AI techniques currently used or investigated for customer profiling are: ensembles of fuzzy inference systems (Type-I FIS, Type-II FIS, fuzzy-neural systems); evolutionary algorithms (evolutionary functional link, genetic algorithms); dynamic filter weight neural networks; sliding-modes; sparse algorithms; swarm intelligence algorithms like particle swarm optimization; and so forth.
Particularly, AI-based ensembles methods and big data analysis are also considered.
I would say that together to what Hiram Ponce quoted, classical techniques such as rule-based reasoning, case-based matching / reasoning, etc. should also deliver satisfactory results.
Qusai Azzam added: " What Are the Features Should I Use For the Customer , Does it Affect the Method I use ?" I would say that the data model used to represent information and relations dealing with the customer and its environment should be adapted to the kind of method / processing that will be later used. But this can be often progressively adapted to the needs. For example, making an object-oriented model of the customer / environment which will enable the writing of rules to select specific customers with specific properties can be an incremental process. In any case you need to check that data required to perform reasoning (rule-based / case-based) will be properly modeled and available at runtime, and / or that the system will overcome the lack thereof without major runtime breakdown.
Best,
Patrice.
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining