Title Page
Contents
Abstract 9
I. Introduction 12
II. Theoretical Backgrounds 15
2.1. Concept and Features of Social Robots 15
2.2. Elderly Care Using Social Robots 17
2.3. Time Series Data and Social Robot 19
2.4. The Social Robot Used in Study, "Hyodol" 21
III. Essay 1: Data-driven Persona Creating Based on Social Robot Usage Patterns 24
3.1. Pattern Derivation: Time Series Clustering 24
3.2. Pattern Optimization: Time Series Analysis 27
3.3. Data-driven Persona Creating 28
3.4. Research Model 30
3.4.1. Research Process 30
3.4.2. Data Collection and Preprocessing 31
3.5. Essay 1: Research Results 33
3.5.1. Pattern Derivation Based on Time Series Clustering 33
3.5.2. Pattern Optimization based on Time Series Analysis 36
3.5.3. Data-driven Persona Creating 41
3.6. Essay 1: Conclusion and Implication 53
IV. Essay 2: Human-robot Interaction's Triggers, Antecedent and Consequences Path 56
4.1. Research Questions 57
4.1.1. Human-robot Interaction Types: Touch and Contents 57
4.1.2. Elderly's Contents Usage Behavior: Dynamic and Static Contents 59
4.1.3. Effect and Types of Touch on Human-robot Interaction 60
4.1.4. Touch Interaction's Expansion with More Cognitive Effort 63
4.1.5. Customized Needs and User's Expectations 64
4.2. Antecedents and Consequences Path: Granger Causality 65
4.3. Pattern Derivation: Multivariate Time Series Clustering 68
4.3.1. Dynamic Time Warping(DTW) 68
4.3.2. Global Alignment Kernel(GAK) 69
4.4. Determining The Number of Clusters: Cluster Validity Index 70
4.4.1. Silhouette Index(Sil) 71
4.4.2. Dunn Index 72
4.4.3. Context-independent Optimality and Partility Index(COP) 72
4.4.4. Davies-Bouldin Index(DB) and Davies-Bouldin* Index(DB*) 73
4.4.5. Calinski-Harabasz Index(CH) 73
4.4.6. Score Function Index(SF) 74
4.5. Research Model 75
4.5.1. Research Process 75
4.5.2. Data Collection and Preprocessing 77
4.6. Essay 2: Research Results 79
4.6.1. Granger Causality: Antecedents and Consequences Path 79
4.6.2. Multivariate Time Series Clustering and Optimal Cluster Number 85
4.6.3. Cluster 1: Antecedents and Consequences Path 88
4.6.4. Cluster 2: Antecedents and Consequences Path 90
4.6.5. Cluster 3: Antecedents and Consequences Path 92
4.6.6. Cluster 4: Antecedents and Consequences Path 94
4.7. Essay 2: Conclusion and Implication 96
V. Essay 3: Examine Social Robot Users' Usage Patterns and Quality of Life Improvement Mechanisms 99
5.1. Positive Effects of Social Robots from Various Perspectives 99
5.3.1. Elderly's Perception of Social Robot 100
5.3.2. Long Term Usage Pattern's Features 102
5.2. Pattern's Characteristics: Exponential Smoothing Model 105
5.3. Research Model 108
5.3.1. Research Process and Model 108
5.3.2. Data Collection and Preprocessing 110
5.4. Essay 3: Research Results 113
5.4.1. Usage Pattern's Features Extraction 114
5.4.2. The Mechanism of Quality of Life Improving 115
5.5. Essay 3: Conclusion and Implication 120
VI. Conclusion and Implication 124
Reference 128
국문 초록 138
Appendix 141
〈Table 1〉 Definitions of Social Robot's Function 23
〈Table 2〉 Essay 1: Summary of Data Collection 32
〈Table 3〉 Time Series Clustering Results 33
〈Table 4〉 Applying Damped-trend Model Result 37
〈Table 5〉 Applying Damped-trend and Holt Model Result 38
〈Table 6〉 Descriptive Statistics of Interaction Usage Cluster 45
〈Table 7〉 Sub-function Usage Ratio by Interaction Usage Cluster 46
〈Table 8〉 Descriptive Statistics of Contents Usage Cluster 50
〈Table 9〉 Sub-function Usage Ratio by Contents Usage Cluster 51
〈Table 10〉 Research Questions 65
〈Table 11〉 Cluster Validity Indices 75
〈Table 12〉 Essay 2: Summary of Data Collection 79
〈Table 13〉 Results of Elderly's Granger Causality 81
〈Table 14〉 Results of Multivariate Time series Clustering 86
〈Table 15〉 Results of Each Cluster's Validity Index 87
〈Table 16〉 Results of Elderly's Granger Causality, Cluster 1 89
〈Table 17〉 Results of Elderly's Granger Causality, Cluster 2 91
〈Table 18〉 Results of Elderly's Granger Causality, Cluster 3 93
〈Table 19〉 Results of Elderly's Granger Causality, Cluster 4 95
〈Table 20〉 Research Hypothesis 105
〈Table 21〉 Essay 3: Summary of Data Collection 111
〈Table 22〉 Construct and Definition of Variables 113
〈Table 23〉 Hypothesis Test Results 118
〈Figure 1〉 Social Robot 'Hyodol' 22
〈Figure 2〉 Time Series Data of 'Hyodol', Sample 23
〈Figure 3〉 Time Series Clustering Method 26
〈Figure 4〉 Exponential Smooting Model 28
〈Figure 5〉 Example of Data-driven Persona 29
〈Figure 6〉 Essay 1: Research Process 30
〈Figure 7〉 Essay 1: Visualized Research Process 31
〈Figure 8〉 Cluster A, B: Interaction Usage 34
〈Figure 9〉 Cluster C, D: Contents Usage 34
〈Figure 10〉 Cluster A, B: Interaction Usage Centroid 35
〈Figure 11〉 Cluster C, D: Contents Usage Centroid 36
〈Figure 12〉 Granger Causal in Time Series Data 66
〈Figure 13〉 Conceptual Example of Granger Causal 67
〈Figure 14〉 Example of Multivariate Time series Clustering, GAK 70
〈Figure 15〉 Essay 2: Research Process 76
〈Figure 16〉 Essay 2: Visualized Research Process 77
〈Figure 17〉 Elderly's Antecedents and Consequences Path 82
〈Figure 18〉 Results of Impulse Response Analysis 84
〈Figure 19〉 Results of Each Cluster's Validity Index Majority 87
〈Figure 20〉 Elderly's Antecedents and Consequences Path: Cluster 1 89
〈Figure 21〉 Elderly's Antecedents and Consequences Path: Cluster 2 91
〈Figure 22〉 Elderly's Antecedents and Consequences Path: Cluster 3 93
〈Figure 23〉 Elderly's Antecedents and Consequences Path: Cluster 4 95
〈Figure 24〉 Essay 3: Research Process 109
〈Figure 25〉 Essay 3: Research Model 110
〈Figure 26〉 Research Model with Hypothesis 114
〈Figure 27〉 Research Model of Mechanism Identification 116