Stock Price Prediction (Short-Term) — Research-Based Project

Overview

This project focuses on predicting the next day’s closing stock price using historical market data. A research-oriented approach is applied to explore relationships between stock features and evaluate machine learning models for short-term forecasting.

The analysis combines data exploration, regression modeling, and visualization to derive meaningful insights about stock price behavior.

Objectives

• Analyze historical stock market data
• Understand relationships between price features
• Build predictive models for short-term forecasting
• Compare performance of linear and non-linear models
• Evaluate predictions using metrics and visualization

Dataset

• Source: Yahoo Finance (via yfinance API)
• Stock Used: Apple (AAPL) (can be changed)
• Time Period: Last 2 years (daily data)

Features Used:

• Open
• High
• Low
• Volume

Target:

• Close (Next Day Closing Price)

Methodology

1. Data Collection

• Retrieved stock data using yfinance

2. Data Exploration

• Checked structure, statistics, and consistency
• Verified absence of major missing values

3. Feature Selection

• Selected price-based and volume features

4. Model Training

• Linear Regression (baseline model)
• Random Forest Regressor (non-linear model)

5. Evaluation Metrics

• Mean Squared Error (MSE)
• R² Score

6. Visualization

• Compared actual vs predicted prices using line plots

Model Performance

Model	MSE ↓	R² ↑	Performance
Linear Regression	2.45	0.975	Excellent
Random Forest	9.81	0.902	Moderate

Key Findings

• Linear Regression outperformed Random Forest in both accuracy and visualization
• Strong linear relationship exists between:
  • Open, High, Low → Close price
• Random Forest showed larger deviations and lower generalization performance

Important Insight

Although stock markets are often considered complex and non-linear, this study shows that:

Short-term stock price prediction can be effectively modeled using simple linear relationships when features are highly correlated.

This highlights the importance of understanding data before choosing complex models.

Limitations

• External factors (news, economy, sentiment) not included
• No technical indicators used (e.g., moving averages)
• Limited to short-term prediction
• Models not hyperparameter-tuned

Future Work

• Add lag features (previous day prices)
• Apply advanced models (LSTM, XGBoost)
• Integrate sentiment analysis from news/social media
• Deploy a real-time prediction system

Conclusion

This project demonstrates that model simplicity can outperform complexity when data relationships are inherently linear.

Linear Regression proved to be highly effective for short-term stock prediction, while Random Forest did not provide additional benefits without tuning.

The key takeaway:
Choosing the right model depends more on data characteristics than model complexity.

Project Structure

Predict_Future_Stock_Prices/ │── stock_prediction.ipynb │── README.md

Tools & Libraries

• Python
• Pandas
• Matplotlib
• Seaborn
• Scikit-learn
• yfinance

Author

Malik Muhammad Mudassir Iqbal
AI/ML Engineering Intern — DevelopersHub Corporation