FR3116639A1 - Method for estimating the consistency of the circadian rhythms of individuals from their telephone call reports - Google Patents

Abstract

A method of tracking the behavior of an individual i using their call log data, CRA, the method comprising: - receiving data from CRA corresponding to the individual l;- determining a value Dself(i,T1,T2) corresponding to a divergence between two law of probability of occurrence of the performance of an activity by the individual l on the associated user terminal;- the determination of a value Dinter (i,l,Tk) corresponding to a divergence between two laws of probability of occurrence of the respective performance of an activity by individual i and of an individual l on the associated user terminal;- determining a score α relating to a level of divergence in the behavior of individual i, the score α being dependent on a value Dself(i,T1,T2) - Dinter(i,l,Tk);- the identification of the user terminal associated with individual i if the score α verifies a condition.

Description

Method for estimating the consistency of the circadian rhythms of individuals from their telephone call reports

The present invention relates to the field of personal assistance and in particular alerting in the event of a problem.

There are several types of active alert systems for the elderly. For example, there are remote alarms allowing the elderly person in distress to press a button on a portable device. The most advanced of these devices include a geolocation system to quickly find the elderly person.

These devices therefore make it possible to alert relatives of occasional and immediate distress.

However, these devices require that the elderly person be able to implement the alert. For example, if the elderly person or more generally the person requiring follow-up does not remember how the device works (for example, Alzheimer's disease) then these devices are useless, or even can encourage the entourage to release the person's follow-up .

In addition, it is necessary that the person followed is aware of the problem encountered. When it comes to a fall, the person can, for example, lose consciousness or lose all their bearings. The person can also stop eating or be in a depressed state, their state of health can then deteriorate without their being aware of it.

Thus, these useful devices in situations of minor and temporally well-identified distress of an individual turn out to be useless when the distress is greater or when it is established very gradually.

The present invention improves the situation.

A first aspect of the invention relates to a method for monitoring the behavior of an individual i over at least two time intervals T1, T2 by means of his call report data, CRA, the method comprising:
- the reception of CRA data corresponding to user terminals respectively associated with each individual I of a group I of individuals, the CRA data comprises time moments included in the time intervals T1 and T2 where an activity is carried out by at least one individual l on the associated user terminal;
- the determination of a value D _self (i, T1, T2) corresponding to a divergence between P _i ^T1 and P _i ^T2 where P _l ^Tk corresponds to a law of probability of occurrence of the performance of the activity by l individual l on the associated user terminal during a time interval Tk;
- the determination of a value D _inter (i,j,Tk) corresponding to a divergence between P _i ^Tk and P _j ^Tk for j different from i and for at least one time interval Tk among the time intervals T1 and T2 ;
- the determination of a score α relating to a level of divergence of the behavior of the individual i , the score being dependent on a value D _self (i, T1, T2) - D _inter (i, j, Tk);
- the identification of the user terminal associated with the individual i if the score verifies a condition.

Thus, by means of CRA data obtained by the telecommunications operators, the method makes it possible to determine via a score whether the individual monitored has an abnormal behavior over the time interval T2 (respectively T1) compared to his behavior over the time interval T1 (respectively T2). Thus, the user terminal of the individual being tracked is identified, allowing him to then be assisted. The abnormal behaviors monitored here are of two different types, either one monitors a different behavior between two time intervals, or a very similar behavior; the invention makes it possible to monitor both types of behavioral changes together. The condition to be checked is used to monitor differences in behavior, very similar behaviors or both. Thus, when the score is high, the behaviors are similar between the two time intervals, the condition may thus be the fact of having a score above a high threshold. When the score is low, the behaviors are distinct, the condition can thus be the fact of having a score below a low threshold. When one wishes to monitor both types of behavior change together then the condition may be that the score is above a high threshold or below a low threshold. When following individuals with a certain type of disorder, the repetition of tasks over a period of several days or even several months can be an important indicator for monitoring the evolution of a disease. Conversely, when monitoring elderly individuals at risk of falling, for example, a significant difference in behavior from one day to the next can be a relevant indicator.

The CRAs held by telecommunications operators represent a very rich source of data. The CRAs contain a set of information recorded on each call or each text exchange, in particular temporal data (timestamp of calls, duration of exchanges), social data (calling interlocutor and called interlocutor) and spatial data (location of antennas involved in the appeal).

