Bitcoin Trading Bot Optimization — CITS4404 Project (Group 13)

This repository contains the implementation of a trading bot that uses adaptive strategies to optimize Bitcoin trading performance based on historical data. It integrates nature-inspired optimization algorithms (PSO, ABC, HCA) to tune the parameters of a moving average crossover strategy.

Repository Structure

File / Folder	Description
AI_Project_Integrated_Code.ipynb	Main notebook: This file integrates all components (data loading, bot logic, evaluation, optimization, plotting). Run this notebook to execute the entire experiment.
ABC.py	Implements the Artificial Bee Colony (ABC) optimization algorithm.
PSO.py	Implements the Particle Swarm Optimization (PSO) algorithm, with support for fitness tracking and iterative output.
HCA.py	Implements the Hill Climbing Algorithm (HCA) for local search-based parameter tuning.
Call ABC	Script that calls the ABC optimizer with appropriate parameters and configuration.
Call PSO	Script that calls the PSO optimizer.
Call HCA	Script that calls the HCA optimizer.
Algorithm Comparison	Contains code and results comparing algorithm performance on training and test datasets (e.g., final cash).
bot.py	Defines the trading bot evaluation logic, including moving average computation, buy/sell signal generation, and capital simulation.
data/	Contains historical BTC price data (BTC-Daily.csv) used as input for training and testing.
legacy/	Archive of earlier or alternative versions of evaluation functions and related code.
.gitignore	Specifies which files and folders to exclude from version control (e.g., __pycache__).
README.md	This file — provides project overview, file documentation, and usage instructions.

How to Run the Code

Requirements:

• Jupyter Notebook or any Python environment that supports .ipynb files.
• Python 3.7+
• Standard scientific libraries: numpy, pandas, matplotlib

Steps:

1. Clone this repository.
2. Ensure the BTC price data file (BTC-Daily.csv) is placed inside the data/ directory.
3. Open the file AI_Project_Integrated_Code.ipynb.
4. Run all cells from top to bottom. This includes:
   • Importing data
   • Constructing moving averages
   • Running optimization algorithms (PSO, ABC, HCA)
   • Visualizing trading signals
   • Comparing results

The final section of the notebook will generate training and testing performance plots and metrics for all three optimization algorithms.

Project Scope and Objectives

This project was conducted as part of the CITS4404 Artificial Intelligence Project unit at The University of Western Australia. The aim was not just to produce a profitable trading bot, but to explore the following AI and optimization concepts:

• Understanding bots as pluggable components.
• Applying nature-inspired algorithms to generate and refine strategy instances.
• Using evaluation functions over real-world data to adaptively improve those strategies.
• Experimentally exploring the trade-offs between model flexibility, search space complexity, and algorithm performance.
• Comparing multiple optimization approaches through fair and controlled experiments.

Algorithms Implemented

• Particle Swarm Optimization (PSO)
  Swarm-based optimizer that balances personal and global experience to find high-performing solutions.

• Artificial Bee Colony (ABC)
  Bee-foraging-inspired algorithm that emphasizes neighborhood search, probabilistic selection, and exploration through scout bees.

• Hill Climbing Algorithm (HCA)
  Greedy local search algorithm used as a simple baseline for comparison.

Each optimizer is used to tune 14 continuous parameters of the bot, including moving average weights, window lengths, and smoothing factors.

Contributors

This project was completed by Group 13 members for CITS4404:

• Eusha Khan (24245979)
• Joel Brooker (23409801)
• Stefan Andonov (23374198)
• Wenbo Gao (23335934)
• Yi Ren (23895642)
• Zhi Wang (24560057)