About The Project

This repository provides Python codes used to perform all calculations reported in the paper: Technical Analysis with Machine Learning Classification Algorithms: Can it still ‘Beat’ the Buy-and-Hold Strategy?

Pre-requisites

• Anaconda Distribution with Python 3.9.21

Calculate rolling-window forecasts

For example, the following script will calculate four-period-ahead forecasts using the rolling-window strategy for BTC-USD using Dataset I as predictors

python ./RF_BTC_all_vars_CE_tau_4.py > /dev/null 2>&1 &

Implement a Trading Strategy

For example, the following script will implement the proposed trading strategy based on Random Forest forecasts and all the performance metrics for BTC-USD using Dataset I as predictors

python ./perf_report_RF_BTC_all_vars_v1.py > /dev/null 2>&1 &

Calculate Statistics of each Performance Metrics

For example, the following script will calculate the medians, IQRs, means, standard deviations, t-statistics of performance metrics of the trading strategy based on Random Forest forecasts for BTC-USD

python ./perf_stats_all_invest_pers_v1.py > /dev/null 2>&1 &

Implement Bootstrap Reality Check

For example, the following script will compare all trading methods based on LightGBM forecasts for SPY

python ./bootstrap_RC.py

Plot Figures

For example, the following Jupyter notebook will plot all the figures in the paper

plots_v1.ipynb

Download Data from Yahoo Finance, Fred, Nasdaq Data Link (API keys are required)

downloadData.ipynb

List of Main Files

Python main file	Description
trading_strategies.py	Python functions to implement the proposed trading strategies
talib_technical_indicators.py	Python functions to calculate technical indicators and candlestick chart patterns
performance_measures.py	Python functions to calculate performance measures
custom_losses_v1.py	Python functions to define custom loss functions and scoring functions
compare_models.py	Python function to implement the bootstrap reality check to compare trading methods
forecast_directional_movement_v4.py	Python function to implement the rolling-window strategy to predict future pricing moving directions
my_classification_algorithms_optuna_v4.py	Python functions to implement and cross validate all ML classification algorithms

License

Distributed under the MIT License. See LICENSE.txt for more information.

Contact

Ba Chu - [email protected]

Project Link: https://github.com/wave1122/trading_strategies