Method and device for predicting rate of return

Abstract

The invention provides a method and device for predicting a rate of return. The method comprises the following steps of establishing an acoustic model and a language model for the call voice between aloan overdue client and a customer service; recognizing the call voice into a call text; performing text mining on the call text to identify the overdue reasons, repayment capability, repayment willingness, special requirements and other multi-dimensional information of rate of return overdue customers; modeling the rate of return by using the information mined by the call text and the call behavior information; and predicting a future rate of return of the user to enable the differentiated collection.

Description

technical field [0001] The present disclosure relates to information processing and application, and in particular, to a method and device for predicting the recovery rate. Background technique [0002] Since 2015, with the rise of mutual funds and cash loans, the collection market has undergone earth-shaking changes. For many emerging financial companies, the recovery rate has replaced compliance as the number one indicator, the industry entry threshold has been greatly reduced, and due to the large amount of overdue mutual funds and cash loans, the market has become more sensitive to collections Demand has increased dramatically. [0003] However, traditional collection rate models for collection ignore call content and call behavior information. The traditional repayment rate model mainly uses the customer's personal information before the loan and overdue conditions to model the repayment rate. After the customer is overdue, especially when the overdue time is long, t...

AI Technical Summary

Problems solved by technology

[0005] (1) The customer service needs to extract and record customer information after talking with the customer, which reduces the work efficiency of the customer service;

[0006] (2) It is a subjective task for customer service to extract customer information from calls, and it is difficult to guarantee the accuracy and consistency of manual labeling;

[0007] (3) The information expressed by the customer during the call is rich and multi-dimensional, and it is difficult for the customer service to record all the valid information, and it is very easy to omit and lose valid information; and

[0008] (4) In order to ensure the accuracy and consistency of customer information extracted from calls by customer service, it is necessary to arrange pre-job training for customer service, and may also need to increase manual quality inspection, all of which increase costs

Images