📊 Income Prediction Using Decision Tree Regression

This project aims to predict individual income using a complete Machine Learning workflow.
The dataset contains demographic, education, employment, and household-related attributes, and the model uses a Decision Tree Regressor with full hyperparameter tuning to estimate income values.

---

🚀 Project Overview

• Perform data cleaning and preprocessing.
• Apply Ordinal and One-Hot Encoding to categorical features.
• Use log transformation to reduce skewness in the target variable.
• Split data into training and testing sets.
• Use GridSearchCV to optimize tree hyperparameters.
• Evaluate model performance (R², RMSE).
• Visualize predicted vs actual income values.

---

🛠️ Technologies Used

• Python
• Pandas
• NumPy
• Scikit-Learn
• Plotly
• Jupyter Notebook
• Google Colab

---

🔧 Model Details

• Algorithm: Decision Tree Regression
• Tuning:
  • max_depth
  • min_samples_leaf
  • min_samples_split
• Scoring: Negative Mean Squared Error (MSE)
• Target: Income (log-transformed during training)

---

📈 Results

Due to the synthetic nature of the dataset, the model shows:

• Training R² : low
• Testing R² : low

This indicates underfitting, meaning the dataset lacks strong relationships between features and income.

Despite this, the project demonstrates a clean, end-to-end ML pipeline suitable for learning and experimentation.

---

📁 Files Included

• data.csv → Dataset used for training and tetsing the model
• Income Prediction Project No LogTransformation.ipynb → Main notebook containing full ML workflow without log transformation applied ( higher in the accuracy )
• Income Prediction with Log Trasnfromation.ipynb → Another version of the notebook containing full ML workflow with log transformation applied ( lower in the accuracy )
• README.md → Project documentation

---

📈 Results

Due to the synthetic nature of the dataset and the selected model ( Decision Tree Regression ) , the model shows underfitting, with low R² scores on the both notebooks .

• Income Prediction Project No LogTransformation.ipyn → Rsquared = 1.68%
• Income Prediction Project with Log Trasnfromation.ipyn → Rsquared = -8.54%

This demonstrates a realistic challenge when datasets lack strong feature–target relationships.

---

🎯 Future Improvements

• Try ensemble models: RandomForest, GradientBoosting, XGBoost
• Use a more realistic dataset
• Apply advanced feature engineering to extract meaningful patterns

---

👤 Author

Developed by Samir Mohamed as part of a regression machine learning practice project.