Measuring Complexity at the Requirements Stage: Spectral Metrics as Development Effort Predictors

Key findings from this study

• The study found that spectral measures predict integration effort with correlations exceeding 0.95, substantially exceeding predictions from structural metrics.
• The researchers demonstrate that density-based metrics lack significant predictive validity for integration effort, despite their prevalence in network analysis.
• The authors report that eigenvalue-derived measures capture cognitive and effort dimensions inaccessible to simpler connectivity-based structural metrics.

Overview

Requirements specifications exhibit structural complexity that propagates downstream through system architecture and implementation. Structural complexity in requirements remains inadequately measured despite its consequences for project cost, schedule, and success. This research applies Natural Language Processing to extract structural networks from textual requirements and employs spectral graph metrics to predict integration effort.

Methods and approach

The authors extracted structural networks from textual requirements using Natural Language Processing methods. They validated metric predictive validity through controlled experiments using molecular integration tasks as structurally isomorphic proxies for requirements integration. This approach eliminated confounding factors including domain expertise and semantic ambiguity while preserving topological equivalence between molecular graphs and requirement networks.

Results

Spectral measures derived from eigenvalue analysis predicted integration effort with correlations exceeding 0.95, substantially outperforming structural metrics that achieved correlations above 0.89. Density-based connectivity metrics demonstrated no significant predictive validity. The superior performance of spectral measures indicates that eigenvalue-derived properties capture cognitive and effort dimensions that simpler structural metrics cannot represent.

Implications

These findings establish that spectral graph metrics provide a validated quantitative foundation for assessing requirements complexity at early development stages. The demonstrated predictive validity suggests that eigenvalue-based measures can identify requirements specifications likely to incur substantial integration effort before downstream system design commences. Organizations applying these metrics during requirements engineering could prospectively detect complexity-driven risk factors and adjust specification strategies accordingly.

Scope and limitations

This summary is based on the study abstract and available metadata. It does not include a full analysis of the complete paper, supplementary materials, or underlying datasets unless explicitly stated. Findings should be interpreted in the context of the original publication.