Feature Selection and Classification of Spinal Abnormalities to Detect Low Back Pain Disorder Using Machine Learning Approaches

Abstract

With the advent of computational intelligence, the analysis of medical data using machine learning techniques benefits accurate classification of different health diseases and disorders. However, disorganization and variability of data make the job difficult. This paper has streamlined an ensemble learning approach for classification of low back pain disorder based on spinal abnormality data of 310 patients with 12 features. To overwhelm the misleading effect of inappropriate attributes, most influential features are identified using evolutionary feature elimination method. Experiments are performed in both way- with or without feature filtering. The basic machine learning algorithms used in the work: Logistic regression, Decision Tree, Naive Bayes, and in addition to the Random Forest ensemble learning method. Random Forest classifier, as expected, is recorded to exhibit the best accuracy of 94% over other classifiers.