Not Professional Advice
MLInsights is an educational and exploratory tool. Nothing it produces is financial, medical, legal, or any other kind of professional advice. For decisions in those areas, consult a qualified professional.
Results Are Estimates
Model metrics are measured on a held-out portion of your data and checked with cross-validation where possible. They estimate how a model might perform, but they aren't guarantees. Real-world performance can be very different, especially when:
- The dataset is small, unbalanced, or doesn't represent real conditions.
- A column leaks information about the target, making scores look too good.
- The data changes over time, so patterns from the past stop holding.
Automated Recommendations
Suggested target columns, task types, and recommended models are based on simple rules about column names, value counts, and dataset size. They're a helpful starting point, not an assessment of which model is best for your problem.
Data Quality and Cleaning
Results depend entirely on the data you provide and the cleaning choices you make. Filling or dropping missing values changes your data, and those changes affect every result that follows. Review cleaned data before relying on it.
Exported Model Files
Exported models are provided as-is for further analysis. Test them thoroughly on new data before using them in any real system.
Model files use Python's pickle format, which can run code when loaded. Only load .pkl files you created yourself or that come from a source you trust.
Availability
MLInsights may be unavailable at times, and uploaded data expires automatically. Keep your own copies of anything important, and export results you want to keep.
Limitation of Liability
You use MLInsights and its results at your own risk. See the Terms of Service for the full limitation of liability.
Contact
Questions about this disclaimer? Open an issue on the MLInsights GitHub repository and we'll respond as soon as we can.