TalkingData has been deeply engaged in the financial field, processing and evaluating massive data in a comprehensive way. The company's self-developed algorithm is based on genetic algorithm feature selection technology, bifurcation semi-supervised learning technology, sample transfer learning algorithm, etc., which is capable of comprehensively processing and evaluating the data of each dimension, deeply understanding the characteristics of the target scenario, integrating the data of users' online behavior, behavior preference on the device, consumption behavior, etc. On this basis, we present the data analysis results, and help partners reduce the bad debt rate among users in the qualification review phase.