Examining interest in a vehicle-insurance offer
Cross-sell models can help a marketing team rank customers for further evaluation. This example evaluates the recorded response to a vehicle-insurance offer using demographic, vehicle and policy information.
1. Business decision
An insurance cross-sell exercise asks which customers show interest in an offer. This project encodes not-interested as the positive class, so interested customers lie at the low-score end of its native output.
Score direction
High scores indicate not-interested.
Offer interest
Use the opposite end of the score when studying interested customers.
Commercial outcome
Interest classification alone does not establish conversion value or campaign profit.
2. Data set
The supplied cross-sell dataset contains 381,109 records. Its category order is interested, not-interested, so the native output is a score for not-interested. Interested customers occupy the low-score end. The distinction is retained in the ROC, rate charts, confusion counts and calculator.
The downloadable project, saved report and supplied source CSV define the exact version used here. Repository: original dataset/source record.
| Subset | Records |
|---|---|
| Training | 228667 |
| Validation / selection | 76221 |
| Testing | 76221 |
| Unused | 0 |
| Variable | Role | Type | Encoding | Unit |
|---|---|---|---|---|
| gender | Input | Binary | female; male | As supplied |
| age | Input | Numeric | As supplied | |
| previously_insured | Input | Binary | no; yes | As supplied |
| vehicle_age | Input | Numeric | As supplied | |
| vehicle_damage | Input | Binary | no; yes | As supplied |
| annual_premium | Input | Numeric | As supplied | |
| vintage | Input | Numeric | As supplied | |
| response | Target | Binary | interested; not-interested | As supplied |
Interactive chart: response pie chart. Enable JavaScript to explore it.
Interactive chart: response Pearson correlations chart. Enable JavaScript to explore it.
3. Model
The final model has 7 encoded input features and 1 outputs. The following dimensions describe the final saved network.
| Layer | Input shape | Output shape | Activation |
|---|---|---|---|
| Scaling | 7 | 7 | — |
| Dense | 7 | 3 | Tanh |
| Dense | 3 | 1 | Sigmoid |
Output semantics: the sigmoid score increases toward not-interested; interested is the other class. Calibration has not been evaluated, so scores are not presented as calibrated probabilities.
4. Training strategy
The saved training configuration uses WeightedSquaredError with QuasiNewton. Training minimizes the recorded objective; the validation subset monitors generalization during fitting. The testing subset is used for the evaluation below.
Interactive chart: Quasi-Newton method error history. Enable JavaScript to explore it.
Quasi-Newton method results
| Measure | Value |
|---|---|
| Epochs number | 98 |
| Elapsed time | 00:00:02 |
| Stopping criterion | Minimum loss decrease |
| Training error | 0.37 |
| Validation error | 0.368 |
5. Model selection
No model selection experiment is recorded in this supplied project. The displayed architecture is the trained model used for testing; earlier article claims about a different selected architecture do not apply to this version.
A transparent test-set comparator is the majority-class rule, with accuracy 87.8%. This is a baseline for interpretation, not an alternative model fitted on the test labels.
6. Testing analysis
The final classifier is evaluated on 76,221 testing records. The confusion counts below were reproduced from the saved model. Rows are actual classes and columns are predicted classes. The decision threshold is 0.5 on the score for not-interested.
Test class prevalence is shown by the support counts. Accuracy is 68.8% and macro F1 is 0.604. ROC AUC is 0.843. The native ROC optimal threshold is descriptive of this test set and is not an independently validated operating policy.
| Actual / predicted | interested | not-interested | Total |
|---|---|---|---|
| interested | 8687 | 611 | 9298 |
| not-interested | 23158 | 43765 | 66923 |
| Class | Test cases | Sensitivity / recall | Specificity | Precision / PPV | F1 |
|---|---|---|---|---|---|
| interested | 9298 | 93.4% | 65.4% | 27.3% | 0.422 |
| not-interested | 66923 | 65.4% | 93.4% | 98.6% | 0.786 |
Interactive chart: ROC chart. Enable JavaScript to explore it.
Interactive chart: Positive (not-interested) rates chart. Enable JavaScript to explore it.
Interactive chart: Negative (interested) rates chart. Enable JavaScript to explore it.
Interactive chart: Cumulative gain chart. Enable JavaScript to explore it.
7. Model deployment
Validated inputs → saved preprocessing → neural network → score or estimate → domain review. The ZIP contains the original project, source CSV, schema, test metrics and standalone interactive chart exports. The project hash in the schema identifies this exact version.
Explore the exported model
This research demonstration runs locally in your browser. Values outside the training range are outside the validated domain and are rejected. A valid input range does not guarantee that a combination is physically or operationally plausible.
This is a research demonstration, not a validated decision policy.
8. Evidence and limitations
Campaign profit is not established here. The exported profit task uses the opposite class from the intended conversion and its sample-ratio output is not valid, so that chart is excluded. Validate contact-time availability, calibration and uplift in a later campaign; include actual conversion value and intervention cost only after the target direction is correct.
No external validation or independent calibration study is included. Preprocessing statistics and model choices should be refitted within a prospective or grouped validation design. Correlations and directional responses describe associations, not causes. Human review is required before an operational decision.
Confidence intervals, subgroup performance, calibration curves and decision-cost validation are not established by these tasks. Predictive values apply to the observed test class distribution and may change when prevalence shifts.
References
Earlier video walkthroughs use a previous model version; the saved project and results above are the current reference.
