Mined-knowledge driven software quality monitoring and diagnosis environment

本研究旨在为组织中的软件开发人员和维护人员设计一个决策支持环境，以使他们能够监控系统的结构和行为方面，并提供有关维护特定软件质量的建议、警报和警告。拟议的环境利用现代技术来指定和合成可定制的决策流程图，作为指导模型，对组织内制定的或软件维护文献中确立的最佳做法进行编码。然后执行指导模型以交互地将决策流从决策流程图的根节点引导到叶节点，在叶节点做出决策并通知用户。在决策流程图的每个节点上，可以向指导模型提供各种类型的信息。这些类型的信息被归类为“系统信息”，包括系统的静态、动态或设计相关的数据和关系，或者被归类为“已挖掘知识”，其形式为模式和趋势，包括：通过数据挖掘算法从“系统信息”中提取的关联规则、执行跟踪模式、概括和聚类。然后，指南模型提供不同类型的建议或警报，以及决策流程图中可能的下一步列表，供用户选择。指南模型可以指定不同的软件维护活动，例如：纠正性、完美性、适应性或预防性操作，以确保相应组织在其开发和演化阶段拥有健康的软件系统资产。拟议环境的主要特点包括纳入挖掘的知识，以确保方法的复杂性，以及个性化和易用性，以确保各组织在不同应用领域的软件维护人员采用。

This research aims at devising a decision support environment for software developers and maintainers in an organization, in order to enable them to monitor the structural and behavioral aspects of the system and to provide recommendations, alerts, and warnings with regard to maintaining specific software qualities. The proposed environment takes advantage of modern technologies to specify and synthesize customizable decision-flow diagrams as guideline models that encode the best practices developed in the organization or  established in the software maintenance literature. The guideline model is then executed to interactively direct the flow of decision from the root node of the decision-flow diagram to a leaf node, where a decision is made and the user is notified.At each node of the decision-flow diagram, various types of information can be supplied to the guideline model. These types of information are categorized as either "system information", including system's static, dynamic, or design related data and relations, or as "mined knowledge", in the form of patterns and trends, including: association rules, execution trace patterns, generalizations, and clusters, that are extracted from the "system information" by the means of data mining algorithms. The guideline model then providesdifferent types of recommendations or alerts, as well as a list of possible next steps in the decision-flow diagram for the user to choose from. The guideline models can specify different software maintenance activities such as: corrective, perfective, adaptive, or preventive operations that ensure a healthy software system asset for the respective organization throughout its development and evolution stages. The key characteristics of the proposed environment include the incorporation of mined knowledge that ensures sophistication of the approach, as well as customization and ease of use that ensures adoption by the software maintainers of the organizations in different application domains.