Scientific Research Journal (SCIRJ), Volume VII, Issue VII, July 2019 45 ISSN 2201-2796 www.scirj.org © 2019, Scientific Research Journal http://dx.doi.org/10.31364/SCIRJ/v7.i7.2019.P0719670 Analysis of Customers Purchase Patterns of E-Commerce Transactions Using Apriori Algorithm and Sales Forecasting Analysis With Weighted Moving Average (WMA) Method Nanang Riyadi Faculty of Information Technology Budi Luhur University Jakarta, Indonesia nanangriyadi2@gmail.com Muhammad Fariz Mulki Faculty of Information Technology Budi Luhur University Jakarta, Indonesia muhammadfarizmulki@gmail.com Richard Susanto Faculty of Information Technology Budi Luhur University Jakarta, Indonesia richardsusantoubl@gmail.com DOI: 10.31364/SCIRJ/v7.i7.2019.P0719670 http://dx.doi.org/10.31364/SCIRJ/v7.i7.2019.P0719670 Abstract: The challenge of today's e-commerce companies is how to extract large data into information for decision making, especially in terms of promoting products to be relevant, effective and efficient. At this time the XYZ company uses product category data as the main parameter in promoting its products to customers, but the method used is not optimal and efficient because promotions are not displayed to potential customers based on customer purchasing patterns. so that the sales target is not achieved, therefore market basket analysis is needed to find and understand the basic patterns of association rules that occur in customer purchase transactions. In this study the algorithm used is the Apriori algorithm, apriori algorithm is chosen because the resulting association rules have higher accuracy than the FP-Growth algorithm. Then the results of the apriori algorithm association rules are used as a reference in determining the items to be promoted, then sales forecasting is carried out with the Weighted Moving Average (WMA) method to predict the estimated total sales. The results of this study are apriori algorithm that has a higher accuracy value of 130.75 accompanied by sales forecasting analysis with a weighted moving average method that can be implemented in association rules generated from the Apriori algorithm so that it can help companies make decisions in the category of products that are sold a lot. Index Terms—data mining, market basket analysis, Apriori, FP-growth sales Forecasting, weighted moving average I. INTRODUCTION At present, the XYZ company uses product category data as the main parameter in promoting its products to customers, but the method used is not running optimally and efficiently because promotions are not displayed to potential customers based on customer purchasing patterns. so that the sales target is not achieved, therefore market basket analysis is needed to find and understand the basic patterns of association rules that occur in customer transactions. Therefore, in this study, the algorithm that will be used is the Apriori algorithm and the frequent pattern growth (FP-Growth) algorithm. Both of these algorithms will be used as a reference for formulating association rules produced by the market basket analysis model using Rapidminer software version 9.0. Association rules that will be generated by each of the Apriori algorithms and FP-Growth will be evaluated and analyzed to find algorithms that have higher accuracy. Algorithms with higher association rules will be used as a reference in determining the items to be promoted, then sales forecasting is done by the weighted moving average method to predict the estimated total sales.