Define the prediction
Choose a target and a useful measure of performance, including costly errors.
Develop predictive models from your business data.
ML. Prepare training data, run experiments, and inspect the errors that matter to the business. Use Models to compare candidates before deciding what to change.
Product walkthrough. Press Space or Enter to pause or play.
ML. From a prediction target to a tested candidate.
Prepare and train
Select training data, keep evaluation cases separate, and run an experiment.
Inspect the tradeoff
Compare overall scores with important subsets, such as hazardous-shipment recall.
Plan the next experiment
Use the findings to revise the data or objective before advancing a candidate.
Predict what matters. Test what could go wrong.
Meridian compares models for late shipments. The strongest overall score still misses a critical hazardous-load guardrail.
Predict which shipments may miss the 21:00 cutoff using information known at the decision time. The intended use is human-reviewed recovery.
Inspect three candidates against the same held-out data. Trial 3 leads overall, but its hazardous-load recall is 84%, below the 90% minimum.
Continue study retains the reason to improve hazardous-load recall. The comparison does not select a new model release.
In practice
Missing a hazardous shipment can matter more than a small gain in average prediction quality. Meridian’s evaluation makes that tradeoff visible: Trial 3 improves the overall score but misses the recall requirement for lithium loads. The team keeps the decision and the reason to continue the study.
Explore the solutionDiscuss your machine learning workflow.
Bring a prediction problem, available data, and examples of the errors that matter. We’ll discuss a first experiment.