collector 2001), the filter 20025 and the categorizer 20026 (of the profiler 2002) and the determiner 20035, the identifier 20036, the correlator 20037 and the thresholder 20038 (of the analyzer 2003) as well as the memory 200x1, the transmitter 200x3 and the receiver 200x4 may at least partially be functionalities running on the CPUs 200x2, or may alternatively be separate functional entities or means controlled by the CPUs 200x1 and supplying the same with information. The transmitter and receiver components 200x3, 200x4 may be realized to comprise suitable interfaces and/or suitable signal generation and evaluation functions.

The CPUs 200x1 may be configured, for example, using software residing in the memories 200x2, to process various data inputs and to control the functions of the memories 200x2, the transmitter 200x3 and the receiver 200x3 (as well as of the builder 20015, the collector 20016, the querier 20017 and the tracker 20018 (of the data collector 2001), the filter 20025 and the categorizer 20026 (of the profiler 2002) and the determiner 20035, the identifier 20036, the correlator 20037 and the thresholder 20038 (of the analyzer 2003)). The memory 200x2 may serve for storing program code for carrying out the methods according to the aspects disclosed herein, when executed by the CPU 200x1.

It is to be noted that the transmitter 200x3 and the receiver 200x4 may be provided as an integral transceiver, as is indicated in Fig. 2. It is further to be noted that the transmitters/receivers 200x3, 200x4 may be implemented as physical transmitters/receivers for transceiving via an air interface or a wired connection, as routing/forwarding entities/interfaces between network elements, as functionalities for writing/reading information into/from a given memory area or as any suitable combination of the above. At least one of the builder 20015, the collector 20016, the querier 20017 and the tracker 20018 (of the data collector 2001), the filter 20025 and the categorizer 20026 (of the profiler 2002) and the determiner 20035, the identifier 20036, the correlator 20037 and the thresholder 20038 (of the analyzer 2003), or the respective functionalities, may also be implemented as a chipset, module or subassembly.

Fig. 3A shows a first method embodiment which also reflects the interaction between the components of the device embodiment, while Fig. 3B shows a second method embodiment which also reflects the interaction between the components of the device embodiment. In the signalling diagram of Fig. 3, time aspects between signalling are reflected in the vertical arrangement of the signalling sequence as well as in the sequence numbers. It is to be noted that the time aspects indicated in Fig. 3 do not necessarily restrict any one of the method steps shown to the step sequence outlined in Fig. 3. This applies in particular to method steps that are functionally disjunctive with each other.

In a first step Sla, the builder 20015 of the data collector 2001 performs building, per each subscriber, a mobility profile by tracking the mobile terminals 202. To this end, in an optional step Sla-1, the collector 20016 of the data collector 2001 performs repeatedly collecting time-stamped location information for each mobile terminal 202. It is to be noted that the repeatedly collecting may be performed periodically (e.g. every minute) for each and every subscriber in the system 200.

In the latter case, in an optional step Sla-1-1, the querier 20017 of the data collector 2001 performs querying, i) for a a second generation (2G) network, a Mobile Switching Center (MSC) node and a Visitor Location Register (VLR) node in, ii) for a in a third generation (3G) network, a Serving GPRS Support Node (SGSN) and Radio Network Controller (RNC) or iii) for a Long Term Evolution (LTE) network or fourth generation (4G) network, a Mobility Management Entity (MME). Accordingly, the system of the present disclosure works in modern systems as well as in legacy systems.

On the one hand, when a tracked mobile terminal 202 of a subscriber is in a ready state, the location information optionally comprises geographical coordinates of a cell of the respective mobile terminal 202. On the other hand, even when the tracked mobile terminal 202 of a subscriber is in a stand-by state or an idle state, in an optional step Sla-1-2, the tracker 20018 of the data collector 2001 performs tracking the respective mobile terminal 202 on the location level, the Routing level, and/or the Tracked Area level. In the latter case, the level may depend on the technology (e.g., the SGSN in a 3G network may be used when the terminal is in the stand-by mode). - -

Still further, although not all subscriber appearances may be grasped precisely (e.g., when in idle mode, the location can be limited only to the routing area), but when observing for a long period (e.g., for year), real connections will be eventually discoverable by collecting cell-level location data.

