應用機器學習於行銷結果之預測

Applying Machine learning for forecast of marketing results

姚緯嬬(Wei-Ju Yao) ； 廖家芯(Jia-Sin Liao) ； 侯東旭(Tung- Hsu Hou)

《TANET2017 臺灣網際網路研討會》 2017年 (2017/10) Pp. 154-158

https://doi.org/10.6728/TANET.201710.0028

摘要

網際網路的發展帶動電子商務活動快速成長，導致金融業者傳統的營運模式都要做重大的改變，其中行銷通路是銷售的重要管道，但影響顧客消費的因素諸多，為了增加獲利與優勢、如何讓企業能及時掌握住客戶需求，不僅能避免資源浪費在成交率低的顧客身上，還能提高銷售量與收益。本研究以行銷結果為主題，透過資料探勘方法做預測，依據UCI公開資料庫的銀行行銷總筆數45211進行研究，利用Clementine 12.0版與MATLAB R2009a軟體進行預測模型之建構，研究結果發現如果客戶資料含有信用狀況、個人貸款、房子貸款等其中之一特徵時，表示此銀行判斷行銷成功的機率將會提高，因此建構的兩種模型皆適合做為預測之參考依據。

關鍵字

銀行行銷 ； 粒子群演算法 ； 倒傳遞類神經 ； 支援向量機 ； 機器學習

並列摘要

The development of Internet has led to the rapid growth of e-commerce activities, resulting in the financial industry's traditional mode of operation have to make major changes, including marketing channels is an important pipeline sales, but the impact of customer consumption factors, in order to increase profits and advantages and how to enable enterprises to grasp the needs of customers in a timely manner, not only to avoid waste of resources in the low turnover of the customer who can increase sales and revenue. This study is based on marketing results, through data exploration methods to do forecast, according to the UCI public database of bank marketing total number 45211 to study, using the Clementine version 12.0 and MATLAB R2009a software to predict the construction of the model, the results found that if the customer information contains credit status, personal loans, house loans, one of the characteristics of that the bank to determine the success rate of marketing will be improved, so the two models are suitable for the construction of the forecast as a basis for reference.

並列關鍵字

Bank marketing ； Particle Swarm Optimization ； Back Propagation Neural Network ； Support Vector Machine ； Machine learning