2018 BAI Global Innovation Award Finalist

BAI Global Innovation Awards 2018 finalist Ping An Technology

Best Application of Data Analytics, AI or Machine Learning in a Product or Service

Ping An Technology – China
Emotion Recognition Based Financial Risk Management System

Financial risk control relies on the credit analyst’s grip on the authenticity of the customer’s information. In the face-to-face interview, the credit analyst has to judge the credit risk through his dialogue with the customer based on his past experience, which is energy-consuming and error-prone. Our solution utilizes the results of psychological study, adopts AI methods, and extracts the abnormal performance of people when they tell lies to provide assistant judgment for the credit review process. It lowers the professional threshold of credit analysts, improves the efficiency of credit review, reduces the bad debt ratio, and realizes AI+banking.

Ping An Team Photo The Problem

Risk control is required by bank credit, insurance audit and so on, but traditional financial risk control relies on the credit analyst’s grip on the authenticity of the customer’s information. The mining and prevention of risks largely depend on the credit analyst’s work experience. In the traditional credit review process, the credit analyst not only needs to make preparations before the interview, but also needs to respond to the situation during the interview, and sum up after the interview, which requires a lot of labor and easily fatigues the credit analyst, as a result, the information mining is not exhaustive and is error-prone.

The Solution

Using the research results of psychology and the methods of AI, our solution can extract the abnormal emotions of people when they conceal something or tell lies, and capture the subtle changes of their facial muscles to provide assistant judgment for the credit review process.

In the credit review process, our solution can analyze the audio/video information in real time and remind abnormal behaviors. Speech recognition algorithm is used to analyze the conversation according to the semantic meaning. A multimedia model combining image, audio and semantic meaning is used to assess the risk probability in real time. A preliminary report is given at the end of the credit review. It not only reduces the requirement on the credit analyst’s experience, but also reduces the labor intensity in the process of credit review, so that the credit analyst can focus more on the credit review.

The Outcome

The application of the risk control system based on emotion analysis has greatly reduced the workload of credit analysts, the labor cost of risk control, the professional threshold of risk control personnel, and the training cycle of credit analysts. In business, it has simplified the business process, accelerated the speed of business transaction, improved the quality of business transaction, reduced the potential risks, reduced the bad debt ratio, and improved the enterprise’s profit margin. In the process of credit review, customers cannot perceive the existence of the system. It does not change the original business process, and can effectively prevent targeted attacks on the system.