Indian Journal of Science and Technology

Abstract

Data mining is used extensively and is applied successfully in various fields like market-basket analysis, e-business, fraud detection, quality control, cross-selling of products etc. More recently, data mining has been successfully applied to healthcare sector and healthcare applications. Objectives: The objective of this research is to study the classification based prediction techniques as applied to healthcare. It also aims at finding the different applications and tools used in classification based prediction in the healthcare sector. Methods: Prevalently the prediction techniques used are Decision Trees, Naive Bayes classifier, Bayesian networks, k-Nearest neighbour and artificial neural networks. A few researchers also have used support vector machines, genetic algorithm and decision rules for prediction. Feature selection techniques have been applied to extract relevant features required for the purpose of prediction. Findings: It is found that there is no single algorithm or technique that is the best of all the other algorithms/technique on any given medical dataset and application. Always there is a need to explore the right technique for the given dataset. A detailed review of the research on classification based prediction techniques reveal that the algorithms and techniques are applied on different data sets, which also has heterogeneous data types. It is observed that work is done on improving the predictive accuracy by applying attribute selection measures and feature selection techniques. Techniques have been developed to diagnose diseases, predict the occurrence of diseases, assess the gravity of the diseases such as cancer, heart, skin, liver, SARS, diabetes to name a few. The various applications explored are SMARTDIAB, H-Cloud, Medical Decision Support System, Evidence based medicine, adverse drug events, Passive In-home Health and Wellness monitoring, Healthcare management are a few applications developed in support of Medical data mining. Application: SMARTDIAB is an automated system for monitoring and management of type 1 Diabetic patients which supports monitoring, management and treatment of patients with type 1 diabetes. Passive In-home health and wellness monitoring is an application for monitoring older adults passively in their own living settings through placing sensors in their living environment.

Keywords: Bayesian, Challenges, Classification, Data Mining, Decision Tree, Healthcare, Survey