Thus, the difference or on the contrary the similarity of the behavior of the individual between two time intervals T1 and T2 (for example time intervals corresponding to days) can be detected by means of the activity of the monitored individual ( that is to say the individual i ) on his user terminal, namely the user terminal associated with the tracked individual.

The method makes it possible to determine the difference or the similarity of the behavior of the individual followed over several time intervals in a reliable manner. Indeed, the activity of the individual via his user terminal can vary from one time interval to another without this constituting an abnormal difference in behavior. Conversely, the activity of the individual via his user terminal can remain constant from one time interval to another without this constancy constituting abnormal behavior.

For this, the method takes into consideration the differences between the behavior of the individual being monitored (individual i ) and that of another individual (individual j ). This makes it possible to avoid taking into consideration changes or constants in the activity of the individual i via his user terminal which would be insignificant. This makes it possible to obtain a more relevant result in the follow-up of the individual i .

By time interval, it is understood any time interval long enough to allow an analysis, for example a time interval of one hour, one day or even six months. The two time intervals T1 and T2 can be two time intervals of the same duration and/or covering similar periods during a day (for example, starting and/or ending at the same time). For example, these periods may each correspond to a different day. These time slots can also follow one another, i.e. T2 corresponds to a time slot starting after time slot T1. For example, the time intervals T1 and T2 could correspond to two successive days or two successive periods of six months.

Call reports, CRA, are information recorded by telecommunications operators. This information relates to each activity carried out via a user terminal which generates the use of the communication network of the telecommunications operator.

By user terminal, it is understood any terminal capable of communicating on the network of a telecommunications operator, for example, a mobile phone, a smartphone, a computer, a connected object, etc.

By user terminal associated with an individual, it is understood the user terminal of which the telecommunications operator is able to know that the user of the user terminal is the individual with whom he is associated.

Group I of individuals means a group of individuals whose CRAs are accessible. The group I of individuals includes the individual i to follow, the other individuals of the group can be individuals randomly selected or having common characteristics with the individual i , for example, individuals of the same age group or in the same geographical area or even individuals also monitored, that is to say individuals for whom the same monitoring is implemented as for individual i .

By activity or even activity carried out via a user terminal, it is understood any activity which requires the use of a communication network. This may involve, for example, making a phone call, receiving a phone call, answering phone calls or even sending or receiving messages (voice, text, etc.). An activity can also be a combination of these tasks, such as sending a text message and making a phone call. The activity is carried out by the individual on the user terminal associated with this individual.

The temporal moments when an activity is carried out via the user terminal corresponds to a set of moments, each of these moments representing a moment when the individual associated with the user terminal carries out the activity. Temporal moments are included in the CRAs. The temporal moments can be timestamps also called “timestamp” in English. These time moments are included in time intervals, for example the time intervals T1 and T2.

By probability law of occurrence of the performance of an activity by the individual l on the associated user terminal during a time interval Tk, is meant the probability law corresponding to the probability law on the time interval Tk of performance of the activity by the individual 1 on his user terminal during a time interval Tk.

By divergence is meant the function which measures the dissimilarity of two probability laws or of two distributions (or distribution functions) corresponding respectively to the two probability laws. The divergence can be determined on the basis of the Kullback-Leibler divergence or the Jensen-Shannon divergence.

The score relating to a level of divergence of the behavior of the user can be increasing according to a level of divergence of the behavior or decreasing according to the level of divergence. The condition to be checked can be of any type, for example that the score must be lower or higher than a threshold or even included in an interval.

The level of divergence of the behavior of individual i is an evaluation of the consistency (or even similarity) of the regularity of the performance of the activity by individual i between at least two time intervals.

Thus, when the score verifies the condition, the level of behavioral divergence can be considered high or very low. The activity of individual i during an interval is then considered suspicious or considered abnormal because it is significantly different or abnormally similar compared to this same activity carried out during other time intervals. In this case the user terminal and therefore the individual i is identified. It is then possible to monitor vulnerable people with great reliability (elderly people, people with anxiety disorders, depressed people, sick people and/or people receiving care) to determine whether the person could be in danger (significantly tired , phycological distress, physically immobilized, social isolation, nocturnal activity, etc.). It is also possible to help physicians to target earlier individuals likely to develop certain diseases or to have risky behaviors (for example, alcoholism disorders) and/or symptomatic (sleep disorders, desynchronization of rhythms circadian activities of a social nature).

