行動商務產品推薦方法

Abstract

隨著第三代行動通訊(3G)的用戶數增加，使得行動資料傳輸量增大，因此促使行動商務的形成。但多通路的公司想要發展行動商務往往遇到困難，因為缺乏對新行動通路使用者消費行為的了解。而傳統協同過濾推薦法因很少產品在行動網站上被瀏覽到所以可能產生資料稀疏的問題。

在這篇研究中，我們首先提出一個以手機特徵為基礎的混合推薦法去解決傳統協同過濾法在行動環境下資料稀疏的問題，我們運用手機的特徵去辨認使用者的偏好，然後依其特性將使用者分群。此混合推薦法結合了手機特徵和產品偏好，並且運用了用戶群產生關聯規則來推薦產品。

接著，我們提出一個混合多通路方法去解決對於新通路使用者消費行為的未知問題，與傳統協同過濾法中因找不到相似使用者所產生的資料稀疏問題。推薦給新行動通路的產品是基於新行動通路的瀏覽行為與既有通路的消費行為以不同的權重混合而成。

最後，我們結合了手機特徵與多通路法成為一個手機特徵多通路混合法，利用關聯規則與最頻繁項目集來推薦產品。我們的實驗顯示手機特徵多通路混合法的推薦品質比手機特徵法和混合多通路法好，亦比傳統協同過濾法好。

Mobile data communications have evolved as the number of third generation (3G) subscribers has increased to conduct mobile commerce. Multichannel companies would like to develop mobile commerce but meet difficulties because of lack of knowledge about users’ consumption behaviors on the new mobile channel. Typical collaborative filtering (CF) recommendations may suffer from the so-called sparsity problem because few products are browsed on the mobile Web.

In this study, we first propose a mobile phone feature-based (MPF) hybrid method to resolve the sparsity issue of the typical CF method in mobile environments. We use the features of mobile phones to identify users’ characteristics and then cluster users into groups with similar interests. The hybrid method combines the MPF-based method and a preference-based method that employs association rule mining to extract recommendation rules from user groups and make recommendations.

Second, we propose a hybrid multiple channels (HMC) method to resolve the lack of knowledge about users’ consumption behaviors on the new channel and the difficulty of finding similar users due to the sparsity problem of typical CF. Products are recommended to the new mobile channel users based on their browsing behaviors on the new mobile channel as well as the consumption behaviors on the existing multiple channels according to different weights.

Finally, we combine MPF with HMC approach into a hybrid MPF-HMC method, which utilizes association rules of product categories and products as well as most frequent items to recommend products. Our experiment results show that the hybrid MPF-HMC combined method performs well compared to the pure MPF-based and HMC-based methods as well as the typical kNN-based CF method.