Prediction of Periventricular Leukomalacia Occurrence in Neonates Using a Novel Support Vector Machine Classifier Optimization Method

Abstract

Prior work has documented that Support Vector Machine (SVM) classifiers can be powerful tools in predicting clinical outcomes of complex diseases such as Periventricular Leukomalacia (PVL). Our previous study showed that SVM performance can be improved significantly by optimizing the supervised training set used during the learning stage of the overall SVM algorithm. This study fully develops the initial idea using the reliable Leave-One-Out Cross-validation (LOOCV) technique. The work presented in this paper confirms previous results and improves the performance of the SVM even further. In addition, using the LOOCV technique, the computational time is decreased and the structure of the algorithm simplified, making this framework more feasible. Furthermore, we evaluate the performance of the resulting optimized SVM classifier on an unseen set of data. This demonstrates that the developed SVM algorithm outperforms normal SVM type classifiers without any loss of generalization.

A Novel Embedded Feature Selection and Dimensionality Reduction Method for an SVM Type Classifier to Predict Periventricular Leukomalacia (PVL) in Neonates

“…To optimize the ML model published by Jalali and colleagues [14,15], the team used a filter approach as a benchmark method to select the best possible feature subset. This method aims to establish a relevance metric based on selected correlation (or dependency) criteria between individual features and the output, ranking them from weak to strong [23][24][25].…”

“…As illustrated in the first part of the results section and explained in the discussion section, the previously used benchmark filter method [14,15] suffers from several crucial drawbacks. To address these drawbacks, the invented feature selection algorithm presented in this study builds on the idea of the wrapper method, shown in Figure 2b.…”

“…Figure 1 illustrates the current approach and compares it with the novel predictive ML method used in our previous study [14]. While the predictive model introduced in our previous studies [14][15][16][17][18] provided satisfying classification results, it did not explain them. That is, the algorithm had no transparency, providing no indication of which portion of the data was relevant to the performance of the ML algorithm.…”