Task8: FeatureEngineering_ModelTuning

This project is part of Data Analysis with Python (Task 8). It focuses on: Feature Engineering & Hyperparameter Tuning (generic example with student dataset). Fraud Detection using Decision Trees (synthetic dataset).

---

🚀 Project Structure

Section 1: Feature Engineering & Model Tuning

• Created new features (Total_Score) in a student dataset.
• Tuned a Random Forest model using GridSearchCV.

Section 2: Fraud Detection with Decision Trees

• Generated a synthetic fraud dataset (fraud_detection.csv).
• Encoded categorical features (credit/debit).
• Engineered new features (Amount_Squared, Log_Amount).
• Trained & tuned a Decision Tree using GridSearchCV.
• Evaluated with Precision, Recall, F1-score.

---

📊 Results

Section 1 (Random Forest – Student Dataset)

• Best Params: {'max_depth': 3, 'n_estimators': 50}
• Accuracy: ~60%

Section 2 (Decision Tree – Fraud Detection)

• Best Params: {'criterion': 'gini', 'max_depth': 10, 'min_samples_split': 5}
• Accuracy: ~90%

⚠️ Accuracy is high because most transactions are legitimate, but the model struggles with detecting fraud due to class imbalance.

---

💡 Recommendations to Improve Fraud Detection Accuracy

Handle Class Imbalance

• Use SMOTE (Synthetic Minority Oversampling) or undersample legitimate transactions.
• Try class weights (class_weight="balanced") in Decision Trees.

Try Advanced Models

• Random Forest, XGBoost, or LightGBM often perform better than a single Decision Tree.
• Ensemble methods can reduce overfitting and capture complex fraud patterns.

Feature Engineering

• Create time-based features (e.g., transactions per hour/day).
• Calculate average transaction amount per user.
• Flag unusual transactions (very high or very frequent).

Anomaly Detection

• Use algorithms like Isolation Forest or One-Class SVM to detect rare frauds.

🛠 Requirements

Install dependencies with:

• pip install pandas numpy scikit-learn

▶️ How to Run

• python Task8_FeatureEngineering_ModelTuning.py

This will:

• Run Section 1 (Random Forest on student dataset).
• Run Section 2 (Decision Tree on fraud detection dataset).

📂 Output

fraud_detection.csv → synthetic fraud dataset generated automatically.

Console output shows:

• Best hyperparameters
• Accuracy (Section 1)
• Classification Report (Section 2)