Skip to content

READ.md

Latest commit

 

History

History
42 lines (31 loc) · 1.84 KB

READ.md

File metadata and controls

42 lines (31 loc) · 1.84 KB

📈 Bitcoin Price Forecasting with LSTM & Real-Time Data

📌 This project is part of a larger research study analyzing Bitcoin from two perspectives:

  • 📈 As a Stock – influenced by trading volume, technical indicators, market sentiment
  • 💱 As a Currency – influenced by exchange rates, transaction volumes, and lag time

🧠 This notebook trains an LSTM model on the currency perspective dataset, but the architecture is generalizable across both perspectives and timeframes (2017–2024).

🧠 Objective

  • Clean and unify historical price data from multiple sources
  • Interpolate missing values and remove outliers using IQR
  • Create sequences for LSTM using a rolling window approach
  • Build an LSTM model with tunable hyperparameters via KerasTuner
  • Integrate real-time Bitcoin price fetching (CoinGecko API)
  • Forecast the next Bitcoin price point

⚙️ Key Features

  • 📅 Multi-sheet Excel ingestion with smart merging
  • 🧹 Full pipeline: datetime parsing, median aggregation, IQR-based outlier removal
  • 🔗 Real-time API integration with fallback handling
  • 🔁 Sequence creation for time-series forecasting
  • 🧪 KerasTuner for LSTM architecture optimization

🧾 Files Included

📄 Bitcoins_forecasting_code.ipynb      # Main Jupyter notebook with code
📘 Bitcon_explaination.pdf        # Explanation of methodology and code
📦 requirements.txt                # Required Python libraries
🧠 model_summary.txt               # Architecture and training process
📊 Bitcon_Research_Dataset.xlsx    #Cleaned dataset representing the currency perspective (2017–2024), used for LSTM training.Other versions (stock/time-split) are compatible with this code structure.

README.md                          # Project overview (this file)