In the above-described cases, the data collector 2001 can make an assumption that the subscriber stayed in the same cell, if the last activity and the subsequent activity were recorded from the same cell and the elapsed time in between was comparatively short (e.g. less than an hour). For this assumption, a low upper time bound can be applied on the inactive time period in order to be strict in terms of positioning since the system has a higher chance to observe joint appearances in the long run (to be described herein below). If the location cannot be determined or assumed on the cell level, then for that time period the mobility record is ignored in the analytics performed by the analyzer 2003 described below.

Naturally, the memory 20012 may also store the obtained records (e.g. time, geographic location or speed of movement) of individual subscribers for a longer period.

Then, in step S2a, the filter 20025 of the profiler 2002 performs filtering, from the mobility profiles, at least one first irregular location visited by a first subscriber and at least one second irregular location visited by a second subscriber. In this case, in an optional step S2a-1, the categorizer 20026 of the profiler 2002 performs categorizing the visited locations into categories.

In this context, the locations visited by the subscriber may be categorized, and subsequently, certain locations for the subscribers may be selected based on the aforementioned records. The locations can represent individual mobile cells 201, or groups of those, e.g., cells belonging to the same base station (not shown). The correct choice of geographical granularity greatly affects the quality of the overall output of the system. On the one hand, the finer the granularity, the higher the chance that the system identifies what is considered a single place in real life as separate locations. In that case, however, a so-called Ping-Pong phenomenon can be tackled by the system seamlessly. On the other hand, a coarse granularity may mask the important difference between real life places, e.g., two points of interests if geographically separable should be handled as different locations in the same city district. The person skilled in the art may adjust the granularity between those two extremes. The above-described categories may comprise two or more of characteristic locations, transit locations and the above-mentioned irregular locations. In addition, the categorizing may be performed based on at least one of a total amount of time (e.g., measured in minutes) spent at a location during an observation period, and an average duration of visits at the location during the observation period (e.g., average of time elapsed between the first and the last mobility record from the same location without leaving said location).

As an implementation example, let L={l,2,3,...,l} be the locations available under the whole coverage area of the mobile network 202 operator. Let t¹, d,¹ denote the above components (i.e., total amount of time spent at a location and average duration of visits) for the given observation period for subscriber i and for location I. By evaluating these components for each location, a spatio-temporal presence categorization for each subscriber and each location is obtained. The categorization may be performed as follows in three main groups.

For the characteristic locations (e.g. the home or work locations) of a subscriber, the module may use those locations I, where ¾'>δ,. Here, δ, is a predefined lower bound parameter for subscriber i which can be defined by the following rules: the number of identified characteristic locations should not exceed three on average per subscriber so the total amount of time spent at these locations continuously should exceed 20% each, e.g., 5 hours per day on the long run.

The transit locations are regularly visited by the subscriber mainly because they are geographically close to the characteristic locations or they are on the way between those locations. For these locations, the repetitiveness is highly similar to the characteristic locations repetitiveness, but the total time spent at those locations is relatively low. A good example for a transit location is a metro station on the way to work. Hence for a transit location I, ¾'>¾, and ά!<ε, where ε, is an upper bound on the average duration at transit locations.

Optionally, the irregular location may be determined if ¾'<δ, and ε,<ύ_\ , wherein δ, is a predefined upper bound parameter for subscriber i, ε, is a predefined lower bound on the average duration at the irregular location, t,¹ is the total amount of time of subscriber i spent at a location I during an observation period, and d,' is the average duration of visits of subscriber i at the location I during the observation period. That is, the irregular locations are those locations that are visited for a relatively long duration, but the total time spent at that location calculated for the whole observa- tion period is low because the subscriber rarely or only occasionally visits the location.

It is noted that the observation period may be selected based on a frequency of occurrences of the irregular locations, and wherein the frequency of occurrences is based on δ,.

