Analysis of stock market volatility: Adjusted VPIN with high-frequency data

The volume-synchronized probability of informed trading (VPIN) is widely accepted as a proxy of volatility in the high-frequency market. We propose a novel VPIN model, called Adjusted VPIN, to improve the performance of VPIN so that it can directly analyze and better predict the information asymmetry of individual stocks. We extend the VPIN model by optimizing the classification algorithm with a neural network method and high-frequency data. Both trading volume and trends are considered to capture stock volatility. Empirical results on three different trading volume groups generate a 37.86% higher relevant result with logarithm stock yield than the VPIN model.