解構群眾募資-以Light GBM洞見開創契機

Deconstructing crowdfunding to generate opportunities with the Light GBM insights

張絡鈞(Lo-Chun Chang)

指導教授 ： 唐瓔璋

國立交通大學/管理學院/經營管理研究所/碩士(2020年)

摘要

2008年以降，平台經濟的快速發展，也帶動群眾募資經濟的蓬勃，群眾募資市場供給跟需求也不斷變動，而群眾募資產業也不斷在變革，從過去的長尾態勢逐漸變成帕累托法則發展的趨勢，因此群眾募資專案發起者，面對動態的市場，勢必不能以靜態的觀點看待市場，因此本研究運用動態的機器學習演算法與動態的可解釋性分析，以期更了解整體市場需求與變化。
本研究以市場為核心去解構整個群眾募資產業，過去社會科學分析所應用的傳統統計方法，受限資料的局限性和預測能力的不足與定態的研究方法，本文以Kickstarter多年的全部群眾募資專案與資料作為研究主軸，採用多種機器學習演算法，並運用多種可解釋性分析如SHAP Value、Permutation Value、Partial Dependency等方法，結合社會科學觀點與資訊科學之長，洞見三大群眾募資管理戰略構面與產生本文問題意識（problematic） 與研究命題，試圖解構整個群眾募資產業。
研究發現，首先就進入市場決策而言，可運用Lean Start up與群眾外包降低產品在市場的風險；對市場認知而言，群眾募資發起者，需要從宏觀至微小處，在動態的市場中，運用滾動式調整市場戰略目標與對與之相對應的戰術，乃至將群眾募資昇華至品牌操作，建立自身群眾募資專案社群；最後就財務面決策來說，須建立可達成的財務準則與目標，同時運用多層級的定價方式達成財務目標，
關鍵字：群眾募資、群眾外包、巨量資料、機器學習、可解釋性分析、Light GBM、帕累托法則、精益創業。

關鍵字

群眾募資 ； 群眾外包 ； 巨量資料 ； 機器學習 ； 可解釋性分析 ； Light GBM ； 帕累托法則 ； 精益創業

並列摘要

After 2008, the rapid development of the platform economy has also boosted the crowdfunding economy. The supply and demand of the crowdfunding market has also changed constantly, and the crowdfunding industry has been constantly changing. From the long tail of the past to the trend of Pareto's law. Therefore, the crowdfunding project sponsors, in the face of dynamic market, must not be able to view the market from a static point of view, so this study proposes a dynamic algorithm, and dynamic interpretability analysis, in order to better understand the overall market demand and changes.
This research takes the market itself as the core to deconstruct the entire crowdfunding industry. Traditional statistical methods used in past social science analysis are often limited by inadequate data, inadequate model prediction capabilities, and steady-state research methods. This research uses the Kickstarter platform data for many years.
The crowdfunding project and data as the main idea of the research, using a variety of machine learning algorithms, combined with the social science perspective and the strengths of information science, using a variety of interpretability analysis methods such as SHAP Value, Permutation Value, Partial Dependency, etc. To gain insights and generate problematic-the three aspects of the crowdfunding management strategy, trying to deconstruct the entire crowdfunding industry, research findings.
First, enter the market. In terms of decision-making, Lean Startup and crowdsourcing can be used to reduce the risk of products in the market. For market awareness, crowdfunding initiators need to use macro-to-micro-levels to adjust the strategic goals of the market in a dynamic market and corresponding tactics, and even sublime crowdfunding to brand operations, establish their own crowdfunding project community. Finally, in terms of financial decisions, it is necessary to establish achievable financial standards and goals, and use multi-level pricing to achieve financial goals.

Keywords: Crowdfunding, Crowdsourcing, Big data, Machine Learning, Interpretability Analysis, Light GBM, Pareto rule, Lean Start Up

並列關鍵字

Crowdfunding ； Crowdsourcing ； Big data ； Machine Learning ； Interpretability Analysis ； Light GBM ； Pareto rule ； Lean Start Up

參考文獻

參考文獻

1. Agrawal, A. K., Catalini, C., & Goldfarb, A. (2011). The geography of crowdfunding (No. w16820). National bureau of economic research.

2. Agrawal, A., Catalini, C., & Goldfarb, A. (2015). Crowdfunding: Geography, social networks, and the timing of investment decisions. Journal of Economics & Management Strategy, 24(2), 253-274.

3. Altmann, A., Toloşi, L., Sander, O., & Lengauer, T. (2010). Permutation importance: a corrected feature importance measure. Bioinformatics, 26(10), 1340-1347.

4. Arlot, S., & Celisse, A. (2010). A survey of cross-validation procedures for model selection. Statistics surveys, 4, 40-79.