In the end, it can be summarized that the profiler 2002 may distil the irregular locations from the mobility patterns with a clear separation between the irregular locations and the characteristic and transit locations.

Then, in step S3a, the determiner 20035 of the analyzer performs determining a match between a first subscriber and a second subscriber when the first and second subscribers visit one common irregular location at the same time. That is, the analyzer 2003 may use these irregular locations for each and every subscriber, and by querying the respective records from the data collector 2001, the analyzer may assign timestamps to the irregular locations, as stated above.

In an optional step S3a-1, the correlator 20037 of the analyzer 2003 performs correlating the visits of a subscriber i and a subscriber j according to the following measure:

wherein Aj'(t) represents a spatio-temporal presence function of subscriber i at the irregular location I, Aj'(t) represents a spatio-temporal presence function of subscriber j at the irregular location I, and A '(t) represents the overlapping spatio-temporal presence of subscribers i, j (i.e., the time spent together at the same irregular location I at the same time period). In this way, a correlation analysis may be performed to find those subscriber pairs i,j that often visit, at the same time, the same locations irregular to both of them.

In other words, in order to obtain the colocation ratio q,j, the joint spatio-temporal presence is divided by the sum of the products of presence at irregular locations of the two subscribers. In that case, optionally, the spatio-temporal presence functions are defined in the following way A'(t) = 1 or Aj'(t) = 1 during the time when subscriber i or j is at location I, and 0 otherwise. In this way, the maximum value of Cij is 1 when subscribers i,j spend all their time at irregular locations together, and it is equal to 0 when they never meet at irregular locations.

Then, in step S4a, the identifier 20036 of the analyzer 2003 performs identifying, based on the number of matches, a spatio-temporal relationship between the first and second subscribers. In that case, in an optional step S4a-1, the thresholder 20038 of the analyzer 2003 performs thresholding S4a-1 the correlation measure such that Ci,j>c*, wherein the threshold c* is calculated with a probability theory approach in order to rule out the effects of randomness. That is, if the correlation measure exceeds the threshold, then subscribers i,j probably know each other. Given the size of the network of the operator, the number of subscribers and their mobility profiles, this threshold can be calculated with the probability theory approach in order to rule out the effects of randomness.

Optionally, the categorizing is performed based on at least one of a route taken to an irregular location during the observation period, and a movement speed on the route during the observation period. In this way, the system can be augmented with the additional information of how the subscribers move between locations. For a given social connection that is discovered by the above disclosure, if the mobility traces indicate that the subscribers travelled on the same route at the same time (with the same derived speed) to the irregular location, then this data gives confirmation to the result of the analyzer 2003.

In the following, a use case is described in order to illustrate the effect attainable by the present disclosure in relation to the first embodiment. Here, an expected value of number of subscriber pairs is calculated who are accidentally collocated once, twice or more times. Assume that subscribers spend exactly 10% of the time they spend at irregular locations (which is assumed in total to be the 10% of the observation period) at once at one location in a network that contains 100 thousand cells and 1,000,000 subscribers. This means that every subscriber visits 10 irregular locations where they spend 1% of the observation period continuously.

If both time and space are chosen from a uniform distribution, the probability of colocation is 10^"5 and the probability of the time overlap is 10² (assuming discrete time units) with another subscriber, i.e., the expected value of the joint presence is

X 0.01 X T % 10^"7 X T

Hence, the expected value of the colocation ratio is E(q,j)«10⁶ taking into account that the sum of the two subscribers' total time spent at irregular locations is 2χ10^_1. In case the locations and times are drawn from a normal distribution instead of a uniform distribution, these expected colocation ratios are higher.

By relaxing the constraint erected by the joint probability of pulling random locations and time windows for all subscribers at once, the expected number of subscriber pairs who are collocated n times can be written based on the probability components in the above equation:

For n=l, the expected number of subscriber pairs is approximately 5,000,000, E(X₂)=20, and for n=3 it is E(X₃)=4xlO ⁵. So, in this use case, if a subscriber pair was collocated at least 3 times, it would be safe to say that it is not a random phenomenon and that the subscribers know each other. In the above example, the threshold may be equivalent to 3 colocations, i.e., Cj,j=0.3.

