Social network analysis for churn prediction
View Patent ↗Traditional methods of prediction of the likelihood of churning in a given market such as wireless telecommunications are improved by inclusion of social network analysis to predict churning by members of a population in response to churning by any given individual member of that population. A sphere of influence of individual members of the population on other members of the population are modeled and analyzed through a plurality of degrees of separation and quantitatively evaluated by a number of relational algebraic joins in which at least one such join includes weights to assert a diminishing influence with number of degrees of separation.
1. A method of predicting churning in a population, said method comprising steps of
counting interactions between respective members of said population,
determining chains of said interactions, said chains of interactions having stages corresponding to individual members of said population, said stages in a chain being limited to a predetermined number of stages,
applying different weights to said number of interactions for respective stages of said chains as determined in a limited number of relational algebra join operations corresponding to said predetermined number of stages of said interactions to produce weighted numbers of said interactions, and
summing said weighted numbers of interactions along said chains of interactions.
2. The method as recited in claim 1 , wherein said predetermined number of said stages of said interactions is three.
3. The method as recited in claim 2 , further including a step of determining overlaps of said interactions between members of said population.
4. The method as recited in claim 3 , including a further step of
counting interactions in said overlaps.
5. The method as recited in claim 4 , including a further step of
sorting said overlaps in accordance with a number of interactions counted in said step of counting interactions.
6. The method as recited in claim 5 , wherein said step of determining overlaps includes steps of
sorting said interactions between members of said population in accordance with respective participants in respective interactions, and
determining a proportion of interactions of a member of said population with another member of said population to all interactions of said member of said population.
7. The method as recited in claim 6 , including a further steps of
applying a threshold to said proportion of interactions, and
performing said steps of counting interactions, applying different weights and summing said weighted numbers in regard to members of said population for which said proportion of interactions exceeds said threshold.
8. The method as recited in claim 7 , wherein said members of said population are subscribers to a telecommunication network and said interactions are calls made using said telecommunication network.
9. The method as recited in claim 8 , wherein said telecommunication network is a wireless communication network.
10. The method as recited in claim 3 , wherein said step of determining overlaps includes steps of
sorting said interactions between members of said population in accordance with respective participants in respective interactions, and
determining a proportion of interactions of a member of said population with another member of said population to all interactions of said member of said population.
11. The method as recited in claim 10 , including a further steps of
applying a threshold to said proportion of interactions, and
performing said steps of counting interactions, applying different weights and summing said weighted numbers in regard to members of said population for which said proportion of interactions exceeds said threshold.
12. The method as recited in claim 11 , wherein said members of said population are subscribers to a telecommunication network and said interactions are calls made using said telecommunication network.
13. The method as recited in claim 12 , wherein said telecommunication network is a wireless communication network.
14. The method as recited in claim 1 , further including a step of determining overlaps of said interactions between members of said population.
15. The method as recited in claim 14 , including a further step of
counting interactions in a said overlaps.
16. The method as recited in claim 15 , including a further step of
sorting said overlaps in accordance with a number of interactions counted in said step of counting interactions.
17. The method as recited in claim 16 , wherein said step of determining overlaps includes steps of
sorting said interactions between members of said population in accordance with respective participants in respective interactions, and
determining a proportion of interactions of a member of said population with another member of said population to all interactions of said member of said population.
18. The method as recited in claim 17 , including further steps of
applying a threshold to said proportion of interactions, and
performing said steps of counting interactions, applying different weights and summing said weighted numbers in regard to members of said population for which said proportion of interactions exceeds said threshold.
19. A method of predicting churning in a population, said method comprising steps of
configuring a data processor to count interactions between respective members of said population,
configuring a data processor to determine chains of said interactions, said chains of interactions having stages corresponding to individual members of said population, said stages in a chain being limited to a predetermined number of stages,
configuring a data processor to apply different weights to said number of interactions for respective stages of said chains as determined in a limited number of relational algebra join operations corresponding to said predetermined number of stages of said interactions to produce weighted numbers of said interactions, and
configuring a data processor to sum said weighted numbers of interactions along said chains of interactions.
20. The method as recited in claim 19 , wherein said predetermined number of said stages of said interactions is three.