Shopping and Basket Analysis by Using an Improved Apriori Algorithm in WEKA

Abstract

In the past years, e-commerce and online shopping grew fast. It became more helpful by letting people buy the desired product online. Also, to help their users to find the product of their desire easily and make the process simpler, the online shopping websites use some kinds of an algorithm to provide recommendation systems. Often, these systems use techniques like basket analyzing and association rules which is finding the relation between the products together or between users too, so apriori algorithm is one of the famous ones among the recommendation systems. Although it has some limitations while implementing which makes the algorithm less confident or even useless, Let us assume we have 100K records in the sold item list in a system in which about 10K refers to the customers buying only one or two items in their purchase. Therefore, this ten per cent will not affect finding the relation between the items, at the same time these records will make the system less efficient and take more time to analyze, in this paper, we try to show how we can improve the apriori algorithm efficiency and accuracy by some preprocessing on the dataset before applying apriori algorithm by eliminating the unnecessary records, this process helps to make the algorithm better because of reducing the number of transactions, hence finding strong relationships between items easier for the rest of the records.