Skip to content

README.md

Latest commit

 

History

History
242 lines (181 loc) · 6.6 KB

README.md

File metadata and controls

242 lines (181 loc) · 6.6 KB

💼 Salary Prediction — Adult Dataset

Binary classification project predicting whether an individual's annual income exceeds $50K based on demographic and employment attributes from the UCI Adult Census dataset.

Python scikit-learn License Status

📑 Table of Contents

📌 Overview

This project applies supervised machine learning to predict whether a person earns more than $50K per year using the well-known UCI Adult Census Income dataset. The pipeline covers data cleaning, exploratory data analysis (EDA), feature engineering, model training, hyperparameter tuning, and evaluation — all packaged in a reproducible workflow suitable for portfolio and internship presentation.

📊 Dataset

Key Features

  • age, workclass, education, education-num
  • marital-status, occupation, relationship
  • race, sex, capital-gain, capital-loss
  • hours-per-week, native-country

🔁 Project Workflow

  1. Data Loading & Inspection
  2. Data Cleaning — handling missing values & duplicates
  3. Exploratory Data Analysis (EDA)
  4. Feature Engineering & Encoding
  5. Train / Test Split & Scaling
  6. Model Training (multiple algorithms)
  7. Hyperparameter Tuning
  8. Model Evaluation & Comparison
  9. Model Persistence (.pkl)
  10. Conclusion & Insights

🛠 Tech Stack

Category Tools
Language Python 3.9+
Data Handling pandas, numpy
Visualization matplotlib, seaborn
Machine Learning scikit-learn
Model Persistence joblib
Environment Jupyter Notebook

📁 Project Structure

salary-prediction/
│
├── data/
│   └── adult.csv
│
├── notebooks/
│   └── salary_prediction.ipynb
│
├── models/
│   └── best_model.pkl
│
├── images/
│   └── eda_plots/
│       ├── correlation_heatmap.png
│       ├── feature_importance.png
│       ├── confusion_matrix.png
│       └── roc_curve.png
│
├── requirements.txt
├── README.md
├── LICENSE
└── .gitignore

⚙️ Installation

1. Clone the repository

git clone https://github.com/ParvathyM155/Salary_Prediction_Using_Machine_Learning.git
cd Salary_Prediction_Using_Machine_Learning

2. Create and activate a virtual environment

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

3. Install dependencies

pip install -r requirements.txt

▶️ Usage

Launch the notebook:

jupyter notebook notebooks/salary_prediction.ipynb

Or load the saved model directly in Python:

import joblib
model = joblib.load("models/best_model.pkl")
prediction = model.predict(new_data)

🔍 Exploratory Data Analysis

Key insights uncovered during EDA:

  • Strong correlation between education level and income.
  • Hours-per-week and age significantly affect earning probability.
  • Marital status and occupation are powerful categorical predictors.
  • The dataset is imbalanced (~76% <=50K, ~24% >50K).

🔥 Correlation Heatmap

Visualizes pairwise relationships between numerical features and the target variable.

Correlation Heatmap

🤖 Modeling

The following classifiers were trained and compared:

  • Logistic Regression
  • Decision Tree
  • Random Forest
  • Gradient Boosting
  • K-Nearest Neighbors
  • Support Vector Machine

Hyperparameter tuning was performed using GridSearchCV with cross-validation.

🌟 Feature Importance

Top features driving the Gradient Boosting model's predictions.

Feature Importance

🏆 Results

Model Accuracy Precision Recall F1-Score ROC-AUC
Logistic Regression 0.85 0.74 0.60 0.66 0.90
Random Forest 0.86 0.76 0.63 0.69 0.91
Gradient Boosting 0.87 0.79 0.65 0.71 0.92

✅ Gradient Boosting achieved the best overall performance and was selected as the final model.

📈 Model Evaluation

Evaluation techniques applied:

  • Confusion Matrix
  • Classification Report
  • ROC Curve & AUC
  • Precision–Recall Curve
  • K-Fold Cross-Validation
  • Feature Importance Analysis

🧩 Confusion Matrix

Breakdown of correct vs incorrect predictions for each income class.

Confusion Matrix

📉 ROC Curve

Trade-off between true-positive and false-positive rates across thresholds.

ROC Curve

✅ Conclusion

The final Gradient Boosting model reliably predicts income brackets with ~87% accuracy and a 0.92 ROC-AUC, demonstrating strong generalization. Education, age, hours-per-week, and capital gain emerged as the most influential features — aligning with real-world economic intuition.

🚀 Future Improvements

  • Address class imbalance using SMOTE or class weighting
  • Experiment with XGBoost and LightGBM
  • Deploy the model as a Streamlit or Flask web app
  • Add MLflow for experiment tracking
  • Build a CI pipeline with GitHub Actions

👤 Author

Parvathy M

📄 License

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

⭐ If you found this project helpful, please consider giving it a star!