Bank Financial Risk Prediction Model Based on Big Data

Abstract

Financial risk prediction is an important technique to systematically predict the unforeseeable risks in banking systems. The issues involving ill-timing and low accuracy in the current risks prediction methods necessitate an effective risk prediction method. Akin to the use of big data in various domains, the technology has a significant role in financial services and can be used to accurately and timely predict the possibilities of risks. In this paper, an effective hybrid method is proposed to aptly and effectively predict financial risks in the banking systems. The method utilizes the Lasso and linear regression algorithms via the big data features and framework technologies. By proper formalization of the bank financial risk problems, the risk data is obtained and processed. To filter the initial text features and preprocess the annual report text data, the information gain method is used. With the Bag-of-Words (BoW) and the word frequency reverse document frequency weighting method, the text features of financial risk prediction are extracted. The bank financial risk prediction model is constructed based on weighted fusion adaptive random subspace algorithm. The prediction results obtained are integrated so as to realize the bank financial risks in a seamless way. The experimental results show that the proposed method can effectively improve the prediction accuracy and consumes comparatively lesser time in risk prediction.