Comparative Analysis of Machine Learning Classifiers on Bioinformatics and Clinical Datasets

摘要

Looking at the advancements in high-throughput techniques, it is easy to generate huge volume of data and with that comes a problem of processing, analyzing and verifying a data. This paper has analyzed and compared different machine learning based classification techniques for analysis of gene expressions generated through microarrays and clinical decision support using different types of health care datasets. Various classification algorithms, like Decision Tree (DT), Naïve Bayes (NB), and Neural Network (NN) were matched to find the optimum accuracy for diagnosis. The efficiencies of these classification algorithms were examined and equated using glaucoma and cancer datasets. The outcomes were compared based on various parameters, viz. accuracy, sensitivity, dimensionality, and specificity for determining classification algorithm performance. The classifiers indicated enhancements in terms of accuracy after the datasets underwent normalization. This relative analysis depicted that, essentially not a single technique of classification that performs the better than the other. Thus the potential of classification algorithms is highly dependent on several factors of the dataset to be analyzed. In the context of health care datasets, it's the features of the dataset that are the primary factors affecting the performance.