Profit Prediction and Mechanism Analysis of Stock Price-Volume Patterns Based on Random Forest

Abstract

This paper collects 25,598 data entries of stocks that experienced a certain magnitude of price decline, followed by volume contraction and subsequent volume expansion, starting from 2008. It investigates the impact of various influencing factors on the subsequent profitability of stocks after volume expansion. First, an indicator system covering trading volume, market index, K-line patterns, and financial metrics is constructed. A random forest model is then established to identify the top 11 important features among 52 features. These top features include the 1-2 day market index and market profit index, the 0-2 day price change rate, the price decline relative to the highest price in the previous 60 trading days, the 1-2 day K-line entity ratio, and the volume ratio. Second, the forward selection method is used to calculate the AUC_ROC values as the number of features increases from small to large. It is found that the AUC_ROC reaches the first peak when the number of features is 3-4 core technical indicators. After that, the AUC_ROC decreases due to the addition of noise information, and then rises to the second peak with the introduction of financial indicators. Finally, the SHAP method is applied to analyze the marginal contribution of the features included in the second peak. The results show that within a certain range, the marginal contributions of K-line entity, volume ratio, price decline, earnings cash coverage multiple, and interest coverage multiple to the future return rate are positive; however, when these indicators are too high or too low, their marginal contributions turn negative.