CRYPTOCURRENCIES FINANCIAL DATA ANALYSIS AND FORECASTING

This project aims to analyze and forecast the future prices of 10 different cryptocurrencies by utilizing historical financial data collected from the Binance API. A web-based interface is developed using Python and Flask to visualize and interact with the predictions.

🔄 Project Workflow

1. Data Collection
   Historical price data for 10 cryptocurrencies were pulled directly from the Binance API and saved as .csv files.

2. Data Preprocessing

   • Handled missing values carefully to ensure data quality
   • Calculated and integrated various technical analysis indicators
   • Scaled data using StandardScaler
   • Prepared time-series input windows for the LSTM model (window size: 21 days)
   • Configured forecasting horizon to predict prices 7 days ahead

3. Data Analysis

   • Explored price trends, volatility, and correlations
   • Visualized data patterns to support model selection

4. Price Forecasting
   Applied three regression-based machine learning models to predict future prices:

   • Linear Regression
   • XGBoost Regressor
   • LSTM (Long Short-Term Memory) deep learning model

   Models were evaluated with metrics including MAE (Mean Absolute Error), RMSE (Root Mean Square Error), and R² Score. The LSTM model showed strong performance in capturing temporal patterns in cryptocurrency prices.

📊 Dataset Features

Each dataset contains 6 main features:

Feature	Description
Date	Timestamp of the price record
Open	Opening price at the start of period
High	Highest price during the period
Low	Lowest price during the period
Close	Closing price at the end of period
Volume	Trading volume in that period

Datasets are stored in .csv format and enriched with technical indicators during preprocessing.

📁 Included Datasets

• ADA-USD.csv (Cardano)
• BNB-USD.csv (Binance Coin)
• BTC-USD.csv (Bitcoin)
• DOGE-USD.csv (Dogecoin)
• DOT-USD.csv (Polkadot)
• ETH-USD.csv (Ethereum)
• LINK-USD.csv (Chainlink)
• LTC-USD.csv (Litecoin)
• XMR-USD.csv (Monero)
• SOL-USD.csv (Solana)

🌐 Web Interface

A user-friendly web interface is developed with Flask, providing:

• Interactive visualization of actual vs. predicted cryptocurrency prices
• Upload and selection of cryptocurrency datasets
• Easy navigation to explore different cryptocurrencies

The main web page (index.html) is located in the templates/ folder.

🧠 Machine Learning Models

Model	Purpose
Linear Regression	Baseline regression model
XGBoost Regressor	Handles non-linear relationships
LSTM Neural Network	Captures time-series dependencies and trends

The LSTM model uses a window size of 21 days (LSTM_WINDOW = 21) to look back on historical data and predicts prices 7 days into the future (FUTURE_STEPS = 7).

🛠️ Technologies & Libraries

• Python
• Flask (Web framework)
• Binance API (Data collection)
• Pandas, NumPy (Data manipulation)
• Matplotlib, Seaborn (Visualization)
• Scikit-learn, XGBoost (Machine learning)
• TensorFlow/Keras (Deep learning with LSTM)

👥 Team Members

• Burak SAYAR – GitHub Profile
• Mert BUYUKNISAN – GitHub Profile
• Rahime GEDİK – GitHub Profile
• Zeynep COL – GitHub Profile