InComeIQ
ML Income Predictor
A production-ready Machine Learning pipeline and web application for demographic income classification. It features an object-oriented training pipeline, REST APIs, and dynamic feature explainability.
Problem Statement
Many ML projects stop at basic Jupyter Notebooks. Real-world applications require transitioning raw data science code into modular software, while also defeating the "black box" nature of ML models to provide transparent AI.
Methodology
Built a modular architecture with object-oriented feature engineering using Sklearn Pipelines. Implemented an automated orchestrator comparing models via GridSearchCV. Developed a REST API with Flask and a custom PredictionPipeline with in-memory artifact caching. Engineered dynamic feature explainability using Matplotlib/Seaborn.
Results
XGBoost emerged as the best model with ~84% accuracy. Delivered a system providing confidence percentages and dynamically generated bar charts to explain predictions. The modern "glassmorphism" web app natively logs all requests to a SQLite database.