PDP report
target bike-rentals · 11 features · 4 plotted
1 · Overview — where to look
An additive surrogate read off the global curves reproduces 71.7% of the model's predicted variance; adding the regional plots kept by the decision sequence, 88.6%.
Each point is a feature: importance (x) against heterogeneity (y). Bottom-left is ignorable; bottom-right is important and fully described by its mean effect; the top-right corner — important and heterogeneous — is where the mean hides something. An arrow marks each split the decision sequence accepted: from the feature's global point to its weighted-mean point across the subregions.
| # | feature | importance | heterogeneity | #regions | regional analysis |
|---|---|---|---|---|---|
| 1 | hr | 0.7314 | 0.2882 | 7 | split into 4 regions → |
| 2 | temp | 0.2281 | 0.2668 | 5 | split found — rejected by the decision sequence → |
| 3 | yr | 0.1878 | 0.2028 | 7 | split found — rejected by the decision sequence → |
| 4 | hum | 0.1020 | 0.1525 | 7 | split into 4 regions → |
| 5 | season | 0.0906 | 0.1341 | · | not plotted (below the coverage cut) |
| 6 | weathersit | 0.0557 | 0.1093 | · | not plotted (below the coverage cut) |
| 7 | workingday | 0.0455 | 0.3711 | · | not plotted (below the coverage cut) |
| 8 | weekday | 0.0448 | 0.1351 | · | not plotted (below the coverage cut) |
| 9 | mnth | 0.0368 | 0.1337 | · | not plotted (below the coverage cut) |
| 10 | windspeed | 0.0190 | 0.1580 | · | not plotted (below the coverage cut) |
| 11 | holiday | 0.0189 | 0.0990 | · | not plotted (below the coverage cut) |
The plotted features carry 80% of the total importance mass (target 80%, ceiling top_k = 5).
The decision sequence. Starting from the global curves, each round applies the split with the largest explained-variance gain, measured on top of the splits above it, and stops when no remaining split adds at least 1.0%. A real split (its heterogeneity does drop) can still add nothing — or even hurt, by double-counting — when its variance is already explained by an earlier split.
| step | regions | heterogeneity | explained variance |
|---|---|---|---|
| global effects (GAM) | · | · | 71.7% |
| + split hr (on temp, workingday, yr) | 4 | 0.478 → 0.288 | +15.5% → 87.2% |
| + split hum (on hr, temp, weathersit) | 4 | 0.174 → 0.153 | +1.4% → 88.6% |
| rejected · temp (on hr, hum) | 3 | 0.267 → 0.241 | +0.9% — below the 1.0% threshold |
| rejected · yr (on hr, hum) | 4 | 0.203 → 0.178 | -0.1% — redundant (variance already explained) |
| rejected · workingday (on hr, yr) | 4 | 0.371 → 0.307 | -4.3% — redundant (variance already explained) |
Bar view — importance and heterogeneity
2 · Regional analysis — the final CALM
The selected snapshot: global effects everywhere except the accepted splits. Features in descending importance — a split feature enters as one group at the instance-weighted mean of its subregions. The split features' global counterparts are in the baseline section at the end.
2.1 · hr
Split on temp, workingday, yr into 4 regions — worth +15.5% of explained variance on top of the splits above it; importance and heterogeneity here are the instance-weighted means over the subregions. The global counterpart is in the baseline.
Partition tree
Feature 3 - Full partition tree:
🌳 Full Tree Structure:
───────────────────────
hr 🔹 [id: 0 | heter: 0.48 | inst: 2000 | w: 1.00]
workingday = no 🔹 [id: 1 | heter: 0.37 | inst: 614 | w: 0.31]
temp < 4.50 🔹 [id: 2 | heter: 0.24 | inst: 248 | w: 0.12]
temp ≥ 4.50 🔹 [id: 3 | heter: 0.32 | inst: 366 | w: 0.18]
workingday = yes 🔹 [id: 4 | heter: 0.34 | inst: 1386 | w: 0.69]
yr = 2011 🔹 [id: 5 | heter: 0.25 | inst: 696 | w: 0.35]
yr = 2012 🔹 [id: 6 | heter: 0.32 | inst: 690 | w: 0.34]
--------------------------------------------------
Feature 3 - Statistics per tree level:
🌳 Tree Summary:
─────────────────
Level 0🔹heter: 0.48
Level 1🔹heter: 0.35 | 🔻0.13 (26.67%)
Level 2🔹heter: 0.29 | 🔻0.06 (17.80%)
2.2 · temp
Global effect
Regional effects
A split on hr, hum into 3 regions was found (heterogeneity 0.267 → 0.241), but the decision sequence skips it: it adds only +0.9%, below the 1.0% threshold. The regional plots are omitted; reproduce them with find_regions.
2.3 · yr
Global effect
Regional effects
A split on hr, hum into 4 regions was found (heterogeneity 0.203 → 0.178), but the decision sequence skips it: it adds no explained variance beyond the splits kept there — the same variance is already read elsewhere. The regional plots are omitted; reproduce them with find_regions.
2.4 · hum
Split on hr, temp, weathersit into 4 regions — worth +1.4% of explained variance on top of the splits above it; importance and heterogeneity here are the instance-weighted means over the subregions. The global counterpart is in the baseline.
Partition tree
Feature 9 - Full partition tree:
🌳 Full Tree Structure:
───────────────────────
hum 🔹 [id: 0 | heter: 0.17 | inst: 2000 | w: 1.00]
temp < 13.71 🔹 [id: 1 | heter: 0.15 | inst: 1338 | w: 0.67]
weathersit = light rain/snow 🔹 [id: 2 | heter: 0.33 | inst: 146 | w: 0.07]
weathersit ∈ {clear, mist, heavy rain} 🔹 [id: 3 | heter: 0.12 | inst: 1192 | w: 0.60]
temp ≥ 13.71 🔹 [id: 4 | heter: 0.18 | inst: 662 | w: 0.33]
hr = 17.00 🔹 [id: 5 | heter: 0.19 | inst: 33 | w: 0.02]
hr ∈ {0.00, 1.00, 2.00, …} (23 levels) 🔹 [id: 6 | heter: 0.17 | inst: 629 | w: 0.31]
--------------------------------------------------
Feature 9 - Statistics per tree level:
🌳 Tree Summary:
─────────────────
Level 0🔹heter: 0.17
Level 1🔹heter: 0.16 | 🔻0.01 (7.69%)
Level 2🔹heter: 0.15 | 🔻0.01 (5.23%)
3 · Global baseline — without regions
What you would believe about the split features without the regional analysis: their global mean effects, with the heterogeneity the accepted splits just explained still hiding inside the band. Compare with their subregions in the regional analysis above.
hr
hum
