The Analysis And Metrics Of The Characteristics In The Large-scale Software Weighted Networks

Researching software architectures with the theory of complex networks provides a new meaning of measuring the complexity of software structure. Previous studies mainly extracting software topology structure to unweighted networks, then analyze the rules and characteristics of software structure. But there are various types of relationships between software networks nodes. The properties of relationships and their linking tightness are different. Taking software structure as unweighted networks ignores the diversity of relationships, and may lead inaccurate conclusions.Combing the theoretical knowledge of weighted complex networks, this paper proposes software structure weighted networks model based on coupling between classes. According to tightness of the relationships between classes, the model gives different weights to different types of edges. The paper designs and implements software topology structure weighted networks analysis platform. The platform extracts software source codes to weighted network, computes its characteristics and analyzes the structural characteristics of the software networks. Using the software structure weighted networks model and analysis platform, the paper analyzes hundreds of object-oriented famous open source software systems with different scale. The results show that the software weighted networks also present characteristics of complex networks and contain statistical laws. This article uses statistical methods to analyze and measure the characteristics value of the weighted networks, which could give a more comprehensive and accurate description of software structure characteristics. The paper compares and analyzes the law of topological features in weighted network and unweighted networks. From the microscopic point, the paper takes the specific nodes for example and analyzes the effectiveness of the software structure weighted network analytical method. The results show that weighted networks analytical method can be more comprehensive and accurate.The study of software structure weighted networks is the complement and improvement to software unweighted networks, which can further understand structural complexity of large-scale software systems, more accurately measure the quality of software structure and provide better guidance for the software development and maintenance.