Probability Correction After Resampling: Training vs Real Prevalence & a Calibration Table
This lesson is part of the course
In this lesson
You correct the fault probabilities of a model trained on a 50/50 resampled set back to the real prevalence of 1 % and compare raw and corrected probabilities with the observed fault rates.
Buy the course once for $9.99 to unlock every lesson and lab, the final exam and the certificate – lifetime access, no subscription.
Lessons in this course
- The Accuracy Trap: 3 Faults in 100 Runs, Confusion Counts & the «Always OK» ModelFree
- Metrics That Respect the Minority: Balanced Accuracy, the Matthews Coefficient, F1 & Fβ Locked
- Random Undersampling & Oversampling: Given Draw Tables, New Class Shares & What Gets Lost Locked
- Synthetic Minority Samples: Neighbour Distances, the k Nearest & New Points by Interpolation Locked
- Class Weights: n ÷ (k × n_class), a Weighted Log-Loss & a New Best Cut-Off Locked
- Cost-Based Thresholds: A Cost Matrix, Expected Cost per Run & the Cheapest Cut-Off Locked
- Probability Correction After Resampling: Training vs Real Prevalence & a Calibration Table Locked
- Capstone – Comparing Strategies: Balanced Accuracy, MCC, Recall, Expected Cost & a Ranked Summary Locked