應用資料探勘技術於新產品推薦之研究

CCUR > College of Business > Department of Information Management & Graduate Institute of Information Management > Thesis > Item 987654321/26566

Please use this identifier to cite or link to this item: https://irlib.pccu.edu.tw/handle/987654321/26566

Title:	應用資料探勘技術於新產品推薦之研究 Using Data Mining Technique for New Product Recommendations
Authors:	王子平
Contributors:	資訊管理學系
Keywords:	期望值最大化演算法分類法決策樹貝氏類神經 expectation maximization decision tree naive bayes neural networks
Date:	2014
Issue Date:	2014-01-22 10:22:30 (UTC+8)
Abstract:	現今資訊技術和網路的快速發展，企業面臨的競爭和壓力有別於以往，整個大環境的改變和市場的競爭下，企業必須對於顧客要有足夠的掌握和瞭解才是致勝的關鍵。當企業要推薦或是銷售產品時，就必須考量到更多的因素來分析消費者的購買型態和意願，才能與顧客有長期合作的關係，根據市場統計數據指出，維持一個舊顧客所花費的成本遠低於開發一個新的顧客所需花費的成本。尤其在新產品的推陳出新下，企業要如何讓顧客可以最快地瞭解新產品，進而購買，這將是一門重要的課題。以往在推薦產品的技術上來說，大致上可分為協同過濾(collaborative filtering)和基於內容的過濾(content-based recommendations) 來分析顧客的資料並做推薦，但對於新產品推薦的方面兩種方法皆會遇到一些問題，且大多是以過去的經驗和選擇已經行之有效的啟發式的方法，較缺乏有系統的步驟去尋求結果，因此本研究利用資料探勘的技術來對顧客的購買資料做預測模型的建立，並配合期望值最大化演算法(E-M)分類法來先行對顧客做分群，以比較個人和分群顧客的成效差異，也嘗試了決策樹、貝氏和類神經演算法來建立預測模型，以比較其不同演算法的準確率高低，找出最佳的預測方法。 Due to the rapid development of Information Technology and the Internet, the competition and the pressure that corporations are facing totally different from the past. Changes and competition among the whole market under the general environment make corporations focus on understanding the customer thoroughly, and that is the key to succeed. When corporations aim to sell products, ones are supposed to consider customers, shopping hobbits and shopping willingness in order to keep a long-term relationship with customers. According to statistics from the market, the cost of keeping a frequenter is less than developing a new one. Especially, there are more and more products coming out, so it is an important issue for corporations to make customers easily understand products and then buy. Basically, the technique of product recommending can be divided into Collaborative Filtering and Content-based Recommendations to conduct data analysis on customers and recommend products. However, if corporations want to sell new products by these two techniques, there will be some problems such as looking for results by depending past experience and adopting other used elicitation methods without taking systematic steps. Therefore, the research adopts data-mining technique to build predictive models from the data of customers’ shopping hobbies. The Expectation-Maximization Algorithm is adopted in the research as well to compare the accuracy of prediction between individual with groups of customers. Other approaches such as Decision Tree, Naive Bayes, and neural network are also adopted to build predictive models in order to compare with other algorithms and look for the best way to predict customers hobbits.
Appears in Collections:	[Department of Information Management & Graduate Institute of Information Management] Thesis

Files in This Item:

There are no files associated with this item.

View Licence

Loading...