Using Deep Residual Networks to Deal with Financial Risk Control Problems
Miao Liu, Mengxing Huang, Yu Zhang, Wenlong Feng, Junxiao Lai, Xinze Li · 2018
Recently, the use of relevant knowledge in the field of machine learning to process high dimensional feature data and solve related problems has become a mainstream method. The use of blending ensemble and stack ensemble techniques has improved the accuracy of predictions, which is a bottleneck problem due to the difficulty of handling high dimensional feature. This paper adopts the deep residual network (ResNet 50) to help the processing of complex features of financial data to replace the traditional data processing methods, and use the bagging method to complete the final classification.The data used in this article is the data of the 2017 Tianchi Competition. The final AUC value of our algorithm is better than the best score of 0.2%.