THE RESULT
What the data shows
The largest computed metric is 0.7461 for Includes PageValues / Random forest; the smallest is 0.1495 for Without PageValues / Prior baseline. Metric: average_precision (Average precision).
The extra-information variant is a leakage investigation, not a recommended feature set.
| model | accuracy | balanced accuracy | macro f1 | roc auc | average precision | brier | information | variant |
|---|---|---|---|---|---|---|---|---|
| Prior baseline | 0.8505 | 0.5 | 0.4596 | 0.5 | 0.1495 | 0.1272 | Without PageValues | Without PageValues / Prior baseline |
| Logistic regression | 0.8493 | 0.5038 | 0.4698 | 0.7342 | 0.2881 | 0.1171 | Without PageValues | Without PageValues / Logistic regression |
| Random forest | 0.8505 | 0.5225 | 0.5084 | 0.7748 | 0.3645 | 0.1109 | Without PageValues | Without PageValues / Random forest |
| Prior baseline | 0.8505 | 0.5 | 0.4596 | 0.5 | 0.1495 | 0.1272 | Includes PageValues | Includes PageValues / Prior baseline |
| Logistic regression | 0.8868 | 0.6875 | 0.7258 | 0.8974 | 0.6348 | 0.08288 | Includes PageValues | Includes PageValues / Logistic regression |
| Random forest | 0.9059 | 0.7695 | 0.796 | 0.9287 | 0.7461 | 0.06842 | Includes PageValues | Includes PageValues / Random forest |
THE METHOD
From source to answer
Paired diagnostic models with grouping held fixed to the base features.
The Python source exposes this study’s transformations. The complete project download includes shared preparation and evaluation routines.
EVALUATION
How the result was checked
Fixed 80/20 split of unique input groups (seed 42); exact input duplicates stay together.
Fixed models; default decision threshold; no holdout tuning.
Target and input features
Target: revenue
Inputs: administrative, administrative_duration, informational, informational_duration, productrelated, productrelated_duration, bouncerates, exitrates, specialday, month, operatingsystems, browser, region, traffictype, visitortype, weekend
THE NEXT DECISION
What follows from the finding
Exclude downstream outcome information from any deployable conversion model.
Where the conclusion stops
One observation is a completed session. End-of-session behavior cannot support an early-session prediction claim. PageValues is excluded from purchase models because it can encode downstream purchase information. Associations are not experiment results.
Related studies may reuse observations or holdouts. These are historical analyses; associations and backtests do not demonstrate commercial impact. Further model tuning needs new, untouched evaluation data.