Online shopping consumers' analysis with data mining techniques

Keywords: E-commerce, Association rules, Apriori , Carma, Frequent Pattern Growth With the development of information technologies, the sources where data collected have diversified and it has gained importance because of the multiplicity of data obtained. This study aims to analyze the top-selling products in a company operating in e-commerce sector, besides identifying the products sold together, improving operational performance in the warehouse, reducing the costs and expediting the release processes. The reasons of preferring algorithms used in the study were; Apriori algorithm was the most common of the association rules, the Frequent Pattern Growth algorithm showed high performance and the Carma algorithm provided further improvement. These algorithms were used to identify the best-selling products in 10000 orders of using the company's 6-month data. The results of the analysis were compared with the sales data of the 9th month, consisting of 1200 order. The X40-X39 coded products, which are the first product group determined according to the Apriori algorithm which gives the best result, cover 5% of all orders. For the first 7 product groups covering 5% to 1% of the orders, the cost effect in bundle of products was analyzed by time study. The analysis results confirm the positive effect of the proposed system on process improvement. The determined product groups included in future orders was analyzed by artificial neural networks. During 6 month from 9. months, it has been proved that the product groups would be included in the order.