The Research Of Trustworthy Software Estimation Model Based On Improved Projection Pursuit Technique

Software has been applied to the political,economic,military,culture and other aspects,because of the continuously increasing of size,complexity and intelligent level of software products,as well as the complexity of application environments and uncontrollability in the process of software development,software products exist many defections that lead to the untrustworthiness.Software trustworthiness estimation methodology is one of the most important and difficult problems in the area of trustworthy software research in recent years.Through theliterature reviewing and investigationof software enterprises,and systematically study the latest research results abouttrustworthy software measurement modelof domestic and abroad,the research of trustworthy software attributes mainly focuses on the security,survivability,fault tolerance,reliability,safety,real time and so on.Due to the diversity of software trustworthy attributes and the difference between those attributes;it is difficult to fully reflect the overall level of software trustworthiness with a single trustworthy attribute.To further explore the relationship between software trustworthy attributes,through the analysis of the complexity of the software trustworthy attributes and the software trustworthy measurement model,according to the ISO/IEC25010 software quality standard,establish a set of software trustworthy index system based on behavior trustworthiness,capability trustworthiness and user's behavior trustworthiness.Consider that different software requires different trustworthy emphasis;users can adjust the measurement index.Due to the high dimensional characteristics of the trustworthy software measurement index attributes,this paper aims at the optimal problem of projection direction optimization in the projection pursuit method,introduce the gravitational search algorithm(GSA),and the penalty function(SUMT)method is used to construct a new projection indicator function,as a result,simplifies the complexity of the projection indicator function.Furthermore,combine the analytic hierarchy process theory and entropy weight method,reduce the trustworthy software attributes from high dimension to low dimension,and a trustworthy software estimation model based on improved projection pursuit technique(PP-TSCMM)is proposed.In order toreflect the trustworthy software level in a more intuitive way,and a trustworthy software classification method based on Affinity Propagation algorithm is proposed.Case study shows that the result of comprehensive software measurement model is basically equal to practical application situation;provide a new method for estimate trustworthiness level of large and complex software systems.