Obesity risk prediction using machine learning models

1. Clinical question and intended use

The research question is whether recorded lifestyle and physical characteristics distinguish seven obesity levels. Since height and weight are inputs, this is a classification of current recorded status.

2. Cohort, measurements and endpoint

The UCI dataset contains 2,111 records associated with Colombia, Peru and Mexico. The repository states that 77% were generated synthetically and 23% were collected through a web platform. Height and weight are inputs, so this is classification of recorded status rather than prediction of a future health outcome.

The downloadable project, saved report and supplied source CSV define the exact version used here. Repository: original dataset/source record.

3. Model

The final model has 20 encoded input features and 7 outputs. The following dimensions describe the final saved network.

Output semantics: one score per class in this order: Normal_weight, Obese_I, Obese_II, Obese_III, Overweight_I, Overweight_II, Underweight. The predicted label is the largest score. Calibration has not been evaluated, so scores are not presented as calibrated probabilities.

4. Training strategy

The saved training configuration uses CrossEntropy with QuasiNewton. Training minimizes the recorded objective; the validation subset monitors generalization during fitting. The testing subset is used for the evaluation below.

Quasi-Newton method results

5. Model selection and baseline

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 17.8%. This is a baseline for interpretation, not an alternative model fitted on the test labels.

6. Clinical validation

The final classifier is evaluated on 422 testing records. The confusion counts below were reproduced from the saved model. Rows are actual classes and columns are predicted classes. The multiclass decision is argmax; no binary threshold is applied.

Test class prevalence is shown by the support counts. Accuracy is 87.4% and macro F1 is 0.856. A multiclass ROC AUC was not reported; the confusion matrix and per-class measures are the available evidence.

7. Workflow and reproducibility

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.

8. Safety, generalizability and governance

Synthetic records and related anthropometric predictors limit interpretation of the internal split. Compare with a transparent BMI-based classification, separate original from synthetic subjects, and obtain external validation. Scores are not calibrated probabilities and must not guide individual diagnosis or treatment; clinician or specialist review is required for medical use.

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

• Dataset source and provenance.
• Neural Designer project and saved task evidence, 21 September 2026.
• Neural Designer testing-analysis documentation.