Exploring patterns in employee attrition
Workforce analytics can describe where retention questions merit further investigation. This example models recorded attrition from the supplied employee attributes and evaluates the resulting scores on held-out records.
1. Business decision
Workforce planning benefits from understanding patterns in historical departures. A predictive association does not explain why an employee leaves or identify which intervention would change that outcome.
Historical patterns
Study the recorded attrition=Yes outcome.
Planning context
Interpret results alongside organizational and employment context.
Intervention evidence
Test proposed retention measures separately from the classifier.
2. Data set
The supplied benchmark contains 1,470 employee records. The endpoint is attrition=Yes. The project provides a retrospective modelling exercise; it does not demonstrate that a particular workplace policy caused departures or will prevent them.
The saved Neural Designer project and report define the exact version used here. The project does not identify an independently verified external data record.
| Subset | Records |
|---|---|
| Training | 882 |
| Validation / selection | 294 |
| Testing | 294 |
| Unused | 0 |
Inspect the full variable schema
| Variable | Role | Type | Encoding | Unit |
|---|---|---|---|---|
| age | Input | Numeric | As supplied | |
| business_travel | Input | Numeric | As supplied | |
| daily_rate | Input | Numeric | As supplied | |
| department | Input | Categorical | Human Resources; Research & Development; Sales | As supplied |
| distance_from_home | Input | Numeric | As supplied | |
| education | Input | Numeric | As supplied | |
| education_field | Input | Categorical | Human Resources; Life Sciences; Marketing; Medical; Other; Technical Degree | As supplied |
| employee_count | None | Constant | As supplied | |
| employee_number | Input | Numeric | As supplied | |
| environment_satisfaction | Input | Numeric | As supplied | |
| gender | Input | Binary | Female; Male | As supplied |
| hourly_rate | Input | Numeric | As supplied | |
| job_involvement | Input | Numeric | As supplied | |
| job_level | Input | Numeric | As supplied | |
| job_role | Input | Categorical | 9 categories; see schema | As supplied |
| job_satisfaction | Input | Numeric | As supplied | |
| marital_status | Input | Categorical | Divorced; Married; Single | As supplied |
| monthly_income | Input | Numeric | As supplied | |
| monthly_rate | Input | Numeric | As supplied | |
| num_companies_worked | Input | Numeric | As supplied | |
| over_18 | None | Constant | As supplied | |
| over_time | Input | Binary | No; Yes | As supplied |
| percent_salary_hike | Input | Numeric | As supplied | |
| performance_rating | Input | Numeric | As supplied | |
| relationship_satisfaction | Input | Numeric | As supplied | |
| standard_hours | None | Constant | As supplied | |
| stock_option_level | Input | Numeric | As supplied | |
| total_working_years | Input | Numeric | As supplied | |
| training_times_last_year | Input | Numeric | As supplied | |
| work_life_balance | Input | Numeric | As supplied | |
| years_at_company | Input | Numeric | As supplied | |
| years_in_current_role | Input | Numeric | As supplied | |
| years_since_last_promotion | Input | Numeric | As supplied | |
| years_with_curr_manager | Input | Numeric | As supplied | |
| attrition | Target | Binary | No; Yes | As supplied |
Interactive chart: attrition pie chart. Enable JavaScript to explore it.
Interactive chart: attrition Pearson correlations chart. Enable JavaScript to explore it.
3. Model
The final model has 48 encoded input features and 1 outputs. The following dimensions describe the final saved network.
| Layer | Input shape | Output shape | Activation |
|---|---|---|---|
| Scaling | 48 | 48 | — |
| Dense | 48 | 3 | Tanh |
| Dense | 3 | 1 | Sigmoid |
Output semantics: the sigmoid score increases toward Yes; No 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 | 120 |
| Elapsed time | 00:00:00 |
| Stopping criterion | Maximum validation error increases |
| Training error | 0.353 |
| Validation error | 0.467 |
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 85.7%. This is a baseline for interpretation, not an alternative model fitted on the test labels.
6. Testing analysis
The final classifier is evaluated on 294 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 Yes.
Test class prevalence is shown by the support counts. Accuracy is 79.6% and macro F1 is 0.66. ROC AUC is 0.818. The native ROC optimal threshold is descriptive of this test set and is not an independently validated operating policy.
| Actual / predicted | No | Yes | Total |
|---|---|---|---|
| No | 210 | 42 | 252 |
| Yes | 18 | 24 | 42 |
| Class | Test cases | Sensitivity / recall | Specificity | Precision / PPV | F1 |
|---|---|---|---|---|---|
| No | 252 | 83.3% | 57.1% | 92.1% | 0.875 |
| Yes | 42 | 57.1% | 83.3% | 36.4% | 0.444 |
Interactive chart: ROC 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.
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
Do not use benchmark scores alone for employment decisions. Establish dataset provenance and representativeness, review privacy and sensitive attributes, validate across time and organizational units, and investigate group-level findings with human review. The model is not an estimate of an intervention's effect.
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.
