對零售商而言，每天都有大量的交易資料產生，如何善加利用手上資料來更了解消費者並進行行銷活動是件重要的事情，其中包括進行消費者區分、了解消費者行為和進行商品推薦，在過往常會利用RFM(recency, frequency and monetary) 或是PPS(Purchased products structure) 分析來進行消費者區分，然而本研究則透過TF-IDF及LDA分析方法來尋找消費者意圖，以此進行消費者區分，並運用區分好的消費者意圖，以MST(minimum spanning tree)進行購物籃分析來了解消費者行為和找出關鍵產品，最後運用Lift值計算出關鍵產品排序，藉此找出不同消費者意圖下的消費行為並進行商品推薦。;For retailers, a large amount of transaction data is generated every day, how to make good use of the data at hand to better understand consumers and carry out marketing activities is important, including consumer differentiation, understanding consumer behavior and conducting product recommendation. For customer segmentation in the past, Recency, Frequency and Monetary (RFM) or Purchased products structure (PPS) analysis was usually used. However, this study uses Term Frequency–Inverse Document Frequency (TF-IDF) and Latent Dirichlet Allocation (LDA) analysis methods to find consumer intentions to use on customer segmentation. Base on the segment of consumer intentions to conduct shopping basket analysis with Minimum Spanning Tree (MST) to understand consumer behavior and find the key products. Finally, use Lift score to calculate the key product ranking, so as to find out the consumption behaviors under different consumer intentions and make product recommendations.
