Factors Selection for Stock Predicting Based on Vertical and Horizontal Analysis of Technology Giants

The stock prediction with machine learning algorithms has been widely known and accepted by the investors and institutions. Whereas, the significances of the factors chosen in the machine learning based prediction are crucial. This article focuses on selecting the significant factors predicting the stock price and volatility from based on the Long short-term memory and Random Forest model. First of all, four years of the data of some technology giants’ stock, industry indexes, market indexes and settle price of the future are taken into consideration and prepared in advance. By creating the time series data and adding the data sets of some other possible factors, the research combines the vertical and horizontal analysis to calculate the importance of the features to forecast either return or volatility. The main factors analysis is used to select the feature, tune the day of time series and improve the model. Last but not least, the Principal Component Analysis and Grey Model are regarded as the farther research field. According to the results, 5 days of time series performs best in both predicting the return and volatility. Apart from that, the index value of Standard and Poor 500 Index Technology Plate and the average of the return are most significant for return prediction, when the standard deviation of Amazon, Microsoft, SPLRCT and USTEC are most important features for volatility. The future research direction lies in the theoretical proof of the effectiveness of the combination of the horizontal and vertical analysis and more empirical research will be done to enhance the confidence of the result. These results shed light on the most significant factors of the return and standard deviation and obviously hence the accuracy of the prediction.