Overdue Prediction of Bank Loans Based on LSTM-SVM

Abstract

In the aspect of bank loans, the accuracy of traditional user loan risk prediction models, such as KNN, Bayesian, DNN, are not benefit from the data growth. This article is based on the work of Overdue Prediction of Bank Loans Based on Deep Neural Network. And we propose to analyze the dynamic behavior of users by LSTM algorithm, and use the SVM algorithm to analyze the user's static data to solve the current prediction problems. This article uses users' basic information, bank records, user browsing behavior, credit card billing records, and loan time information to evaluate whether users are delinquent. These static data are the basic input for SVM. For LSTM model, we extract user's recent transaction type from browsing behavior as input to LSTM, to predict the probability of users' overdue behavior. Finally, we calculate the average of the two algorithms as the final result. From the experimental results, this LSTM-SVM model shows a great improvement than traditional algorithms.

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 — 13 citations exist upstream.