Research on Deep Learning Model Architecture in Financial Big Data Analysis

This paper proposes a multi-task learning adaptive reinforcement convolutional neural network (MTL-RCNN). The model combines the advantages of multi-task learning and adaptive convolutional layers, and dynamically adjusts the convolution kernel through reinforcement learning to improve the learning ability of the model when processing financial big data. The model framework enhances the synergy between multiple tasks by sharing the underlying convolutional layer and designing independent task-specific optimization layers for each financial task (such as risk prediction, market analysis, credit scoring, etc.), thereby improving the overall performance. Reinforcement learning is adopted to optimize the size of convolution kernel and other parameters in order to improve the precision and convergence rate of the model. To test the validity of this model, the author makes experiments on the historical data of the public stock market and the loan default data. Experiments show that MTL-RCNN is better than the traditional deep learning models in multi-evaluation. In particular, in the stock market risk prediction task, the model's accuracy reaches 91.2%, which is nearly 7 percentage points higher than the 84.3<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> of the traditional LSTM model. At the same time, the training time and computing resource consumption of the model have also been effectively optimized, proving its high efficiency in financial big data analysis.