ForeCoin

This is the implementation of project 4.2: Financial Advisor Bot.

This web application is a real-time cryptocurrency analysis tool that uses artificial intelligence to provide market insights in order to make recommendations for a dynamic investment strategy.

---

About the Bot

The analysis tool uses artificial intelligence to provide market insights. It combines two primary AI models: a sentiment analysis model that scans live news to gauge market mood, and a time-series forecasting model (Chronos-T5) to predict future price movements. Navigate to the Dashboard for a market overview, or visit the "Stable" and "Volatile" coin pages for detailed, asset-specific predictions.

The bot currently works with 5 cryptocurrencies

Stable Coins: BTCUSDT, ETHUSDT, BNBUSDT

Volatile Coins: DOGEUSDT, SHIBUSDT

---

🖥️ Getting Started 🏃

When forking this repository, download the following folders from ForeCoin's Additional Folders

As some files are relatively big, not all files are loaded into this github repository.

List of folders that can be downloaded:

• historic_data
• apps
• prediction_logs
• pycache

How to Run

Run the requirements.txt file

• pip install -r requirements.txt

then run the following

• pip install --upgrade --force-reinstall "feedparser>=6.0.10"

(As pygooglenews strictly requires feedparser to be less than version 6.0.0, we would need to forcibly upgrade it)

---

The Machine Learning Models

KNN With Supertrend

We monitor major cryptocurrencies including Bitcoin (BTC), Ethereum (ETH), and Binance Coin (BNB). These stable assets are analyzed using specialized KNN Supertrend models optimized for less volatile market movements.

LGBM with Quantile Regression

For high-volatility cryptocurrencies like Dogecoin (DOGE) and Shiba Inu (SHIB), we employ LGBM Quantile models that provide low, median, and high prediction ranges to account for their unpredictable price movements.

Chronos T5

The Chronos-T5 time-series forecasting model analyzes historical price data to predict future cryptocurrency prices. Combined with traditional machine learning models like KNN and LGBM, we provide multiple prediction perspectives for enhanced accuracy.

FinBERT for Sentiment Analysis

The sentiment analysis component scans cryptocurrency-related news articles and social media posts to determine the overall market sentiment. This helps identify potential market movements based on public opinion and news events.

---

🗃️System Architecture🗃️

ForeCoin/
│
├── apps/
│   ├── static/
│   │   └── assets/
│   │        ├── package.json/
│   |        ├── css/
│   |        ├── img/
│   |        ├── vendor/
│   |        └── js/
│   |            ├── dashboard_updater.js
│   |            ├── index_search.js
│   |            ├── stable_updater.js
│   |            ├── volatile_updater.js
│   |            └── volt.js
|   |
│   └── templates/
│       └── home/
│           ├── index.html
│           ├── dashboard.html
│           ├── stable.html
│           └── volatile.html
│       └── includes/
│           ├── footer.html
│           ├── navigation.html
│           ├── scripts.html
│           └── sidebar.html
│       └── layouts/
│           ├── base-fullscreen.html
│           └── base.html
|
├── historic_data/
│   ├── sentiment (can remove)/
│   ├── stable/
│   ├── volatile/
│   └── *all_coins_chronos_pred.csv*
│
├── trained_models/
│   ├── BNBUSDT_knn_supertrend_model.pkl
│   ├── BTCUSDT_knn_supertrend_model.pkl
│   ├── ETHUSDT_knn_supertrend_model.pkl
│   ├── DOGEUSDT_lgbm_quantile_model.pkl
│   ├── SHIBUSDT_lgbm_quantile_model.pkl
│   └── *all_coins_chronos_pred.csv*
|
├── trained_models/
│   ├── backtest_log.json
│   ├── BNBUSDT_predictions.json
│   ├── BTCUSDT_predictions.json
│   ├── ETHUSDT_predictions.json
│   ├── SHIBUSDT_predictions.json
│   └── dogeUSDT_predictions.json
|
├── app.py
├── backtesting.py
├── data_collect.py
├── models.py
├── predict_chronos.py
├── sentiment_bert.py
├── webscraper.py
├── readme.md
└── requirements.txt