AI-Powered ETA Prediction System
We developed an AI-powered predictive model to estimate the ETA of railcars from origin to destination. By using historical and real-time Car Location Messages (CLMs), weather conditions and other relevant data, our solution provides highly accurate ETAs, significantly improving upon the previously scheduled-based estimates.
Robust Data Ingestion and Preprocessing
We set up a robust data ingestion pipeline to collect, clean, and preprocess Railinc data. This pipeline handles both user-fed and historical data, addressing data quality issues such as missing values and inconsistencies. This ensures that the data used for predictions is reliable and comprehensive.
Advanced Feature Engineering and Model Development
We performed feature engineering to extract new features from raw data, enhancing model performance. We selected and implemented appropriate machine learning algorithms, trained models using historical data on AWS sagemaker, and validated them through cross-validation techniques.
User-Friendly Web Application
We developed an interface which allows users to input data and view ETA predictions triggered by lambda functions. Users can upload CSV files in a predefined format, which the system processes to derive new ETAs. The application displays both the new ETA and the ETA given in the events data, making it easy for users to compare and act.
Tech Stack
