Filter
Remove algorithms incompatible with the task and required output.
Describe your task, data and priorities through a guided eight-step process. The selector will rank suitable starting choices, alternatives, warnings and evaluation methods.
Important: the selector identifies algorithms worth testing. It cannot guarantee the best model before appropriate baselines, experiments and validation are completed.
Every recommendation should be understandable and testable.
Remove algorithms incompatible with the task and required output.
Compare labels, data type, dataset size and learning approach.
Apply priorities such as explainability, speed, compute and edge use.
Present a starting choice, strong alternatives and advanced options.
Final selection requires appropriate baselines, validation and domain review.
Start with the simplest credible method and record its result.
Prevent leakage and preserve time order or groups where required.
Compare several candidates under the same data and evaluation process.
Choose measures that reflect the real cost of mistakes.
Inspect errors, subgroups, uncertainty, fairness and operational risks.
Confirm usefulness with domain experts and representative real-world data.
The ranking is educational guidance based on the information supplied. Data quality, preprocessing, tuning, domain constraints, deployment conditions and evaluation design can change the final choice. Your selections remain in this browser and are not uploaded.
Use the guided selector now or continue into the structured algorithm library.