📈 FinPredict: End-to-End Stock Price Direction Prediction Using Machine Learning & Deep Learning

A comprehensive data science project covering EDA, feature engineering, classical ML, and neural networks — all applied to real financial time-series data.

---

🧭 Project Overview

This project aims to predict whether a stock’s closing price will go UP or DOWN the next trading day using historical OHLCV (Open, High, Low, Close, Volume) data. We follow a full pipeline:

1. ✅ Exploratory Data Analysis (EDA) – Understand data structure, trends, and seasonality
2. 🧹 Data Preprocessing & Feature Engineering – Create technical indicators, handle missing values, create lag features
3. 📊 Statistical Analysis & Visualization – Correlation, stationarity tests, returns distribution
4. 🤖 Classical Machine Learning – Train Logistic Regression, Random Forest, XGBoost models
5. 🧠 Deep Learning (LSTM/GRU) – Build sequence models for time-series forecasting
6. 📈 Model Evaluation & Comparison – Compare performance across models using time-based splits

---

📦 Dataset

We use the S&P 500 Stock Prices (2010–2024) dataset from Kaggle.

Columns:

• date – Trading date
• open, high, low, close – Daily price levels
• volume – Number of shares traded
• Name – Ticker symbol (e.g., AAPL, MSFT)

💡 Note: Your Excel screenshot shows all_stocks_5yr.csv — this is likely the same dataset. We’ll load it into pandas for analysis.

---

🛠️ Tech Stack

• Language: Python 3.9+
• Libraries:
  • pandas, numpy → Data handling
  • matplotlib, seaborn, plotly → Visualization
  • scikit-learn → Classical ML
  • tensorflow / keras → Neural Networks (LSTM)
  • statsmodels, scipy → Statistical tests
  • ta (Technical Analysis library) → Feature engineering

---

📁 Project Structure

FinPredict/
│
├── data/                   # Raw and processed datasets
├── notebooks/              # Jupyter notebooks for each phase
│   ├── 01_eda.ipynb
│   ├── 02_preprocessing.ipynb
│   ├── 03_statistics_visualization.ipynb
│   ├── 04_machine_learning.ipynb
│   └── 05_neural_networks.ipynb
├── models/                 # Saved trained models
├── utils/                  # Helper functions (feature engineering, plotting, etc.)
├── README.md               # This file
└── requirements.txt        # Python dependencies

---

🚀 Getting Started

1. Clone this repository
2. Install dependencies:

   pip install -r requirements.txt

3. Download the dataset from Kaggle and place it in data/all_stocks_5yr.csv
4. Open notebooks/01_eda.ipynb to begin!

---

📌 Key Insights (To Be Filled After Analysis)

• Stationarity of returns?
• Most predictive features?
• Best performing model? (XGBoost vs LSTM)
• Accuracy on test set?

---

📝 Future Work

• Add fundamental data (P/E ratio, EPS) from Yahoo Finance
• Try Transformer models for multi-stock forecasting
• Deploy model via Streamlit or FastAPI
• Backtest trading strategy based on predictions

---

🙋‍♂️ Author

Madhav Madupu | [LinkedIn/GitHub] | Date: December 10, 2025

---

🌟 Built for learning, portfolio showcase, and real-world finance applications.