Data-driven diagnosis of spinal cord abnormalities.

The purpose of this study is to examine potential machine learning techniques that could have been applied to predict spinal abnormalities. Univariate extraction of features as just a filtration system for selecting the features and principal component analysis (PCA) as just a tool for features extracted are one of the data preprocessing stages. A variety of machine learning methodologies, SVM classifier (SVM), regression models (LR), and bagging classification techniques, are now being examined as well for the diagnosis of spinal abnormalities. The Support Vector Machine, Logistic Regression, bagging Svm Classifier, bagging Logistic Multiple regressions are used on a data of 300 sample images inside the Kaggle source. The effectiveness of identification of aberrant and healthy spinal patients is assessed using several parameters, such traas is n and test correct True Negatives, recall, and miss rates. The categorization model was analyzed to use the receiver- operating characteristic (ROC) and highly accurate curves, and the kernel parameters. The obtained training prediction accuracy for Svm Classifier, Logistic Regression, bagged Svm Classifier, and bag Logistic Regression are 86 %, 86 %, and 85 percent, correspondingly, if % of input could be characterized for training. For testing data, the prediction accuracy for Svm Classifier, Logistic Regression, bagging Support Vector Machine, and bagging Regression Analysis all are 87%. [ABSTRACT FROM AUTHOR]