Research on Factors Affecting Financial Stock Price Fluctuations based on Improved Convolutional Neural Network

This paper studies the factors affecting the volatility of financial stock prices, inputs stock related information into the convolutional neural network on the basis of understanding the convolutional neural network, completes the model training, and obtains the neuronal threshold and the gradient of connection weight of the whole convolutional network through calculation. A convolutional neural network model improved by genetic algorithm is constructed, and the factors affecting stock price changes are predicted by this model. It is found that the significance of five factors, such as trading volume and exchange rate, is the highest, and all pass the bilateral significance level test of 0.1. Five factors, such as the growth rate of industrial added value and retail sales of social consumer goods, were identified as "strong indicators". When the stock price fluctuation is greater than 0.05, the prediction accuracy of this model reaches 100%. Therefore, the method in this paper can accurately determine the main factors affecting stock price volatility.