Comparative Analysis of Machine Learning Techniques for Impeccable Detection of Diabetes

Abstract

This paper presents an analysis of machine learning approach in detection and diagnosis in the field of health sciences. Extremely high use of such tools and techniques now a days are found predominantly used in medical and life saving sciences, first of all ensures the very high level of accuracy. Secondly, apart from high level of accurate results, these tools and techniques provide the speediest results which are not possible through any other means. The machine learning techniques fundamentally use artificial intelligence in which a large amount of data set is processed using a well established algorithm. Usually, such algorithm uses two sets of data in which the first set is trained data set and the another data set is physically collected or observed data set. These two data sets are compared either in real time or referral time and accordingly deviations are recorded to detect and diagnose the presence or absence of a particular deficiency and disease. The use of engineering domain based technical algorithms and it's integration with medical sciences has revolutionised the advancements in health and life saving acumen with the speediest possible assistance. In this paper the most common disease now a days which has affected more than 30% population i. e. Diabetes has been chosen for the purpose of research. Various machine learning techniques presently being used in such detection and diagnosis along with the relevant algorithm has been put up. Machine Learning Algorithms popularly known as MLA, analyses the real time data present in the data set and helps predicting the outcome relevant to presence or absence of any disease. Researchers across the globe have used in recent past different MLA techniques at different times to predict the collected or the observed data. In the study presented here the classification techniques such as k-NN, Naive Bayes, Decision Tree and SVM have been analysed which are being used for accurate and fast results leading to detection and diagnosis of diabetes which are well supported by python machine learning modules like pandas, sklearn, Seaborn.