According to one embodiment, the method further comprises the determination of values D _inter (i, l , Tk) corresponding respectively to divergences between P _i ^Tk and P _l ^Tk for the individuals l of the group I of individuals different from i and for the time intervals T1 and T2, in which the score is dependent on the values D _self (i,T1,T2) - D _inter (i, l ,Tk) with l ∈ I and k ∈ {1,2}.

The score is determined by taking into account all the individuals of group I. Thus, the evaluation of the difference (or divergence) in behavior of the individual i followed is more reliable (and the score more significant) because the differences in behavior insignificant are not taken into account.

The score α can be obtained by counting the number of pairs ( l , k) ∈ I\{i}×{1,2} for which D _self (i,T1,T2) - D _inter (i, l ,Tk) is above a threshold. The threshold can be, for example, equal to 0. The threshold can be chosen greater than 0 in this case the score will be lower and the evaluation errors reduced. The threshold can be chosen to be less than 0, in which case the score will be higher and the identification of the user terminal more frequent. Thus, it is possible to vary the threshold from which the identification is carried out and possibly the alert given. Obtaining the score in this way requires little computing power.

According to one embodiment, the probability laws P _I ^Tk are defined by functions , where t_k,mis a sub-time interval of the time interval Tk and U_myou_k,mis equal to Tk and n _I (t_k,m) represents the number of occurrences of the activity during the sub-time interval t_k,mfor the individualI.

The functions f _{l ,Tk} (t _k,m ) are respectively the distribution functions of the probability laws P _l ^Tk . The sub-intervals of time t _k,m cover (or are equal to) the interval Tk, i.e. the union of the t _k,m (U _m t _k,m ) covers (or is equal to a) Tk. The sub-intervals of time t _k,m can be pairwise disjoint, ie the intersection of t _k,m1 with U _m\m1 t _k,m is empty for all m1.

According to one embodiment, the method further comprises:
- the reception of data from CRA corresponding to the user terminals of the individuals l of the group I of individuals, the data from CRA comprises time moments comprised in time intervals Tk with k ∈ {3,…N} where an activity is carried out by at least one individual l on the associated user terminal;
- the determination of the values D _self (i,Tk,Tk+1) corresponding respectively to divergences between P _i ^Tk and P _i ^Tk+1 for k ∈ {2,…N-1};
- the determination of the values D _inter (i, l ,Tk) corresponding respectively to divergences between P _i ^Tk and P _l ^Tk for l different from i and for at least the time intervals Tk for k ∈ {1,…N- 1};
in which the determination of the score α is dependent on the values D _self (i,Tk,Tk+1) - D _inter (i, l ,Tk) and/or D _self (i,Tk,Tk+1) - D _inter ( i, l , Tk+1).

The score is determined by following the individuals of group I over N time intervals, i.e. over more than two time intervals. Thus, the evaluation of the difference (or divergence) in behavior of the individual i followed is more reliable (and the score more significant) because the significant differences in behavior are more represented. The method is thus less sensitive to isolated variations in the data received.

The time intervals Tk for k ∈ {1,…N} can be time intervals of the same duration and/or covering similar periods during a day (e.g., start and/or end at the same time) . For example, these periods may each correspond to different days. These time intervals can also follow one another, that is to say that Tk (also denoted T _k ) corresponds to a time interval starting after the time interval Tk-1 (also denoted T _k-1 ). For example, the time intervals Tk for k ∈ {1,…N} could respectively correspond to successive days.

According to one embodiment, if a value β is greater than a predetermined threshold β _lim , where the value β corresponding to P _B (X=α) and P _B is the binomial law, the identification of the user terminal associated with the individual i is not realized.

This ensures that the α score is significant. Thus, the user terminal is only identified if the score α is significant. Indeed, this mainly makes it possible to avoid situations in which, for example, a score α would be wrongly considered as significant when it is non-zero. This also makes it possible to determine the significance value of a calculated non-zero score α by ensuring its conformity with the meaning of the determined value β. Identification therefore only takes place when the value β is lower than the predetermined threshold β _lim . The threshold can be for example 0.05, 0.10, 0.025 or even 0.01. This threshold is predetermined, that is to say determined in a configuration step of the system preceding the steps of the invention.

