Adaptive Weighted Diversity Ensemble Learning Approach for Fetal Health Classification on Cardiotocography Data

Abstract

Accurate classification of cardiotocography (CTG) data is crucial for monitoring fetal health during pregnancy. However, existing methodologies face challenges in achieving precise classifications. This research aims to enhance fetal health assessment accuracy by developing a robust model through the integration of advanced ensemble learning techniques with feature selection, scaling, and adaptive weighting mechanisms. Our proposed Adaptive Weighted Diversity Ensemble Model (AWD) incorporates Random Forest (RF), AdaBoost (AB), Gradient Boost (GB), and Support Vector Classifier (SVC) as base classifiers. The framework includes three modules: Exploratory Data Analysis (EDA), data pre-processing, and an adaptive weighted diversity ensemble module. EDA provided insights into class distribution and feature correlations, guiding subsequent analysis. Data pre-processing ensured uniformity in model training. The adaptive weighting mechanism in the ensemble module assigns weights based on accuracy scores, with diversity computation ensuring a diverse ensemble capturing various prediction patterns. The findings of the experiments prove that the proposed AWD model performed better than other methods in terms of accuracy, F-Score, precision, and recall metrics.

Keywords

• Adaptive ensemble learning model
• adaptive weighting
• cardiotocography data analysis
• diversity computation
• fetal health classification
• machine learning