An Exploratory Analysis of Semantic Network Complexity for Data Modeling Performance

Database modeling performance varies across different constructs. For example, it is usually easier to model a binary relationship than a ternary relationship. Based upon the empirical performance data, ternary relationships are thought to be more complex than binary ones. This paper investigates the relationship between user modeling performance and the complexity of the data model constructs. A complexity estimate is proposed that measures complexity based on three different aspects: component, coordinative and coupling complexity. The aggregation of the three provides the total complexity estimate. Two semantic network variations that users might use are suggested, and their complexity values compared against known user performance. Regression results show reasonable R square values. This analysis suggests that using semantic networks could be a practical way to estimate modeling complexity and user performance. Future research can consider more refined variations of the semantic network, to account for training, experience, and different data models.