ADVANCED KERNEL FEATURE RANKING WITH OUTLIER HANDLING AND OPTIMIZED DECISION TREE MODEL FOR CARDIOTOCOGRAPHY ANALYSIS

Abstract

Intelligent systems play a crucial part in forecasting health-related conditions in dynamic scenarios. Traditional algorithms for analyzing cardiotocography (CTG) data often rely on static metrics, limited datasets, and a narrow feature range, primarily due to constraints in processing capacity. Additionally, these conventional methods struggle to isolate critical attributes within CTG signals, particularly in the CTU-UHB dataset.This study introduces a dual-phase filtering strategy along with advanced attribute prioritization to enhance the predictive framework for identifying irregularities in cardiotocography patterns. The proposed filtering mechanism identifies anomalies within the data, streamlining the subsequent attribute ranking stage.Moreover, a composite classification approach is designed to boost the accuracy of abnormality detection and improve runtime efficiency on the CTU-UHB dataset. The experimental outcomes demonstrate that the newly developed feature-based classification framework surpasses traditional methods in terms of outlier detection, feature ranking, and predictive performance.