Customer Churn Prediction Based on Feature Clustering and Nonparallel Support Vector Machine

Abstract

Bank customer churn prediction is one of the key businesses for modern commercial banks. Recently, various methods have been investigated to identify the customers who would leave away. This paper proposed a new framework based on feature clustering and classification technique to help commercial banks make an effective decision on customer churn problem. The proposed method benefits from the result of data explorations, clusters the customer features, and makes a decision with a state-of-the-art classifier. When facing the data with large amount of missing items, it does not directly remove the features by some subjective threshold, but clusters the features through the consideration of the relationship and the missing ratio. Real-world data from a major commercial bank of China verifies the feasibility of our framework in industrial applications.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 14 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.