Without loss of generality, the above first embodiment may be summarized as follows: the calculation presented above is based on the assumption that those subscribers who know each other sometimes meet at locations which do not belong to their routine mobility traces. If two subscribers are recorded at the same time at the same place, and if this happens multiple times, then the system assumes that a social connection exists between them. In order to make the confidence on the existence of the discovered social connections as high as possible, the system may strive to filter those locations that fall outside of the highly visited places of the subscribers.

The steps of finding the most probable offline social connections may be the following:

1. Record the location and the respective timestamp of every subscriber at all times.

2. Calculate, for each subscriber, the total time and the number of visits and then derive the average duration spent at every location visited during the observation period based on the mobility records: t},d_\ . For the sum of total time values T>∑it'vi holds, where T denotes the observation period. This is because only those records are taken into account in which the location of the subscriber is determined on the cell level.

3. Determine the set of irregular locations denoted by L, for each subscriber i by the upper and lower bounds on t,¹ and d' respectively: ί'<δί, and dj'>sj.

4. Look up all the recorded timestamps to the found irregular locations Li from the data collector 2001 to construct the spatio-temporal presence A' values that are basically on-off processes.

5. Calculate the collocation ratio C values for all subscriber pairs

6. Take those subscriber pairs i,j as offline social connections to become new edges in the social network built by the operator for which q,j>c* where c* is the social significance threshold.

Fig. 3B shows the second method embodiment which also reflects the interaction between the components of the apparatus embodiment.

In this context, in step Sib, the builder 20015 of the data collector performs building, per each subscriber, an application traffic usage profile from application usages via the mobile terminals 202. For instance, the data collector may consider the Gn and Gi interfaces e.g. in 3G networks in order to collect the time-stamped application traffic data for each and every subscriber e.g. by means of a Deep Packet Inspection (DPI) component.

Naturally, as in the first embodiment, the memory 20012 of the data collector 2001 may store the records (e.g. time, application traffic) of individual subscribers for a longer period.

Then, in step S2b, the determiner 20035 of the analyzer 2003 performs determining, from the application traffic usage profiles, a match between a first subscriber and a second subscriber when the first and second subscriber use the same application at the same time. For instance, the analyzer 2003 may use the DPI output of

applications and their respective traffic amounts as a function of time, and may determine the periods in which the individual subscribers were active in respect of the applications.

Finally, in step S3b, the identifier 20036 of the analyzer performs identifying, based on the number of matches, a temporal relationship between the first and second subscribers. For example, the analyzer 2003 may use the application usage profiles as time series for each and every subscriber, and by calculating their simultaneity within the respective applications, after aggregating the results for the observation period, similarly to the first embodiment, a threshold-based decision may be obtained.

Optionally, the application traffic usage profile for a particular subscriber comprises time windows during which the particular subscriber is active in an over-the-top, OTT, mobile application in which the particular subscriber interacts with at least one other subscriber; and the OTT mobile application is such that the particular subscriber interacts with the at least one other subscriber in real-time. In that case, the application may be one of a Voice over Internet Protocol (VoIP) application, and an online multiplayer gaming application. In the latter case, the VoIP application may be one of Skype™, Viber™, Whatsapp™ and Line™.

Without loss of generality, the present disclosure according to the second

embodiment may be summarized as follows: the calculation of the second

embodiment is based on the assumption that those subscribers who know each other are recorded using the same OTT application at the same time, and that this happens multiple times. Only those OTT applications are considered in which the users interact with each other in real time. In order to make the confidence on the existence of the discovered social connections as high as possible, the system strives to find frequently occurring interactions between the same two subscribers.

1. Record the application usage and the respective timestamp of every

subscriber at all times.

2. From the output of the data collector 2001, the profiler 2002 constructs the temporal activity A,¹ values for subscriber i and for application I that are basically on-off processes.

