Evaluating infant development through a novel hybrid intuitionistic fuzzy classification system

Abstract

Preterm birth is a global public health concern, necessitating efficient and accurate neurodevelopmental assessments. This study firstly attempts that develops a novel hybrid classification system combining clustering algorithms, intuitionistic fuzzy sets (IFSs), and classification methods to improve neurodevelopmental assessments for preterm infants by analyzing growth trends between 6 months and 1 year of age. The proposed system integrates clustering techniques, including k-means (KM), fuzzy C-means (FCM), and kernel FCM (KFCM), with IFSs to address the vagueness in growth data. These clusters are then classified using advanced algorithms, such as long short-term memory networks (LSTM), support vector machines (SVM), K-nearest neighbours (KNN), naive Bayes (NB), and backpropagation neural networks (BPNN), to predict cognitive, language, and motor outcomes at 1 year old. Experimental results demonstrated that the KFCM+IFS+LSTM approach achieved the highest accuracy (average cognitive score: 0.95) and robustness (low variability) compared to other methods. The LSTM models consistently outperformed KNN and BPNN, which showed weaker and less consistent results. This hybrid system offers a scalable and precise solution for assessing neurodevelopment in very low birth weight preterm infants, reducing reliance on resource-intensive specialist evaluations and advancing early childhood healthcare methodologies.