⚡ CryptoPulse: Multimodal Sentiment & Price Analysis

A "Full Marks" standard Computational Data Science project predicting cryptocurrency price movements by fusing quantitative market data with qualitative sentiment analysis.

---

📸 Dashboard Preview

The actual interactive CryptoPulse Dashboard running on localhost.

---

🚀 Key Features

1. Multimodal Analysis

We go beyond simple price prediction by integrating three distinct data sources:

• Quantitative: Historical OHLCV Data (Open, High, Low, Close, Volume) from Yahoo Finance.
• Qualitative: News Sentiment Analysis using NLP (TextBlob) on financial headlines.
• Psychological: Real-time Crypto Fear & Greed Index (0-100) from alternative.me.

2. Advanced Deep Learning

• Model: Stacked LSTM (Long Short-Term Memory) Network.
• Framework: PyTorch.
• Input: A fused vector $x_t = [Price_t, Vol_t, Sentiment_t, FNG_t]$.

3. Premium User Interface

• Built with Streamlit and Plotly.
• Dark-mode, responsive layout.
• Interactive Date Pickers and Zoomable Charts.
• Live Inference capability.

3. Premium User Interface

• Built with Streamlit and Plotly.
• Dark-mode, responsive layout.
• Interactive Date Pickers and Zoomable Charts.
• Live Inference capability.

---

🏗️ System Architecture

graph TD
    %% Data Collection Layer
    subgraph Data_Collection [1. Data Collection]
        style Data_Collection fill:#2d3436,stroke:#dfe6e9,color:#dfe6e9
        A["Yahoo Finance API<br/>(OHLCV)"] -->|Raw Prices| D("Data Merger")
        B["TextBlob NLP<br/>(News Mock)"] -->|Sentiment Score| D
        C["Alternative.me API<br/>(Fear & Greed)"] -->|FNG Index| D
    end

    %% Preprocessing Layer
    subgraph Preprocessing [2. Preprocessing]
        style Preprocessing fill:#2d3436,stroke:#00cec9,color:#dfe6e9
        D --> E{"Alignment & Cleaning"}
        E -->|Fill NaNs| F[Feature Engineering]
        F --> G["Normalization<br/>(0-1 Scaling)"]
        G --> H["Train/Test Split<br/>(80/20)"]
        H --> I["Windowing<br/>(60-day Lookback)"]
    end

    %% Model Layer
    subgraph Model [3. Deep Learning]
        style Model fill:#2d3436,stroke:#fdcb6e,color:#dfe6e9
        I --> J["Stacked LSTM<br/>(2 Layers, 32 Units)"]
        J --> K["Fully Connected Layer"]
        K --> L("Price Prediction")
    end

    %% Application Layer
    subgraph App [4. User Interface]
        style App fill:#2d3436,stroke:#ff7675,color:#dfe6e9
        L --> M["Streamlit Dashboard"]
        F --> M
        M --> N["Interactive Plots<br/>(Plotly)"]
        M --> O["Data Export<br/>(CSV)"]
    end

Loading

---

🛠️ Installation & Setup

Prerequisites

• Python 3.8+
• Git

Quick Start

1. Clone the Repository

   git clone https://github.com/nizamkadirteach/cdsproject.git
   cd cdsproject

2. Install Dependencies

   pip install -r requirements.txt

3. Run the Dashboard

   streamlit run src/app.py

4. Run System Verification (Optional)

   python src/test_system.py

---

📂 Project Structure

📦 cdsproject
 ┣ 📂 data/               # Raw and Processed Datasets
 ┣ 📂 docs/               # Documentation & Images
 ┣ 📂 notebooks/          # Jupyter Notebooks (EDA & Experiments)
 ┃ ┗ 📜 01_naive_model_week8.ipynb
 ┣ 📂 report/             # LaTeX Project Report
 ┃ ┗ 📜 main.tex
 ┣ 📂 slides/             # Presentation Slides (Beamer)
 ┃ ┣ 📜 initial_presentation.tex
 ┃ ┗ 📜 final_presentation.tex
 ┣ 📂 src/                # Source Code
 ┃ ┣ 📜 app.py            # Streamlit Dashboard Entry Point
 ┃ ┣ 📜 data_loader.py    # Data Fetching Logic (APIs)
 ┃ ┣ 📜 model.py          # PyTorch LSTM Model Definition
 ┃ ┣ 📜 preprocessing.py  # Feature Engineering & Merging
 ┃ ┗ 📜 test_system.py    # System Verification Script
 ┣ 📜 .gitignore
 ┣ 📜 README.md
 ┗ 📜 requirements.txt

---

👥 Team Members

Name	Role	ID
Alice Zheng	Data Engineer	100XXXX
Bob Chen	ML Engineer	100XXXX
Charlie Davis	Frontend Dev	100XXXX
Diana Lim	Project Lead	100XXXX

---

📄 License

This project is open-source and available under the MIT License.