Assessment of Association Rule Mining Using Interest Measures on the Gene Data

Abstract

Aim: Data mining is the discovery process of beneficial information, not revealed from large-scale data beforehand. One of the fields in which data mining is widely used is health. With data mining, the diagnosis and treatment of the disease and the risk factors affecting the disease can be determined quickly. Association rules are one of the data mining techniques. The aim of this study is to determine patient profiles by obtaining strong association rules with the apriori algorithm, which is one of the association rule algorithms.
 Material and Method: The data set used in the study consists of 205 acute myocardial infarction (AMI) patients. The patients have also carried the genotype of the FNDC5 (rs3480, rs726344, rs16835198) polymorphisms. Support and confidence measures are used to evaluate the rules obtained in the Apriori algorithm. The rules obtained by these measures are correct but not strong. Therefore, interest measures are used, besides two basic measures, with the aim of obtaining stronger rules. In this study For reaching stronger rules, interest measures lift, conviction, certainty factor, cosine, phi and mutual information are applied.
 Results: In this study, 108 rules were obtained. The proposed interest measures were implemented to reach stronger rules and as a result 29 of the rules were qualified as strong.
 Conclusion: As a result, stronger rules have been obtained with the use of interest measures in the clinical decision making process. Thanks to the strong rules obtained, it will facilitate the patient profile determination and clinical decision-making process of AMI patients.