Hybrid Deep Learning for AI-Based Financial Time Series Prediction

Abstract

As the influencing factors of the development of things are not clear or the data collection is difficult, many forecasting problems have evolved into univariate time series forecasting problems, that is, mining its own inherent laws from the past time series data, so as to predict its future development trend. The purpose of this study is to explore the feasibility of using random forest classifier to predict the long-term trend of stocks. By analyzing data sets such as Apple, Samsung, and General Electric, we built a random forest model and found that its prediction accuracy was 85 to 95 percent. With the increase of the number of decision trees, the prediction results of the model tend to be stable, indicating that increasing the number of decision trees can improve the prediction accuracy. In addition, we point out that the method is also suitable for short-term trend forecasting, and suggest training with more fine-grained transaction data to improve forecasting accuracy. Finally, we look forward to the potential of the field of artificial intelligence in time series forecasting, emphasizing that the application of technologies such as deep learning will further improve forecasting accuracy and provide more reliable decision support for financial investors.