商品之目標客群分析研究

Abstract

近年來社群媒體已經成為熱門的資訊分享及資訊傳播的媒介，許多廠商也意 識到其重要性，紛紛透過社群媒體舉辦行銷活動。尤其在發表新商品時，許多公 司行銷人員會舉辦新商品上市的試用活動，希望藉由這些試用者分享的力量，進 一步影響更多消費者，帶動商品的銷售，並提高大眾對品牌的忠誠度。然而，多 數的行銷人員面臨到一個重要的問題，在有限的預算下，到底該選擇哪些人發送 試用品，才能為公司帶來較大的利潤呢? 目前有許多找尋具影響力使用者的研究，是建立在貪婪演算法的基礎發展， 雖然其結果相當接近最佳解，卻有很嚴重的缺點，它的高時間複雜度導致難以應 用在大型的社群網路上。此外，我們也認為挑選出來的目標客群除了應該在社群 中具有高影響力外，若他們對於商品的適配度較高、對商品越感興趣，他們會更 願意分享正面的使用評價，並推薦其他使用者購買該商品。 本研究提出新商品之目標客群辨識分析方法，根據目前欲推廣的商品類別， 找出適合該產品、同時在社群網站上活躍的使用者。透過研究使用者彼此間的關 注關係，深入分析使用者的影響網絡之強度，來挑選最終的目標客群。本研究實 驗結果顯示，我們提出的方法確實能改善影響的成效。

Recently, online social networking websites have become popular platforms of information sharing and disseminating, and numerous firms proposed marketing campaigns especially while they have new products to release. However, restricting to the marketing budgets, an important issue, “who are the right people to target?”, lifts up. Most marketers are facing the problem of target audience selection; it is hard for marketers to decide whom to select as initial users, providing them product samples to use, and expecting to profit from word-of-mouth effects. To resolve such problems, many researchers use greedy algorithm as approximated solution, and provide further improvements based on it. Nevertheless, lack of efficiency makes it hard to apply them to large-scale networks. Additionally, the target audience should not only have great influence in the online community, but also interest in the target product and are willing to recommend others to buy. Therefore, we take these important factors into account to design our methods. In this work, we propose novel target audience identifying methods based on the category of the target product, to solve the target marketing problem. We conduct our experiments using real-world data to test our algorithms. The experimental results showed that our proposed methods outperformed other baseline methods in the performance of MAP, influence spread, and time efficiency.