AI-Based Predictive System for Public Transportation Delay Analysis

Developed an end-to-end AI-powered predictive system for analyzing and forecasting public transportation delays using a real-world, highly noisy dataset. The project focused on handling imperfect data scenarios, including missing values, inconsistent formats, outliers, and corrupted GPS records.

Implemented a full data science pipeline starting with advanced data cleaning and preprocessing techniques such as data imputation, normalization, and outlier detection. Performed in-depth exploratory data analysis (EDA) to uncover patterns and correlations affecting delay behavior.

Engineered meaningful features including delay duration, time-based categories, weather impact levels, and route-based metrics to enhance model performance. Built and evaluated multiple machine learning models such as Linear Regression and ensemble models to accurately predict delays.

Integrated model evaluation metrics (MAE, RMSE, R²) and applied explainability techniques like feature importance analysis to interpret model decisions and identify key factors influencing delays.

This project demonstrates strong capabilities in real-world data handling, machine learning modeling, feature engineering, and delivering actionable insights from complex datasets.