Exploring the Value of Online Word-of-Mouth Dynamics in Predicting Movie Success with a Hybrid Machine Learning Approach

다국어 초록 (Multilingual Abstract)

Exploring the Value of Online Word-of-Mouth Dynamics in Predicting Movie Success with a Hybrid Machine Learning Approach by LIU QI Doctor of Philosophy in Business Administration Graduate School of Kyung Hee University Advised by Dr. Sunho Jung In traditional analyses of word-of-mouth (WOM) effects in movie market, the focus has primarily been on the static characteristics of online WOM, represented by four dimensi ons such as volume, valence, variance, and sentiment. Similarly, traditional machine lear ning techniques have predominantly concentrated on evaluating the predictive accuracy of movie revenue. In this context, this study aims to conduct an exploratory investigation i nto the dynamic nature of online WOM patterns and their relationship with movie reven ue from a marketing analytics perspective. The core research problem involves identifyin g key influencing factors for enhancing movie success based on the dynamic patterns of key characteristics of online WOM. The paper is structured as follows. Firstly, functional data analysis (Ramsay & Silverman, 2005), a non-parametric statistical approach, is employed to conduct dynamic modeling of online WOM . Secondly, elastic net regression (Zou & Hastie, 2005) is utilized to implement the statistical significance testing of a variety of factors influencing movie box office performance. Thirdly, relative importance analysis (Tonidandel & LeBreton, 2011) is employed to calculate the relative importance of each significant factor. This study aims to provide insights into the dynamic relationship between online WOM patterns and movie revenue. By employing a hybrid two-step approach, expected to provide insights into how studios can strategically manage the online review process to improve box office performance of their films. Key words: Online word-of-mouth, marketing analytics, hybrid machine learning approach, functional data analysis, elastic net regression, relative importance analys

목차 (Table of Contents)

• Abstract v
• 1. Introduction 1
• 2. Literature Review 2
• 2.1 Online WOM 2
• 2.1.1 Online Communication 2
• 2.1.2 Traditional vs. eWOM 3
• 2.1.3 opinion leadership and eWOM 4
• 2.2 Methodologies for measuring online word-of-mouth 5
• 2.3 Dynamic features of online word-of-mouth 6
• 2.4 Predictive studies in the movie domain 8
• 3. Methods 9
• 3.1 Data 9
• 3.2 Procedures 11
• 3.3 Data analysis 14
• 4. Results 15
• 4.1 Baseline Characteristics 15
• 4.2 Summary Measures Of Predictor Variables 15
• 4.2.1 Cumulative Box Office Revenue (Cum_Revenue) 16
• 4.2.2 Pre-release Strategy (Release_star) 16
• 4.2.3 Star Presence (Star) 16
• 4.2.4 Number of Screens (Num_SCR) and Screenings (Num_Screenings) 16
• 4.2.5 Early Box Office Revenue (Early_Revenue) 16
• 4.2.6 Principal Components of Review Counts, Ratings, and Text Entropy 17
• 4.2.7 Daily Review Counts (Daily_Volume_PC1 and Daily_Volume_PC2) 17
• 4.2.8 Cumulative Ratings (Cum_Rating_PC1 and Cum_Rating_PC2) 17
• 4.2.9 Review Text Entropy (Entropy_PC1 and Entropy_PC2) 17
• 4.2.10 Rating Variability (Std_Rating_PC1 and Std_Rating_PC2)17
• 4.3 FPCA Result 18
• 4.4 Elastic Net Result 28
• 5. Discussion 31
• 5.1 Empirical findings and managerial implications 31
• 5.1.1 Key Findings 31
• 5.1.2 Managerial Implications 32
• 5.2 Limitations and Further Research 33
• 5.2.1 Limitations 33
• 5.2.2 Further Research 33
• References 35
• Table Contents
• Table 1 11
• Table 2 15
• Table 3 28
• Table 4 29
• Figure Contents
• Figure 1 12
• Figure 2 13
• Figure 3 14
• Figure 4a 18
• Figure 4b 19
• Figure 4c 20
• Figure 4d 21
• Figure5a 22
• Figure5b 22
• Figure5c 23
• Figure5d 24
• Figure5e 25
• Figure5f 25
• Figure5g 26
• Figure5h 27