FusionNet-CXG: Hypothyroidism Prediction Dataset with 3,772 Records

Divya Kesavulu's FusionNet-CXG model dataset contains 3,772 records with 30 features for hypothyroidism prediction. The dataset was used to train a hybrid deep-learning model combining CNN, eXtended LSTM, and GRU components, with performance evaluated via 50 fold-level evaluations. The data was last updated on 2026-04-21 and is shared under a CC-BY-4.0 license.

Use Cases

• Training diagnostic classifiers for hypothyroidism based on 30 clinical features.
• Benchmarking hybrid deep-learning architectures like CNN-LSTM-GRU combinations.
• Evaluating class imbalance mitigation techniques like SMOTE-NC on medical data.
• Conducting feature importance analysis for clinical predictions using SHAP.

Strengths

• Contains 3,772 patient records, providing a substantive sample size.
• Includes 30 features per record for model training.
• Performance was rigorously estimated using 10 repetitions of 5-fold stratified cross-validation.
• Model achieved a reported mean accuracy of 0.9394 and AUC-ROC of 0.94.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect bias inherent to the original public source; external validation is noted as future work.

Provenance

Source

Author Divya Kesavulu, sourced from a public dataset.

Collection Method

Experiments performed on a public dataset; class imbalance was mitigated using SMOTE-NC.

Time Range

null

Freshness

Last updated 2026-04-21 17:48:55; freshness should be verified.

Geography

null