According to one embodiment, the method further comprises transmitting an emergency instruction and/or an alert message to a tracking user terminal when the user terminal associated with the individual i has been identified.

This alerts people following the individual that their behavior either differs significantly or differs very little between one or more time intervals. Thus, the people following the individual will be able to trigger the appropriate measures of action. Monitoring is thus more efficient. The emergency instruction transmitted to the monitoring user terminal, instructs the user terminal to implement an emergency measure, for example triggering an audible alert.

To avoid unnecessary alerts, provision is also made to transmit to the user terminal associated with the individual i a message requiring a return from the individual i when the user terminal associated with the individual i has been identified, the transmission of the alert message being conditioned on the absence of receipt of a return from the individual i.

A second aspect of the invention relates to a computer program comprising instructions for implementing the steps of the method described above, when these instructions are executed by a processor.

A third aspect of the invention relates to a server for monitoring the behavior of an individual i over at least two time intervals T1, T2 by means of his call report data, CRA, the server comprising:
a network interface;
a processor; And
a memory storing instructions, so that when these instructions are executed by the processor, they configure the server to:
- receiving, via the network interface, CRA data corresponding to user terminals respectively associated with each individual I of a group I of individuals, the CRA data comprising time moments comprised in the time intervals T1 and T2 where a activity is performed by at least one individual l on the associated user terminal;
- determining a value D _self (i,T1,T2) corresponding to a divergence between P _i ^T1 and P _i ^T2 where P _l ^Tk corresponds to a law of probability of occurrence of the performance of the activity by the individual l on the associated user terminal during a time interval Tk;
- determining a value D _inter (i, l , Tk) corresponding to a divergence between P _i ^Tk and P _l ^Tk for l different from i and for at least one time interval Tk among the time intervals T1 and T2;
- determining a score α relating to a level of divergence of the behavior of the individual i, the score α being dependent on a value D _self (i,T1,T2) - D _inter (i, l , Tk);
- identification of the user terminal associated with the individual i if the score α verifies a condition.

Other characteristics and advantages of the invention will become apparent on reading the description which follows. This is purely illustrative and must be read in conjunction with the appended drawings on which:
- there illustrates a server and user terminals according to the invention;
- there illustrates a flowchart representing the method according to the invention.

The example of the describes a group I of individuals 12, 22, 32 and 42 and user terminals 11, 21, 31 and 41 respectively associated with individuals 12, 22, 32 and 42.

The user terminal 11 sends and receives 5.1 text messages (for example SMS) with the user terminal 41 and sends and receives 5.2 text messages (for example SMS) with the user terminal 31. The user terminal 41 sends and receives 5.3 text messages (for example SMS) with the user terminal 31. These exchanges 5.1, 5.2 and 5.3 take place via a communication network 1 of one or more telephone operators. The user terminals 31 and 11, which are in the example of the mobile phones, access the communication network 1 via radio access by the respective base stations 2.1 and 2.2. The user terminal 41, which is a computer, accesses the network via a wired or optical connection, that is to say via a telephone modem and a router 2.3. The user terminal 21 does not exchange text messages with the user terminals 11, 31 and 41 associated with the individuals 12, 32 and 42 of the group I of individuals, nor with other user terminals.

The relevant activity for monitoring an individual or a group of individuals depends on the type of vulnerability or problem that this group of individuals is likely to encounter. In the example of the individuals 12, 22, 32 and 42 of group I of individuals require significant follow-up, for example these individuals all have a neurodegenerative disease. These individuals 12, 22, 32 and 42 exchange text messages 5.1, 5.2, 5.3 with other individuals (for example members of their families). Thus, in order to adequately monitor the behavior of these individuals 12, 22, 32 and 42, the activity considered is that of the exchanges (reception/transmission) of text messages.

As previously indicated, the relevant activity could be different, for example, making a phone call. Individuals 12, 22, 32, 42 can also be healthy individuals with similar characteristics (age, sex, occupation, etc.).

