An Intelligent Approach to Support the Making of Architectural Design Decisions in the Context of the Evaluation of Software Architectures

Abstract

Software Engineering needs novel tools to pursue further the goals of achieving software quality, facing the changing role of software. In this context, Software Architecture plays a key role because it directly affects the final quality. Software Architecture Evaluation validates if the architecture achieves the quality requirements, and triggers a set of design decisions. The decision-making is a very complex process driven by several human factors. Nowadays, it is argued that Artificial Intelligence-based practices can assist this process. In this work, an Artificial Intelligence-based approach for assisting architects in the design decision-making process driven by quality attributes is proposed. This first version combines quality-attribute models and an intelligent agent to support software architecture evaluation. It applies Reinforcement Learning tools to obtain a sequential architectural pattern application policy by simulation. A case study and a set of experiments illustrate the proposal with patterns commonly used in software industry.