Abstract
The objective of this research is to speed up the classification efficiency of machine learning (ML) approaches through the use of preprocessing techniques such as balancing and selection of features over imbalanced CTG (Cardiotocography) data. After analyzing the class distribution of the CTG data, it is found that the dataset is imbalanced, as most of the samples belong to the same class. So, to deal with the disparity in the data, an over-sampling technique called SMOTE is employed to boost the performance of various classifiers. Also, the suggested work intends to find out the critical attributes affecting fetal health. So, a crowding distance-based multi-objective Ant lion optimization (MOALO-CD) is presented here as a feature selection (FS) procedure and its results are compared with another well-known FS mechanism: the multi-objective genetic algorithm (MOGA). The experimental results revealed that most of the classifiers performed well in the balanced and reduced dataset while compared with the imbalanced and whole CTG dataset. The performance of Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), and K-Nearest Neighbor (KNN) over the balanced dataset has improved after applying SMOTE to the original data. By applying MOALO-CD, we found some of the most influential factors affecting fetal health status, resulting in a reduced CTG dataset with very little compromise on the classification performance of balanced CTG data.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have