運用類神經網路技術於發掘對群眾募資活動有影響力推特使 用者之研究

Abstract

為了累積用戶，並達到募集公司初期資本之目的，許多新創公司會選擇將自己 的產品/ 服務作為專案於群眾募資網站上線。以美國知名募資網站(Kickstarter) 為例，每日上線專案更是超過千餘件，如何制定推廣策略，並從眾多專案中脫穎而 出，是所有新創公司必須考量的首要關鍵。然而，新創公司資源有限，傳統的數位 廣告推廣方式費用高昂，且轉換率偏低，容易受到廣告屏蔽(Ad block)服務的影 響。現今，影響力行銷(Influencer Marketing)儼然成為新創公司推廣募資專案的 優先策略。影響力行銷泛指企業在社群網路中找出關鍵影響者(Key Influencers)， 與其合作行銷活動，推廣產品/服務的過程。本研究將於知名社群服務提供商推特 (Twitter)上蒐集資料，並以所有討論群眾募資活動的推特使用者做為分析標的， 透過類神經網路網路(Artificial Neural Networks)技術分析高達 20 維度的用戶資 訊，同時比較知名推特用戶影響力評比服務(Followerwonk)之用戶影響力等級， 訓練出一系統模型，本研究之貢獻為提供一更即時、經濟的預測模型，找出社群網 路上的關鍵影響者，促使專案成功。

Resource and time management is an important topic for a brand to develop its marketing strategy, so brands should focus on the tasks that are most profitable. Influencer marketing allows brands to leverage the influencer’s ability to engage highly relevant audiences and create authentic contents. Although influencer marketing is profitable and efficient, it comes with a unique set of challenges that can impact the results of a brand’s marketing campaign, how to identify the right influencers is one of them. In this thesis, a neural network model has been constructed to rank potential influencers for crowdfunding campaigns on Twitter. We applied the social authority value, a mechanism developed by Followerwonk, one of the most popular Twitter marketing platforms in the United States to examine the influential strength of a Twitter user. We found our model a cost-efficient and effective model for identifying categorical influencers. Also, 13 out of 20 different factors of Twitter influence had been evaluated as significant for measuring the influential strength, which improved time-efficiency of evaluating a potential influencer.