A Stacked Machine Learning-Based Investment Sector Recommendation System with Explainable Insights

This project proposes an explainable AI framework that leverages Machine Learning (ML), Deep Learning (DL), and Explainable AI (XAI) to recommend optimal investment sectors based on investor profiles. The system provides personalized investment recommendations and builds user trust using transparent explanations powered by LIME.

📌 Dataset: Investment Survey Dataset – Kaggle

• ✅ Predicts suitable investment sectors like Stocks, Real Estate, Mutual Funds, etc., based on financial goals and risk profile.
• 🤖 Combines ML models: Random Forest, SVM, KNN, Logistic Regression, Naive Bayes, Decision Tree, and Gradient Boosting.
• 🧠 Stacked and Voting Ensemble Classifiers for increased performance (up to 94.12% accuracy).
• 💬 Transparent decision-making using LIME and Feature Importance.
• 🧪 Custom dataset gathered through structured surveys (2,000+ responses).
• 🌐 Optional Flask-based frontend for user interaction.

Investment_Sector_Recommendation/
│
├── data/ # Dataset & preprocessing scripts
├── models/ # Training scripts & saved model files
├── explainer/ # LIME-based interpretation scripts
├── app/ # Flask frontend (optional)
├── requirements.txt # Python dependencies
├── README.md # Project documentation

git clone https://github.com/TMRatul49/Investment_Sector_Recommendation.git
cd Investment_Sector_Recommendation

- python3 -m venv venv
- source venv/bin/activate        # For Windows: venv\Scripts\activate
- pip install -r requirements.txt

- cd app
- flask run

Then open http://localhost:5000 to interact with the recommendation system via a web interface.

• Source: Custom-structured survey
• Hosted on: Kaggle Dataset
• Size: 2,000+ entries

• Age, Gender, Annual Income, Working Status
• Monthly Investment Amount, Duration
• Investment Goal, Motivation, Resource Used
• Mode of Investment (Categorical + Numeric)

• Logistic Regression
• Naive Bayes
• Support Vector Machine
• Random Forest
• K-Nearest Neighbors
• Decision Tree
• Gradient Boosting
• Deep Learning Neural Network
• Ensemble Methods: Voting & Stacking Classifiers

We use LIME and Feature Importance Analysis to:

• 🧩 Provide local instance-level explanations
• 📊 Visualize top influential features per prediction
• 🔍 Ensure transparency and trustworthiness

• 📊 Accurate predictions with stacked ensemble learning
• 🔍 Interpretable insights using LIME
• 🤝 User-friendly system for both novice and expert investors
• 🧪 Ethically sourced dataset reflecting real investor behavior
• 💼 Decision support for financial institutions, policy makers, and individuals

• 🌐 Integrate SHAP for global explainability
• 📡 Add real-time financial data streams
• 📱 Deploy as full-stack or mobile app
• ⏱️ Optimize runtime for low-resource devices

This project is licensed under the MIT License.
See the LICENSE file for details.