3. Calculate the coactivity ratio q,j values for all subscriber pairs i,j.

4. Take those subscriber pairs i,j as offline social connections to become new edges in the social network built by the operator for which c_i(j>c* where c* is the social significance threshold.

The present disclosure provides one or more of the following advantages:

• Providing an analytical framework for the mobile network operator to

contribute to its customer importance calculation: customer importance is highly related to social influence, i.e., the ability of the customer to drive the opinion of their so- - -

cial connections. Besides data sources directly available to the operator (e.g., CDR databases) and extracted online social interactions from network data traffic, the system of the present disclosure adds to the user importance calculation the spatio-temporal relationships (e.g. offline social connections). The system of the present disclosure provides that such offline social connections can be discovered that cannot be - although very important in real life - discovered from electronic communication activities.

• Enabling re-usability of existing data, i.e. by exploiting solely the data available to the mobile network operator

• Enabling simple integration: the system is based on passive measurements, i.e., no special mobile application or additional user activity is needed from the user side since all the inputs are restricted only to the user location/application usage data from the mobile network.

• Extending the social graph: The "offline" social connections are important additions to the social graph the operator maintains for selecting influential subscribers.

• Allowing better assessment of market success: the quality of experience and value for money judgments about services, or the success of marketing campaigns of new products highly depend on those people who raise their voices about a subject in the social media or in semi-private discussions. Identifying customer importance is crucial when it comes to the general perception of the mobile network operator, because those subscribers who have high influence on other people are more probably the opinion leaders about the services and products of the operator.

It is believed that the advantages of the technique presented herein will be fully understood from the foregoing description, and it will be apparent that various changes may be made in the form, constructions and arrangement of the exemplary aspects thereof without departing from the scope of the invention or without sacrificing all of its advantageous effects. Because the technique presented herein can be varied in many ways, it will be recognized that the invention should be limited only by the scope of the claims that follow.

Claims

1. A method for analytical identification of spatio-temporal relationships in a mobile communications network (200) comprising a plurality of subscribers with associated mobile terminals (202), the method comprising:

building (Sla), per each subscriber, a mobility profile by tracking the mobile terminals (202);

filtering (S2a), from the mobility profiles, at least one first irregular location visited by a first subscriber and at least one second irregular location visited by a second subscriber;

determining (S3a) a match between a first subscriber and a second subscriber when the first and second subscriber visit one common irregular location at the same time; and

identifying (S4a), based on the number of matches, a spatio-temporal relationship between the first and second subscribers.

2. The method according to claim 1, wherein the building step comprises:

repeatedly collecting (Sla-1) time-stamped location information for each mobile terminal (202).

3. The method according to claim 2, wherein the collecting step comprises querying (Sla-1-1) one of:

a Mobile Switching Center, MSC, node and a Visitor Location Register, VLR, node in a second generation, 2G, network;

a Serving GPRS Support Node, SGSN, and Radio Network Controller, RNC, in a third generation, 3G, network; and

a Mobility Management Entity, MME, in a Long Term Evolution, LTE, network or fourth generation, 4G, network.

4. The method according to claim 2 or 3, wherein, when the tracked mobile terminal (202) of a subscriber is in a ready state, the location information comprises geographical coordinates of a cell of the respective mobile terminal (202).

5. The method according to claim 2 or 3, wherein, when the tracked mobile terminal (202) of a subscriber is in one of a stand-by state and an idle state, the collecting step further comprises tracking (Sla-1-2) the respective mobile terminal (202) on at least one of:

the location level;

the Routing level; and

the Tracked Area level.

6. The method according to any one of claims 1 to 5, wherein the filtering step further comprises:

categorizing (S2a-1) the visited locations into categories.

7. The method according to claim 6, wherein the categories comprise two or more of:

characteristic locations;

transit locations; and

the irregular locations.

8. The method according to claim 6 or 7, wherein the categorizing is performed based on at least one of the following parameters:

a total amount of time spent at a location during an observation period; and an average duration of visits at the location during the observation period.

