Machine Learning: Technical Trading Strategies
Objective
Our client posed the idea to our team that they wanted to build out a proof-of-concept Machine Learning program for various technical trading strategies in financial markets. The end goal of this program was to dynamically adjust each technical strategy the client was analyzing to their optimized parameters, based on the trained Machine Learning model’s prediction of market price. The program would then live testing for pre-determined amount of time on each chosen market, using the optimized parameters as determined by the ML model.
Client: Financial Data Analytics Firm
Solution
First our team had to build a system to extract the necessary market data for the model and automatically feed the data to the system for training the LSTM model on a daily basis. We built the system to utilize TendorFlow and Keras in Python for training the LSTM model. A C# trading code utilized TensorFlowSharp to pass the live trading data to the model for evaluation and prediction. The system then provided the necessary output and parameters for running live testing.
Project Implementation
- Machine Learning program built with the ability to train and learn on any automated trading strategy fed into the model.
- Program hosted on cloud-based server with ability to run on a 24x7 basis, providing client opportunity to test markets such as crypto and test longer term trading strategies on a fully automated basis.
- User features built into the program allowing for CSV output and various charting capabilities.
- Simple reporting output features, allowing for quick analysis and validation/in-validation of model on chosen trading strategy.
Technologies Used
