kipash-prog/spotter-ai-assessment ? reverse-engineered prompt

Build me a freight rate prediction project that trains on historical load data and predicts posted rates for new shipments. I want it to learn from things like date, origin and destination location, equipment type, lane history, and a couple of market signals, then output a CSV of predictions for a validation set and another CSV for a December chart input file.

Please make the training script do the full pipeline end to end, including feature engineering, model training, and saving final predictions. Use a strong ensemble approach if that helps, and keep the split time based so later dates are only used for validation. I also want a scoring script that can compare my predictions against the provided inputs and generate a chart image showing the December predicted rates.

If you need to check current library details or best practices, look up the latest docs online.