MINING TRANSACTIONAL DATA TO PRODUCE EXTENDED ASSOCIATION RULES USING COLLABORATIVE APRIORI, FSA-RED AND M5P PREDICTIVE ALGORITHM AS A BASIS OF BUSINESS ACTIONS
Laksana, Eka Angga
Liong, Thee Houw
MetadataShow full item record
There are large amounts of transactional data which showed consumer shopping cart at a store that sells more than 150 types of products. In this case, the company is utilizing these data in making business action. In previous studies, the data that has a lot of attributes and record data reduction algorithms handled by the FSA Red (Feature Selection for Association Rules) are then mined using Apriori algorithm. The resulting association rules have high levels of accuracy and excellent test results, which rely more than 90%. In this study, the association rules generated in previous research will be updated by using prediction algorithms M5P, so that the reliability of association rules can be updated for the next day forward. Furthermore, some data mining technique such as: clustering and time series pattern will be implemented to examine the truth and to extend the validity of association rules which were built. It can be concluded that the association rules were established after will generate strong association rules with confidence equal or higher than 70% and the truth of the rules can be seen from the time series pattern on each group of goods which are then used as the basis of business actions.