網路社群媒體的普及提供了消費者一個抒發消費體驗、交換對產品與服務意見的便利平台，也提供企業了解消費者心態的極佳管道。本研究運用機器學習方法分析社群媒體Instagram使用者之貼文資料，建構一套預測使用者購物傾向的模型; 使用者貼文資料包括貼文圖片、貼文內容和貼文特徵等三種類型，購物傾向類型包括市場行家、享樂購物、比較購物、物質主義、衝動購物等五種。本研究首先以問卷調查方式分析147位Instagram使用者之購物傾向，接著以隨機森林 (Random Forest)、決策樹 (Decision Trees)、支援向量機 (Support Vector Machine) 等六種機器學習演算法分析受測者的Instagram貼文資料，最後由模型來判斷受測者之購物傾向。研究結果顯示，預測模型的分類準確率介於72.3%-89.5%，具有良好之判斷能力。本研究成果有有助於企業規劃社群行銷與個人化之產品推薦。;The popularity of online social media has provided consumers with convenient platforms on which they can share their consumption experiences and exchange opinions on products and services and also provided businesses with excellent channels through which they can understand the mentality of consumers. This study employed machine learning to analyze user posts on the social media, Instagram, to construct a user shopping orientation prediction model. The user post data included post image, post content, and post characteristics. The shopping orientation categories included market maven, hedonic shopping, comparison shopping, materialism, and impulse buying. We first investigated the shopping orientations of 147 Instagram users using a questionnaire and then employed five machine learning algorithms including random forest, decision trees, and support vector machine to analyze the Instagram post data of the participants. Finally, the models were utilized to determine the shopping orientations of the participants. The resulting accuracy rates of category prediction in the models ranged from 72.3% to 89.5%, which was fairly good. The results of this study can help businesses plan social media marketing and personalized product recommendations.
