Traditional Machine Learning Projects

Description

This repository contains various traditional machine learning projects implemented using Python and relevant libraries such as scikit-learn, pandas, numpy, matplotlib, and PyTorch. The projects focus on solving real-world problems using classical machine learning techniques like regression, classification, and time series forecasting.

---

Environment Setup

All dependencies required to run the projects are listed in requirements.txt.

Installation Instructions

1. Clone the repository:

   git clone https://github.com/gajeraYash/ml-projects.git
   cd ml-projects
   

2. Create a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate  # On Mac/Linux
   venv\Scripts\activate     # On Windows
   

3. Install dependencies:

   pip install -r requirements.txt
   

---

Projects

1. Stock Price Prediction

Overview

This project predicts future stock prices using Linear Regression with PyTorch. The model is trained on historical stock data and makes future predictions based on learned patterns.

Features

✅ Fetches real-time stock data from Yahoo Finance (yfinance).
✅ Uses Min-Max Scaling for data normalization.
✅ Implements Linear Regression in PyTorch for forecasting.
✅ Saves the trained model for future use.
✅ Separates training (main.py) and prediction (predict.py) for modularity.
✅ Plots actual vs. predicted stock prices for easy comparison.

Files

• preprocessor.py – Handles data fetching & preprocessing.
• model.py – Defines PyTorch Linear Regression model.
• train.py – Trains & saves the stock price prediction model.
• main.py – Runs training & saves model under the stock ticker name.
• predict.py – Loads trained model, predicts future stock prices, and compares them with actual values.

How to Run

1. Train the model on past stock data (e.g., 2024):

   python main.py
   

   This saves the trained model as {TICKER}_model.th (e.g., AAPL_model.th).

2. Predict & Compare Future Prices (e.g., 2025):

   python predict.py
   

   This will load the trained model, predict 2025 prices, and compare them with actual stock prices.

---

License

This repository is licensed under the MIT License.

---