The base stations 2.1, 2.2 and the router 2.3 which allow the user terminals 11, 31, 41 access to the network 1, participate in the transmission of the messages 5.1, 5.2, 5.3. These entities 2.1, 2.2, 2.3 of the communication network 1 record in databases 3.1, 3.2, 3.3 control data and metadata of the streams passing through them. These data form the CRA (call reports). The CRAs include for each text message transmission in particular:
an interaction type (text, phone call);
a timestamp of the transmission of the text message;
the sender of the text message (ID, telephone number, etc.);
the recipient of the text message (username, phone number, etc.);
the number of characters in the text message;
the location of the sender and/or the recipient through the identity of the telephone antenna involved and/or its associated INSEE code;
etc

The CRAs relating to the individuals monitored 12, 22, 32, 42, that is to say the individuals of group I of individuals, and to the relevant activity within the framework of this monitoring are transmitted via the connections 4.1, 4.2 and 4.3 to tracking server 50.

The computer 50 comprises a module (INT) 51 for interfacing with the communication network 1, a processor (PROC) 52 and a memory (MEMO) 53.

The memory 53 comprises a non-volatile memory on which the computer program is stored and a volatile memory on which the parameters for the implementation of the invention are stored, for example, the activities to be considered, the group I of individuals to track, the number of the tracking user terminal, the divergence function to implement, the time intervals to implement T1,…,TN, the sampling step of these time intervals, namely the time sub-intervals t _{k ,m} or even the thresholds δ _lim , δ _max , α _lim , α _max or even β _lim . Volatile memory also stores CRA data and individual occurrences of activity during each sub-time interval.

The interface module with the communication network (INT) 51 allows the computer 50 to receive the CRAs and to transmit emergency instructions and/or an alert message to a monitoring user terminal. The interface module can be an internet modem and/or a computer network card or even a wireless communication module making it possible to receive and decode a radio signal coming from an antenna of the communication network 1.

The processor 52 is in particular configured to control the interface module (INT) 51 in order to be able to receive via the network 1 the CRAs and possibly to send via the communication network 1 to transmit the emergency instructions and/or the message alert and/or the identified user terminal.

The processor 52 is also configured to determine the score α (or more generally the scores α _q,i ), identify the user terminal, determine and compare the value β.

There represents a flowchart of a method according to a particular embodiment.

At step S1, the tracking server 50 and the databases 3.1, 3.2, 3.3 are configured.

Thus, the parameters relating to the tracking server 50 are defined and stored in the memory 53.

The databases 3.1, 3.2, 3.3 are also configured to allow the recording of CRA data and the transmission of these CRA data to the server 51 via the communication network 1.

In step S2, the databases 3.1, 3.2, 3.3 transmit to the tracking server 51 the CRAs formed following the exchanges of text messages in which the various user terminals associated with the individuals of the group I of individuals are involved.

At step S3, the tracking server 50 determines on the basis of the CRAs received the numbers n _I (t_k,m) of occurrences of the activity during the sub-time interval t_{k, m}for each individualI. The tracking server 50 then calculates for each individualIand for each time interval Tk. The tracking server 51 then deduces therefrom the probability laws P _I ^Tk (ForI∈I and k ∈ {1,…N-1}) whose functionsf _{I ,Tk} are respectively the distribution functions.

At step S4, the tracking server 50 determines the values D _self ( i ,T1,T2) obtained from a divergence D between P _i ^Tk and P _i ^Tk+1 for i ∈I and k ∈ {1 ,…N-1}. The divergence D is the function that measures the dissimilarity of two probability distributions. Thus, the divergence used can be the Kullback-Leibler divergence or the Jensen-Shannon divergence. Thus, the values D _self ( i , T1, T2) can be equal to D(P _i ^Tk , P _i ^Tk+1 ) or even the square root of D(P _i ^Tk , P _i ^Tk+1 ).

At step S5, the tracking server 50 determines the values D _inter (i, l , Tk) obtained from a divergence D between P _i ^Tk and P _l ^Tk for (i , l )∈I ² with l ≠ i and k ∈ {1,…N}. Divergence D is the same function employed in step S4. Thus, the values D _inter (i, l ,Tk) can be equal to D(P _i ^Tk , P _l ^Tk ) or even the square root of D(P _i ^Tk , P _l ^{T k} ). Since the square root is employed in step S4, it is also employed in step S5.

At step S6, the tracking server 50 determines the scores α _{q , i} for q ∈ {1,…N-1} and i∈I. Each score α _{q , i} depends on the values D _self (i,Tq,Tq+1) - D _inter (i, l ,Tq) for l ∈I\{ l } with i≠ l . For example, each α _q,i can be obtained by counting the pairs ( l , r) ∈ I\{i}×{0;1} such that D _self (i,Tq,Tq+1) - D _inter (i , l ,Tq+r) is greater than a threshold s. This threshold can for example be 0.

