Performance accuracy between classifiers in sustain of disease conversion for clinical trial tuberculosis data: Data mining approach.

Ponnuraja, C.; Lakshmanan, B.C.; Valarmathi, S.

IOSR Journal of Dental and Medical Sciences; 2016; 15; 105-111.

Abstract: Data mining has been used intensively and extensively by many organizations. In healthcare, data mining is becoming increasingly popular, if not increasingly essential. Data mining applications can greatly benefit all parties involved in the healthcare industry. For example, data mining can help physicians to identify risk factors, effective treatments and best practices in health care industry which helps patients to receive better and affordable healthcare services. The huge amounts of data generated by healthcare transactions are too complex and voluminous to be processed and analyzed by traditional methods. Data mining provides the methodology and technology to transform these mounds of data into useful information for decision making. This work will explore data mining applications within healthcare in one of the major area such as the evaluation of treatment effectiveness as well as disease conversion in tuberculosis patients; it aims to compare various classification techniques and its performance accuracy to build a predictive modelling with controlled clinical trial tuberculosis (TB) data. The classification of tuberculosis patients is of substantial importance in TB disease conversion. During the last few years, many algorithms have been proposed for this task. In this paper, we review different supervised machine learning techniques and its performance accuracy for classification with an original dataset and carry out a methodological comparison using “WEKA” software (Wang, 2010). We used the C4.5 (J48) tree classifier, Iterative Dichotomiser- 3 (ID3), a Multilayer Perceptron (MLP) and a naive Bayes 7classifier over a large set of TB data. It is found that Multilayer Perceptron achieves a competitive performance than naive Bayes, and when the number of features to be classified is reduced naive bayes performs well.

Back to List of publications / Home