Stock Market Trend Prediction Using Machine Learning and High Order Information of Time Series

Given the financial timeline series such as the S&P 500, and any historical data on the stock markets, how can we find useful information in the latest transaction data to predict ups and downs in the future? Recent work on this problem shows the first evidence that machine learning methods are able to detect dependence (non-linear) on the stock market price track. However, due to the high volatility and the volatile nature of the stock market, predicting the timing of a series of financial periods remains a major challenge. In this paper, we have introduced a new way to simplify a temporary financial system full of sound through successive reconstruction by leveraging motifs (common patterns), and then use the neural convolution network to capture the local structure of the time series. The test results show the effectiveness of our proposed method in feature learning and over-performance with 4% _7% accuracy improvement compared to conventional signal processing methods and how to mimic typical trading patterns with machine learning in stock trend predictions.