In step S7.1, the tracking server 50 determines whether the α _{q, i} are significant. For this, the server determines the values β _{q, i} equal to P _B (X= α _{q, i} ) where P _B is the binomial distribution, that is to say that where u is a parameter of the predetermined binomial distribution (for example, u can be chosen at ½ or even at a higher value to make the test more demanding). When the value β _{q, i} is greater than a predetermined threshold β _lim (for example, 0.05, 0.10 or even 0.025) the value α _{q, i} is considered to be insignificant. This value is then ignored. When the value β _{q, i} is lower than the predetermined threshold β _lim the value α _{q, i} is considered significant. The following steps can be implemented with the meaningful values. Steps S7.1 and S7.2 are optional.

At step S7.2, the tracking server 50 determines the scores δ _i obtained by counting the q ∈ {1,…N-1} such that the value α _{q, i} is greater than a threshold α _lim and is considered as significant in step S7.1 (if step S7.1 is performed). The test performed in step S7.1 can also be performed to determine whether the scores δ _i are significant. For this, the server determines if P _B (X= δ _i ), i.e. , is less than a threshold, for example β _lim . N'' corresponds to the number of α _{q, i} considered significant in step S7.2. If step S7.2 is not implemented then N'' is taken equal to N-1.

At step S8, the tracking server 50 identifies the user terminals associated with the individuals of group I.

The user terminal associated with individual i can be identified when:
- at least one α score_{q, I}for q ∈ {1,…N-1} is less than the threshold α_limitif steps S7.1 and S7.2 are not implemented; Or
- at least one α score_{q, I}for q ∈ {1,…N-1} is less than the threshold α_limitand β_{q, I}is below threshold β_limit whether step S7.1 is implemented; Or
- the score δ_Iis less than a threshold δ_limitand δ_Iis considered significant.

This involves determining a difference in behavior of individual i during time intervals Tk for k ∈ {1,…N} _.

Alternatively, the user terminal associated with individual i can be identified when:
- a large percentage (e.g. 90%) of α scores_{q, I}for q ∈ {1,…N-1} are greater than a high threshold α_maxif steps S7.1 and S7.2 are not implemented; Or
- a large percentage (e.g. 90%) of α scores_{q, I}for q ∈ {1,…N-1} are greater than a high threshold α_maxand have a β_{q, I}below threshold β_limit if step S7.1 is implemented; Or
- the score δ_Iis greater than a threshold δ_maxand δ_Iis considered significant.

The aim here is to determine an abnormal regularity in the behavior of individual i during the time intervals T k for k ∈{1,…N} _. This could also reflect a risk of a particular pathology (eg depression, obsessive disorder, etc.).

In step S9 the tracking server 50 sends a message to the individual i whose associated user terminal has been identified. This message requires feedback from individual i . For example, a message like "are you okay?"". It can also be an instruction for the user terminal so that it asks the individual i to perform an action, for example, pressing a key on the user terminal, or even dragging an element displayed on the screen. 'screen. This step can be optional.

In step S10 the tracking server 50 sends an emergency instruction and/or user terminal identification data and/or an alert message to a tracking user terminal. This dispatch can be conditioned on the absence of a response or reaction from the individual i following step S9 if the latter is implemented.

The alert message can indicate the type of behavior that was observed (difference in behavior or high similarity in behavior). This message may also indicate information (telephone number, name, address, contact person, etc.) about the individual i who may be the patient of the user of the monitoring user terminal. It is particularly advantageous when behavioral differences are sought to implement emergency measures. Thus, emergency instructions can trigger an alarm on the practitioner's tracking terminal or even notify the emergency services.

