Improvement Of Performance Of Association Rule Mining Using Apriori Algorithm and Logistic Regression

Abstract

In this research we like to apply popular data mining techniques e.g. Apriori algorithm on Health and Demographic Surveillance System , Matlab data of International Center for Diarrhoeal Disease Research, Bangladesh (ICDDR,B) to extract association rule for learning the knowledge considering existing population according to their socioeconomic, educational, migration, birth, death, marriage and divorce condition. Apriori algorithm has some limitations which are wasting of time for rule mining by scanning the enter database and generated some unexpected association rules. Now we present an Apriori with Logistic Regression model which is save that wasted time depending on scanning only selected transactions and stopped to generated unexpected association rules. The thesis shows by investigational observations with different values of minimum support that applied on only the Apriori model and newly implemented Apriori with Logistic Regression model that implemented new model (Apriori with Logistic Regression model) saves the time consumed by 82.09% in contrast with the only Apriori model as well as reduces the unexpected rule mining 37.93 %, and makes the association rule mining more efficient and minimum time spend. The models will develop in this research could be helpful during sample selection in any Demographic Surveillance System (DSS) in the ICDDRB and over the world when a research project runs for a long time duration.