📈 Stock Price Classifier

This project aims to predict whether a stock’s closing price will increase or decrease on the next day using logistic regression implemented from scratch in Python. It is part of a broader roadmap focused on machine learning applications in finance and internship preparation for programs like GSoC and quantitative research roles.

---

🚀 Project Goals

• Predict binary stock movement (up/down) using past price data and technical indicators
• Implement logistic regression from scratch (no ML libraries)
• Visualize learning progress (loss, accuracy)
• Build reproducible, modular codebase for extension (e.g., regression models, sentiment analysis)
• Strengthen experience in ML for finance domain

---

🛠️ Tech Stack

• Python (core logic)
• NumPy (matrix ops)
• Pandas (data preprocessing)
• Matplotlib (visualization)
• scikit-learn (only for evaluation metrics)

---

📊 Results Summary

• ✅ Training Accuracy: ~55%
• ✅ Test Accuracy: ~51–52%
• 📉 Custom learning curves (loss/accuracy) show convergence
• ⚠️ Class imbalance and market noise present natural challenges

---

📁 Project Structure

stock-price-classifier/

• data/ # Raw & processed stock data (CSV)
• src/ # Custom ML modules & training pipeline
• models/ # Saved weights (theta.npy)
• results/ # Accuracy & loss plots
• README.md
• requirements.txt

---

🧠 Core Implementation

• ✅ logistic_model.py: Logistic Regression from scratch (no sklearn!)
• ✅ train.py: Trains over multiple learning rates, logs accuracy/loss
• ✅ evaluate.py: Loads saved weights and reports classification metrics
• ✅ data_loader.py: Loads and processes CSV stock data

---

📈 Sample Outputs

Plots saved in /results/ folder:

• accuracy_vs_epochs_lr_0p001.png
• loss_vs_epochs_lr_0p001.png
• lr_vs_accuracy.png

(Example plot: Accuracy vs Epochs at different learning rates)

---

✅ How to Run

# 1. Clone the repo
git clone https://github.com/CodeHermit17/stock-price-classifier.git
cd stock-price-classifier

# 2. Install dependencies
pip install -r requirements.txt

# 3. Train model
python src/train.py

# 4. Evaluate best saved model
python src/evaluate.py