Abstract:
Electronic commerce especially online market increased so much about the last quarter of century; which it makes people's life easier by providing to buying the required goods on the web. Also, it let customers to find their demand easier than the classical shopping especially some websites which are implementing some technique to recommending the items for their costumers, so these online systems are needed to utilize some data mining approaches like basket analyzing and association rules which these algorithms are trying to finding correlation among the items together or among customers too, so Apriori approach is a powerful algorithm in data mining for the recommendation systems area. however, it has some problems in the execution time, and these problems reduces the confident of the approach or even useless and less efficient, let’s assume we have 100.000 transactions in the sold item list in a system in which about 10.000 refers to the buyers which buying only an item or just two products in list of their purchase, because of this issue these records has not effect of the finding the correlation among the objects, simultansly these records have bad effect on efficiency of the algorithm and time consuming to analyze, this thesis will try to show how it could improve the efficiency and accuracy of the apriori approach by some preprocessing against to the dataset instead of applying apriori algorithm directly by and this processes are eliminating the unwanted transactions, this task assists to make the algorithm more better because of eliminating the number of records, by this way the process of discovering strong relationships among products more simpler to the rest of the transactions
