Model performance evaluation is one of the key tasks in the ML model lifecycle. A model's performance indicates how successful the model is at making useful predictions on data.
Once your trained model is loaded into Fiddler, click on Evaluate to see its performance.
To measure model performance for regression tasks, we provide some useful performance metrics and tools.
- Root Mean Square Error (RMSE)
- Measures the variation between the predicted and the actual value.
- RMSE = SQRT[Sum of all observation (predicted value - actual value)^2/number of observations]
- Mean Absolute Error (MAE)
- Measures the average magnitude of the error in a set of predictions, without considering their direction.
- MAE = Sum of all observation[Abs(predicted value - actual value)]/number of observations
- Coefficient of Determination (R2)
- Measures how much better the model's predictions are than just predicting a single value for all examples.
- R2 = variance explained by the model / total variance
- Prediction Scatterplot
- Plots the predicted values against the actual values. The more closely the plot hugs the y=x line, the better the fit of the model.
- Error Distribution
- A histogram showing the distribution of errors (differences between model predictions and actuals). The closer to 0 the errors are, the better the fit of the model.
To measure model performance for classification tasks, we provide some useful performance metrics and tools.
- Measures the proportion of positive predictions which were correctly classified.
- Measures the proportion of positive examples which were correctly classified.
- Measures the proportion of all examples which were correctly classified.
- Measures the harmonic mean of precision and recall.
- Measures the area under the Receiver Operating Characteristic (ROC) curve.
- Log Loss
- Measures the performance of a classification model where the prediction input is a probability value between 0 and 1. The goal of the ML model is to minimize this value.
- Confusion Matrix
- A table that shows how many predicted and actual values exist for different classes. Also referred as an error matrix.
- Receiver Operating Characteristic (ROC) Curve
- A graph showing the performance of a classification model at different classification thresholds. Plots the true positive rate (TPR), also known as recall, against the false positive rate (FPR).
- Precision-Recall Curve
- A graph that plots the precision against the recall for different classification thresholds.
- Calibration Plot
- A graph that tell us how well the model is calibrated. The plot is obtained by dividing the predictions into 10 quantile buckets (0-10th percentile, 10-20th percentile, etc.). The average predicted probability is plotted against the true observed probability for that set of points.
Updated 8 months ago