v1.0.0
Release date: Mar 6, 2023
The first GitHub release for this project introduces a new labeling method called "Long Reward Labeling" for training supervised machine learning models on stock market and cryptocurrency market data. The traditional labeling method based on price movement of future price changes has low prediction accuracy in many machine learning models due to the chaotic nature of the cryptocurrency market. The project aims to utilize the concept of the Bellman equation to improve the prediction accuracy of supervised machine learning methods and bring reinforcement learning alike results to simpler and less expensive methods. The labeling method determines the reward of taking a long position at a given time by calculating trend reversal points. This has shown being able to deliver a profitable trading strategy across different time horizons and time frames.
Features: Data fetching library, Long reward labeling