鄭翔安

Hsiang-AnCheng

碩士論文(2016)

以社群媒體為基之價值觀類型分析方法

Approach of Value Analysis Based on Social Media

關鍵字 Keywords

商務模式、社群媒體、主題標籤、Schwartz價值觀、關聯規則

Business Model、Social Media、Hashtag、Schwartz Value Survey、Association rule

摘要

在現今越來越競爭的商業環境中,商務模式的創新已經是一個重要的議題。不過要設計商務模式的時候,首先必須先了解目標客群的想法及背後的特徵。因此了解目標客群價值觀的資訊是可以提供企業參考的資訊。不過傳統的價值觀評測方法對於企業來說時間和人力的成本過於高昂,因此如何有效地分析大量顧客之價值觀是一個值得研究的議題。

近年來社群媒體(Social Media)越來越發達,已經成為人們用來分享意見和觀點的數位平台,有助於實現自動化分析價值觀之需求。因此本研究以世界上最多人使用的社群媒體Facebook和旗下的社群媒體Instagram資料為基礎,發展可以由使用者之「動態文章」進行價值觀類型的分析方法。研究中使用Schwartz價值觀作為價值觀模型,首先透過關聯規則演算法找出每個價值觀之價值觀標籤,並且藉由以文字探勘技術TF-IDF、根據Schwartz價值觀特性所設計之一致性模型、相對性模型和相鄰性模型來進行加權,接著計算使用者動態文章對應價值觀標籤之特徵值,最後設計兩種方法進行計算:(1)基於特徵值的計算;(2)基於區分特徵值的計算。實驗結果可以發現,第二種的準確率優於第一種方法,在分析利他性-利己性的準確度最佳可以達到83.6%,分析開放性-保守性的準確度最佳可以達到73.7%,分析四個價值觀象限準確度最佳可以達到62.2%,因此可以發現價值觀標籤跟Facebook使用者之動態文章存在一定的關聯性。

Abstract

In today's competitive environment, business model innovation has been an important issue. However, in order to design a business model, the first thing you must understand the feature and the idea behind the customer segments. Therefore, understanding the customer segments values that can provide enterprise reference information. However, the traditional method for value evaluation is extremely costly because of time and labor consuming. Therefore, the manner in which to effectively conduct automated value analysis for a large number of objects is an important issue. In recent years, more and more developed social media, has become used to share opinions and views of digital platforms. Perhaps social media can serve the needs of value analysis. Therefore, this study used the world most widely used social media Facebook and a subsidiary of Facebook Instagram as the basis. Developed the methods for value analysis based on the user’s status. In this research, Schwartz value is used as the value. First, find the value tag for each High order value through association rule algorithms. With text mining technology TF-IDF, according to Schwartz value to design the consistency model, relativity model and contiguity model to calculate the characteristic value of the corresponding user status values of the value tag. Finally, designing two analysis methods: (1) Based on the characteristic value; (2) Based on the distinguishing characteristic value. The results can be found method two is better than method one. In the analysis of Self-Transcendence and Self-Enhancement the accuracy can reach 83.6%, the analysis of Openness to Change and Conservation the accuracy can reach 73.7%, the analysis of Quadrant values the accuracy can reach 62.2 %. It is possible to find the value tag and Facebook user's status there is a certain relevance.