royshan / ml_hft Goto Github PK

This project proposed a machine learning (Logistic Regreesion and SVM) framework to predict price movement in a high frequency environment and build trading strategy based on this prediction.

###The project has three main parts:
#####Part 1: Data processing and visualization

• Data for this project is the limit order book of E-mini S&P500 future. A sample data is in sample_data folder.
  
• Training set is created at every trade entry with attributes calculate from previous book entry.
  
• Target value is the directional movement of next trade price.
  
• 'ggplot' package is used for visualization

#####Part 2: Model training and testing

• Logistic Regression with nearly 78% accuracy
• Support Vector Machines with nearly 87% accuracy

#####Part 3: Iceberg Detection

• Introduce the basic idea of iceberg order detection algorithm
• Found more than 11,000 iceberg orders within one day