社群媒體內容分析與顧客參與度預測

Social Media Content Analysis and Customer Engagement Prediction

本研究提出了一套文字內容分析與預測方法(TCAPM)的創新框架，目的在於分析社群媒體文字內容並預測顧客參與程度。TCAPM整合了TF-IDF重要字詞探勘和Word2Vec語意分析向量資訊，透過系統化的詞群分析流程，有效識別影響顧客參與度的正負面詞群特徵。該方法結合了迴歸分析和機器學習技術，不僅能解釋影響因素，還可透過參數微調提升預測能力。TCAPM的創新之處在於其獨特的文本特徵處理方式：成功捕捉了高維度的語意資訊，同時透過將相似詞彙歸類為主題分群，既保留了詞彙的豐富語意，又提升了文本特徵輸入的意義。這種方法不僅增強了模型的解釋力，也優化了計算效率。TCAPM的自動化分析功能使其無需人工編碼即可應用於大規模資料集。同時，TCAPM也適用於小規模樣本，使研究者能夠深入探討特定類別的媒體內容表現。TCAPM的優點在於其模型兼具預測與解釋能力，為文本分析與文字探勘技術提供了一種新穎的分析方法。

This study introduces the Text Content Analysis and Predictive Modeling (TCAPM) framework, an innovative approach for analyzing social media text content and predicting customer engagement levels. TCAPM integrates TF-IDF keyword mining and Word2Vec word embedding techniques, employing a systematic word cluster analysis process to identify positive and negative word clusters that influence customer engagement. The method combines regression analysis and machine learning, explaining influencing factors while enhancing predictive capabilities through parameter finetuning. TCAPM’s innovation lies in its unique feature processing approach, which captures high-dimensional semantic information by categorizing similar words into thematic clusters. This process preserves rich word semantics while enhancing the significance of textual feature inputs, strengthening the model’s explanatory power and optimizing its computational efficiency. TCAPM’s automated analysis function enables its application to large sample datasets without manual coding, while also being suitable for in-depth analysis of small samples in specific subcategories. The framework’s strengths lie in its practicality and ability to balance interpretability with predictive power. TCAPM not only applies to media content marketing but also provides a new analytical framework for text content analysis.