Claims (10)

Method for monitoring the behavior of an individual i over at least two time intervals T1, T2 by means of his call report data, CRA, the method comprising:
- the reception of CRA data corresponding to user terminals respectively associated with each individual I of a group I of individuals, the CRA data comprising time moments included in the time intervals T1 and T2 where an activity is carried out by at least one individual l on the associated user terminal;
- the determination of a value D _self (i, T1, T2) corresponding to a divergence between P _i ^T1 and P _i ^T2 where P _l ^Tk corresponds to a law of probability of occurrence of the performance of the activity by l individual l on the associated user terminal during a time interval Tk;
- the determination of a value D _inter (i,j,Tk) corresponding to a divergence between P _i ^Tk and P _j ^Tk for j different from i and for at least one time interval Tk among the time intervals T1 and T2 ;
- the determination of a score α relating to a level of divergence of the behavior of the individual i , the score α being dependent on a value D _self (i,T1,T2) - D _inter (i,j,Tk) ;
- the identification of the user terminal associated with the individual i if the score α verifies a condition. Method according to claim 1 further comprising the determination of values D _inter (i, l ,Tk) corresponding respectively to divergences between P _i ^Tk and P _l ^Tk for the individuals l of the group I of individuals different from the individual i and for the time intervals T1 and T2, and in which the score α is dependent on the values D _self (i,T1,T2) - D _inter (i,l,Tk) with l ∈I and k ∈ {1,2 }. Method according to Claim 2, in which the score α is obtained by counting the number of pairs ( l , k) ∈ I\{i}×{1,2} for which D _self (i,T1,T2) - D _inter ( i,l,Tk) is greater than a threshold. Method according to one of the preceding claims, in which the probability distributions P_I ^Tk are defined by functions , where t_k,mis a sub-time interval of the time interval Tk and U_myou_k,mis equal to Tk and n _I (t_k,m) represents the number of occurrences of the activity during the sub-time interval t_k,mfor the individualI. Method according to one of the preceding claims further comprising:
- the reception of data from CRA corresponding to the user terminals of the individuals of group I of individuals, the data from CRA comprises time moments comprised in time intervals Tk with k ∈ {3,…N} where an activity is carried out by at least one individual l on the associated user terminal, the time intervals Tk and Tk+1 for k ∈ {3,…N-1};
- the determination of the values D _self (i,Tk,Tk+1) corresponding respectively to divergences between P _i ^Tk and P _i ^Tk+1 for k ∈ {2,…N-1};
- the determination of the values D _inter (i, l ,Tk) corresponding respectively to divergences between P _i ^Tk and P _l ^Tk for l different from i and for at least the time intervals Tk for k ∈ {1,…N- 1};
in which the determination of the score α is dependent on the values D _self (i,Tk,Tk+1) - D _inter (i, l ,Tk) and/or D _self (i,Tk,Tk+1) - D _inter ( i, l , Tk+1). Method according to one of the preceding claims, in which if a value β is greater than a predetermined threshold β _lim , where the value β corresponding to P _B (X=α) and P _B is the binomial law, the identification of the user terminal associated with individual i is not carried out. Method according to one of the preceding claims, further comprising transmitting an emergency instruction and/or user terminal identification data and/or an alert message to a monitoring user terminal when the user terminal associated with individual i has been identified. The method of claim 7 further comprising transmitting to the user terminal associated with individual i a message requesting feedback from individual i when the user terminal associated with individual i has been identified, and wherein transmitting the alert message, of the emergency instruction, of user terminal identification data is carried out in the absence of receipt of a return from the individual i. Computer program comprising instructions for implementing the steps of the method according to one of the preceding claims, when these instructions are executed by a processor. Server for monitoring the behavior of an individual i over at least two time intervals T1, T2 by means of his call report data, CRA, the server comprising:
a network interface module;
a processor; And
a memory storing instructions, so that when these instructions are executed by the processor, they configure the server to:
- receiving, via the network interface module, data from CRA corresponding to user terminals respectively associated with each individual I of a group I of individuals, the data from CRA comprises time instants comprised in the time intervals T1 and T2 wherein an activity is performed by at least one individual 1 on the associated user terminal;
- determining a value D _self (i,T1,T2) corresponding to a divergence between P _i ^T1 and P _i ^T2 where P _l ^Tk corresponds to a law of probability of occurrence of the performance of the activity by the individual l on the associated user terminal during a time interval Tk;
- determining a value D _inter (i, l , Tk) corresponding to a divergence between P _i ^Tk and P _l ^Tk for l different from i and for at least one time interval Tk among the time intervals T1 and T2;
- determining a score α relating to a level of divergence in the behavior of the individual i , the score α being dependent on a value D _self (i,T1,T2) - D _inter (i, l ,Tk);
- identification of the user terminal associated with the individual i if the score α verifies a condition.