9. The method according to claim 7 and 8, wherein an irregular location is determined if:

ti'<6j and 8j<dj', wherein

6j is a predefined upper bound parameter for subscriber i;

8j is a predefined lower bound on the average duration at the irregular location;

ti¹ is the total amount of time of subscriber i spent at a location I during an observation period; and

d ' is the average duration of visits of subscriber i at the location I during the observation period.

10. The method according to any one of claims 7 to 9, wherein the determining step further comprises correlating (S3a-1) the visits of a subscriber i and a subscriber j according to the following measure:

_{= .} 2 x∑_l f A[ (t)dt wherein:

Ai'(t) represents a spatio-temporal presence function of subscriber i at the irregular location I,

A_]'(t) represents a spatio-temporal presence function of subscriber j at the irregular location I, and

Ai,_j'(t) represents the overlapping spatio-temporal presence of subscribers i, j.

11. The method according to claim 10, wherein the spatio-temporal presence functions are defined in the following way:

Ai'(t) = 1 or A_j'(t) = 1 during the time when subscriber i or j is at location I, and

0 otherwise.

12. The method according to claim 10 or 11, wherein the identifying step further comprises thresholding (S4a-1) the correlation measure such that:

Ci,j>C* ,

wherein the threshold c* is calculated with a probability theory approach in order to rule out the effects of randomness.

13. The method according to claim 8 and any one of claims 9 to 12 when depending on claim 8, wherein the categorizing is performed based on at least one of the following additional parameters:

a route taken to an irregular location during the observation period; and a movement speed on the route during the observation period.

14. The method according to claim 9, wherein the observation period is selected based on a frequency of occurrences of the irregular locations, and wherein the frequency of occurrences is based on δ,.

15. A method for analytical identification of temporal relationships in a mobile communications network (200) comprising a plurality of subscribers with associated mobile terminals (202), the method comprising:

building (Sib), per each subscriber, an application traffic usage profile from application usages via the mobile terminals (202); determining (S2b), from the application traffic usage profiles, a match between a first subscriber and a second subscriber when the first and second subscriber use the same application at the same time; and

identifying (S3b), based on the number of matches, a temporal relationship between the first and second subscribers.

16. The method according to claim 15, wherein:

the application traffic usage profile for a particular subscriber comprises time windows during which the particular subscriber is active in an over-the-top, OTT, mobile application in which the particular subscriber interacts with at least one other subscriber; and

the OTT mobile application is such that the particular subscriber interacts with the at least one other subscriber in real-time.

17. The method according to claim 15 or 16, wherein the application is one of:

a Voice over Internet Protocol, VoIP, application; and an

online multiplayer gaming application.

18. The method according to claim 17, wherein the VoIP application is one of:

Skype™;

Viber™;

Whatsapp™; and

Line™.

19. A computer program product comprising program code portions for performing the method of any one of claims 1 to 18 when the computer program product is executed on one or more computing devices.

20. The computer program product of claim 19, stored on a computer readable recording medium.

21. An apparatus (2001, 2002, 2003) for analytical identification of spatio-temporal relationships in a mobile communications network (200) comprising a plurality of subscribers (202) with associated mobile terminals (202), the apparatus comprising a processor (20015, 20025, 20035, 20036) configured to:

build, per each subscriber, a mobility profile by tracking the mobile terminals

(202); filter, from the mobility profiles, at least one first irregular location visited by a first subscriber and at least one second irregular location visited by a second subscriber;

determine a match between a first subscriber and a second subscriber when the first and second subscriber visit one common irregular location at the same time; and

identify, based on the number of matches, a spatio-temporal relationship between the first and second subscribers.

22. An apparatus (2001, 2002, 2003) for analytical identification of temporal relationships in a mobile communications network (200) comprising a plurality of subscribers with associated mobile terminals (202), the apparatus comprising a processor configured to:

build, per each subscriber, an application traffic usage profile from application usages via the mobile terminals (202);

determine, from the application traffic usage profiles, a match between a first subscriber and a second subscriber when the first and second subscriber use the same application at the same time; and

identify, based on the number of matches, a temporal relationship between the first and second subscribers.