Predicting Customer Bank Card Operations

Predicting Customer Bank Card Operations

by from Herzliya Pitu"akh, Israel

Design

Tech

Usability

Creativity

Content

Type: Software
Client from Russia
Category: Finance
Style: Big photo
Color: Black

    Goal: to develop prediction algorithms:1. Total customer spending for the next 30 days for each service/product category2. Individual customer spending in the next month for each service/product categorySource data: 7 mln card transactions by 15K customers, 184 categories of services/product.Result:• For Goal 1, an ensemble of models was a solution: time series prediction, linear regression, decision trees.• For Goal 2, we built a model based on statistics for prior periods and extended statistics on various combination of attributes.