China Market Impact Cost Model Based on Machine Learning

Abstract

With the rapid development of quantitative finance, trading is becoming faster and more frequent. But it also comes with more trading cost. Impact cost is an inevitable cost in stock trading, as the implicit costs, it cannot be measured directly. In this paper, we simulate market trading using 3 second snapshot in the limit order book of CSI300. Next, we calculate and analyze impact cost including both permanent impact cost and temporary impact cost. Then we predict permanent impact cost and temporary impact cost using linear regressor and machine learning models. As a result, we find out decision tree regressor performs best, far better than previous paper. It is meaningful to study the impact cost for we can reduce it and increase profits.