Information theory based feature selection for customer classification
The application of Information Theory techniques in customer feature selection is analyzed. This method, usually called information gain has been demonstrated to be simple and fast for feature selection. The important concept of mutual information, originally introduced to analyze and model a noisy...
| Autor principal: | |
|---|---|
| Outros Autores: | , , |
| Formato: | conferenceObject |
| Idioma: | eng |
| Publicado em: |
2021
|
| Assuntos: | |
| Texto completo: | http://hdl.handle.net/10071/23251 |
| País: | Portugal |
| Oai: | oai:repositorio.iscte-iul.pt:10071/23251 |
| Resumo: | The application of Information Theory techniques in customer feature selection is analyzed. This method, usually called information gain has been demonstrated to be simple and fast for feature selection. The important concept of mutual information, originally introduced to analyze and model a noisy channel is used in order to measure relations between characteristics of given customers. An application to a bank customers data set of telemarketing calls for selling bank long-term deposits is shown. We show that with our method, 80% of the subscribers can be reached by contacting just the better half of the classified clients. |
|---|