Data compression method for a classifier in a telecommunication environment utilizing aggregation of data over time periods of increasing length

1. A method of classifying a sample of values related to the use of a telephone service, the method comprising the steps of:

(a) recording call data records in a server, each call data record identifying a subscriber to the telephone service and comprising at least a duration value and a date value;

(b) reading on the fly the call data records from the server and identifying the call data records relative to the subscriber;

(c) aggregating the identified call data records over a basic time interval in a respective database record assigned to the subscriber and the basic time interval, wherein

the duration value of a current call data record is aggregated by incrementing a counter field of the database record, and

the date value of the current call data record is aggregated by overwriting a respective field of the database record with the difference between the date value and a current date;

(d) repeating step (c) for a number of successive basic time intervals, to form a first group of database records assigned to the subscriber;

(e) when the first group is complete, aggregating the values of the database records of the first group in a database record assigned to the subscriber and an extended time interval;

(f) repeating steps (c) thru (e) for a number of the extended time intervals, to form a second group of database records assigned to the subscriber;

(g) configuring a classifier tool with a behavioral model formed of a weighted list of parameters;

(h) forming a sample of values based on the contents of the first and second groups of database records; and

(i) providing the sample of values as parameters to the classifier tool, whereby the classifier tool calculates a score representative of the adequacy of the sample with a target category.

2. The method according to claim 1 , wherein the records are used circularly in each group.

3. The method according to claim 1 , further comprising:

identifying samples that are effectively in the target category at the end of a reference period;

configuring the classifier tool in learning mode;

submitting the identified samples to the classifier tool, such that the classifier tool refines the behavioral model for conformance of the calculated score to the effective results; and

configuring the classifier tool with the refined behavioral model to calculate the following scores.

4. The method according to claim 1 , wherein the parameters of the behavioral model comprise statistical variables.

5. The method according to claim 4 , wherein the statistical variables comprise tendencies, averages, maximum values, and minimum values over different durations.

6. The method according to claim 1 , wherein the behavioral model comprises distinct parameters for the different groups of records.

7. The method according to claim 1 , wherein

the target category corresponds to subscriber chum, and

the method further comprises the following steps:

aggregating the identified call data records each day in database records of a first group of at least seven records:

aggregating the contents of the records of the first group in one record of a second group of at least four records; and

aggregating the contents of the records of the second group in one record of a third group of records.

8. The method according to claim 7 , wherein the telephone service is based on prepaid refills, the parameters of the behavioral model further comprising:

the amount of time until the refill expires;

the amount of time since the last refill;

the monthly tendency of the number of purchased refills;

the amount of time since the last call; and

the monthly deviation of the number of purchased refills.