• Mark Louie San Juan
• Christian James D. Villapando

Investigating fetal health classification with UCI cardiotocography (CTG) data, this study employed a pipeline integrating advanced machine learning models (MLP, LSTM, KAN) and oversampling techniques (ADASYN, SMOTE, SMOTEENN). Results consistently showed that oversampling significantly improved evaluation metrics, with SMOTEENN being particularly effective. The MLP model, when paired with SMOTEENN, achieved the best overall performance, including 97.82% accuracy and high F1-score, ROC-AUC, and MCC, alongside an efficient runtime. While Kolmogorov-Arnold Networks (KAN) also improved with SMOTEENN (94.04% accuracy), they lagged behind MLP in both performance and efficiency. CTG is a critical diagnostic tool in obstetrics for monitoring fetal health, and accurate classification of its signals is essential for timely medical intervention. These findings highlight the critical role of robust oversampling and appropriate model selection for high-performing and interpretable healthcare AI.

Fetal health classification, Cardiotocography (CTG), Machine learning, MLP, LSTM, KAN, Oversampling, ADASYN, SMOTE, SMOTEENN, Class imbalance, Diagnostic tool, Healthcare AI.

"Predicting Fetal Health Using Cardiotocography Data with Kolmogorov-Arnold Networks ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 10, page no.l436-l490, October-2025, Available :http://www.jetir.org/papers/JETIRTHE2224.pdf

"Predicting Fetal Health Using Cardiotocography Data with Kolmogorov-Arnold Networks ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 10, page no. ppl436-l490, October-2025, Available at : http://www.jetir.org/papers/JETIRTHE2224.pdf