Machine Learning Techniques for the Diagnosis of Disc Disorders: Comparative Analysis

Abstract

In recent years, machine learning (ML) techniques have emerged as leading solution providers in almost all real-world problems. Advancements in development of diagnostic procedures have been obtained while implementing these techniques in medical sciences. In this work, five ML techniques: Multilayer Perceptron (MLP), Gaussian Naïve Bayes (GNB), Logistic Regression (LR), Support Vector Machine (SVM) and Artificial Neural Networks (ANN) have been evaluated on various parameters i.e., accuracy, precision, f1-score, recall, weighted average and macro-average. Experimentation is performed to determine the most precise and accurate forecasting method for the diagnosis of disc disorders. The dataset of 310 records with 12 attributes was considered for the experiment purpose and was divided into two ratios of 80:20 and 90:10 to train and test the different ML models. In addition, different type of model and training based hyperparameter (train-test ratio, depth, width, dropout) optimization was performed to determine the effect on training process of deep learning models and in optimizing the results. Results obtained shows that LR and MLP techniques performed better than other techniques in all instances and best results were obtained for all the methods in the train-test dataset split of 90:10.