Developing the RFM Recommender System of Association Rule Based on Big Data Process

• Main Authors: PAN,CHUNG-WEI, 潘宗暐
• Other Authors: CHOU,TUNG-HSIANG
• Format: Others
• Language: en_US
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/8kys52

碩士 === 國立高雄科技大學 === 資訊管理系 === 107 === Numerous firms accumulate large quantities of data or transactions after importing information systems and services, which leads to troubles with data procedure. Firms also have demands to find customers’ information from large datasets and to understand how to develop marketing strategies accurately to adjust their operational methods. Therefore, this study proposed customer ranking combined Big Data process based on the RFM model (Recency, Frequency, Monetary) to develop a recommendation algorithm using an association rule, which finds greater recommendation to promote operational effects of firms. We developed a hybrid recommender system based on the RFM model and by using an association rule and additional element, “Revenue”. We adjust the weight of potential information to perform the customer ranking, which is conducted by using agglomerative hierarchical clustering. Finally, we present the recommendation by the association rule for each customer level. The data in this study were obtained from a Taiwan apparel agent; therefore, further evaluation in using experiments in other domains is necessary. In addition, this dataset collected transactions including 500 customer purchase records and performed agglomerative hierarchical clustering to rank customers’ levels. Because the time complexity of the clustering algorithm was high, it is not suitable for use with transactions including numerous customers. The datasets in this study use actual sales data; therefore, they are authentic and have been practically applied. The metrics of evaluation showed that the recommender system his study proposes is highly accurate. This method reduced the operation time required when processing large quantities of data by using customer ranking, which also can serve as a decision-maker for firms to implement effective marketing strategies. Managers can use the method proposed in this study to calculate customers’ values based on their potential information and implement adaptive recommendation strategies for each type of customer.