Chen, Jian-An
碩士論文(2021)
以影響力為基之市場區隔趨勢預測方法與技術研發
Development of Methodology and Technology for Influence-based Market Segment Trend Prediction
關鍵字 Keywords
摘要
鑒於 (1)即時掌握市場為企業經營之要務、(2)市場區隔分析與趨勢預測為「掌握市場」之首要工作、(3)社群媒體內容潛藏實體世界活動的樣貌，為市場分析”可用”之資料來源(Data Source)、(4)市場可視為一個「系統」，其內部有多樣元素 (消費者、商品、供應商、市場區隔等)且相互影響，導致市場區隔變化，本研究研發以社群分析為基、影響力導向之「市場區隔趨勢預測方法」以及其實現技術，並驗證技術之正確性與方法之有效性。
首先，從系統的角度分析市場內元素及其影響力，分析市場之元素影響力概念模型。再依此模型，設計針對社群媒體內容之社群媒體為基之市場元素影響力模型。接著根據影響力模型，設計一市場區隔趨勢預測方法，包含兩步驟，分別為(1)以影響力為基之商品評價預測，預測出各商品之未來評價，(2)市場區隔趨勢分析，將前步驟所預測之商品評價用於市場區隔趨勢分析，最終得到未來之市場區隔趨勢。
本研究經由實驗，取得較適當之商品評價預測參數，再依此參數進行商品評價預測，由實驗結果證明，用影響力所訓練之商品評價預測模型之平均誤差值為0.00562，相較於僅利用商品評價訓練之商品評價預測模型之預測誤差值0.00972，誤差約降低42.15%，故證明影響力為基商品評價預測可行且有效。再者，利用訓練好之商品評價預測模型，預測各商品未來之評價，再依評價進行市場區隔分析，得到預測之市場區隔。經由與實際區隔比較，其最差準確度為0.70769，其餘準確度皆高於7成以上，有效反映出實際區隔。上述結果證明，以市場內部元素之影響力預測市場商品評價趨勢，再以商品評價趨勢最終分析出市場區隔變化，為有效之方法。因此，驗證本研究所提「以影響力為基之市場區隔趨勢預測方法」可行且有效。
Abstract
Due to the facts that: (1) knowledge of the market is the key to business operation in the digital era; (2) market segment analysis and trend prediction are the primary tasks of "market mastery"; (3) social media content conceals the physical world activities and is a "usable" data source for market analysis; and (4) the market can be considered as a "system" with various elements that interact with each other, this research developed a social media analysis-based, influence-oriented "market segmentation trend prediction method" and enabling technologies, and verified the correctness of the technologies and the validity of the method.
First, the elements of the market and their influence from a systemic perspective are analyzed, then the conceptual model of elemental influence in the market is analyzed. Based on this model, a social media-based influence model for social media content is designed. Then, based on the influence model, a market segmentation trend prediction method is designed, which consists of two steps: (1) influence-based product evaluation prediction to predict the future evaluation of each product, and (2) market segmentation trend analysis, in which the product evaluation predicted in the previous step is applied to the market segmentation trend analysis to finally obtain the future market segmentation trend.
The experimental results proved that the accuracy of the influence-based segment trend prediction method for this study are higher than 70%, and the worst is 0.70769, which effectively reflects the actual interval. Therefore, market segmentation trend prediction based on influence is feasible and effective