Credit score establishing method based on machine learning and big data technology

Abstract

The invention discloses a credit score establishing method based on a machine learning and big data technology. The method is characterized by specifically including the following steps of establishing unified user ID of credit and loan subjects, extracting and preprocessing data of the credit and loan subjects with the unified user ID to obtain training sample data, establishing a credit risk model through a machine learning classification algorithm-integration tree model, obtaining the risk probability according to the credit risk model, automatically converting the risk probability into a credit risk score. The efficient and accacute integration and fusion of global multidimensional big data of the credit and loan subjects are realized through an ID-Mapping technology, the global data of the credit and loan subjects is provided for the establishment of the credit risk model, the quantitative credit risk analysis is conducted on the credit and loan subjects through the machine learning and big data technology on the basis, and therefore the financial risk management capacity is improved and the credit and loan risks are reduced.

Description

technical field [0001] The invention relates to the technical field of financial risk control, in particular to a credit score construction method based on machine learning and big data technology. Background technique [0002] At present, my country's financial reform continues to deepen, and inclusive finance represented by Internet finance is experiencing explosive growth. In 2015, the scale of my country's consumer credit reached 19 trillion, a year-on-year increase of 23.3%. According to a third-party authoritative report, it is expected to reach 41.1 trillion in 2019. Behind the trend, on the one hand, there is a huge base of people who have not been served by traditional finance, and there is a long-term lack of financial products. Therefore, inclusive finance is just needed, with huge scale and potential; Credit efficiency, and the era of data explosion has greatly reduced the cost and difficulty of acquiring massive data. On top of this, it is not only possible to ...

AI Technical Summary

Problems solved by technology

[0004] However, in actual operation, due to the natural huge difference between Internet data (such as behavior, e-commerce, social networking, etc.) and traditional credit data (such as credit records, bank records, real estate certificates, etc.), traditional financial risk data technology It is often difficult to effectively extract the risk value of new Internet data, let alone support the high concurrency and real-time financial business needs under inclusive finance. The specific difficulties are broken down as follows:

[0005] (1) Data fusion is difficult. Data generally come from multiple channels and systems, are heterogeneous and heterogeneous, and come in various forms, such as text, time series, images, etc. It is difficult to connect data;

[0006] (2) It is difficult to use data. Due to the great increase in data complexity, and the characteristics of unstructured, low saturation, and sparseness, manually defining features is generally time-consuming and labor-intensive, and the efficiency is low;

[0007] (3) Data risk modeling is difficult. Thousands or even tens of thousands of dimensional variables are often generated after feature processing, which is far beyond the processing capabilities of traditional risk control modeling based on LR and scorecard systems. More cutting-edge machine learning algorithm processing is urgently needed Corresponding features;

[0008] (4) Model integration is difficult. Because a single model may have unstable performance, it is often necessary to integrate different models to enhance stability and generalization capabilities. Traditional methods lack corresponding exploration and verification;

[0009] (5) Data chain integration is difficult. From data access, preprocessing, feature processing to risk modeling and iteration, a complete closed-loop system with continuous optimization is formed, and it can be quickly migrated and reused into different financial services to achieve practical results. Accumulated and polished over a long period of time

Images