Companion robots (CRs) have been proposed as alternatives to companion animals such as dogs and cats. For developing and improving CRs that owners will love for a long time, constant assessment of the attachment that CR owners feel toward them is important. However, pervious methods based on questionnaires are costly because they need to be conducted periodically to monitor the owners’ subjective level of attachment over time. Therefore, this study investigated alternative indicators for the subjective level of attachment based on low-cost and continuously accessible CR log data. In this study, we analyzed the relationship between the subjective level of attachment, obtained through questionnaires administered to 287 CR owners, and owner behaviors extracted from CR log data. The results showed that two behavioral features – (1) the number of days when owner behavior occurred within one hour before CR bedtime and (2) the frequency of owner behavior while the CR is being held by the owner – exhibited stronger correlations with subjective level of attachment compared to behavioral features measured without temporal or spatial contextual constraints. Furthermore, factor analysis conducted on the questionnaire responses extracted three factors, Comfort, Protectiveness, and Intimacy, among which the third factor, Intimacy, was found to be correlated with the space-specific behavioral feature. These findings suggest that the behavioral features extracted from the CR log data can serve as more effective indicators of the subjective level of attachment when behaviors are constrained to specific temporal and spatial contexts, and when attachment is represented as multidimensional data.