Published run
Logistic regression with C 1.0 and the lbfgs solver, seed 42, test size 0.2, stratified. Train 5,634 rows, hold-out 1,409. Before the split, 26.5 percent of 7,043 customers left (1,869 Yes). One-hot encoding. customerID stays out of the features. Churn stays out because it is the answer. Hold-out accuracy 0.8062. For the customers who left, precision 0.6573, recall 0.5642, F1 0.6072, on 374 rows. ROC-AUC 0.8416. Average precision 0.6308. Of the 374 who left, 211 were predicted Yes and 163 were predicted No. scikit-learn 1.4.2, computed on 2026-11-04. The canvas run standardises the train rows inside logistic regression because no scaler node is on the graph. The Telco churn blueprint opens this graph and plays the same run. The voice says the hold-out has 374 customers who left out of 1,409, 26.54 percent. Accuracy 80.62 percent stays high because most customers stayed. For the customers who left, precision is 65.73 percent and recall is 56.42 percent: 211 were caught and 163 were missed. ROC-AUC is 0.8416. Average precision is 0.6308.
Segmentation run
A second fit asks which of these 7,043 customers look alike. It uses every row. There is no hold-out. Drop Data removes customerID, gender, TotalCharges, and Churn. Normalization standardises tenure and MonthlyCharges only. One-hot encoding leaves the service and account columns at 0 and 1. K-Means uses k=2, the silhouette peak for k from 2 through 8, seed 42. That peak is the candidate. It is not proof that two is the true number of segments. Silhouette 0.2784. Calinski-Harabasz 2372.0745. Davies-Bouldin 1.2649. scikit-learn 1.4.2, computed on 2026-11-05.
Churn rate inside each group, counted after the fit. Churn was not a feature.| Group | Customers | Churn rate |
|---|
| 0 | 1,526 | 0.0740 (7.40 percent) |
| 1 | 5,517 | 0.3183 (31.83 percent) |
k from 2 through 8 on this preparation and seed 42. Higher is preferred for silhouette and Calinski-Harabasz. Lower is preferred for Davies-Bouldin. The best value of each is k=2. Silhouette and Davies-Bouldin are not monotone after that.| k | Silhouette | Calinski-Harabasz | Davies-Bouldin |
|---|
| 2 candidate | 0.2784 | 2372.0745 | 1.2649 |
| 3 | 0.2364 | 2134.3050 | 1.7405 |
| 4 | 0.2055 | 1756.7069 | 1.8857 |
| 5 | 0.1897 | 1488.0142 | 2.1130 |
| 6 | 0.1523 | 1310.7782 | 2.0961 |
| 7 | 0.1292 | 1168.8318 | 2.2092 |
| 8 | 0.1338 | 1079.1258 | 2.2419 |
The Telco segmentation blueprint opens this graph and plays the same run. The voice says silhouette 0.2784, Calinski-Harabasz 2372.0745, and Davies-Bouldin 1.2649.