Data analytics for decision support in software release management
Date
2018-04-26
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The dynamism and uncertainty in iterative software development demand for frequent releases with shorter duration, reduced scope, and strong stakeholder involvement. The lack of custom-built integrated support for release management in iterative releases mainly motivated this research. Key contribution of this thesis is the proposed Plan-Monitor-Improve (PMI) Framework. PMI integrates i) strategic planning of a release, ii) monitoring release readiness and iii) improving release readiness. Methodologically, this thesis is positioned at the intersection of three key concepts: i) release management ii) data analytics techniques, and iii) decision support systems. The PMI-Plan method introduces a multi-objective particle swarm optimization-based planning method to support the multi-dimensional nature of iterative releases. Maximum stakeholder incorporation is achieved in a three-stage interactive modeling. The research is triggered by industry collaborators in an NSERC CRD project and studies with a prototype tool in real world planning problems. Inspired by statistical process control approaches, the PMI-Monitor method continuously monitors release readiness. PMI-Monitor formulates release readiness prediction as a binary classification problem and solves using machine learning classifiers. The PMI-Improve method addresses release readiness by identifying key attributes to be used for decision-making. PMI-Improve also applies analogy-based reasoning for proactively identifying bottlenecks and their characteristics. The release readiness improvement factor identification is formulated as a search problem and solved using a genetic algorithm. The PMI is an integrated decision centric framework, custom built to support iterative release management. PMI integrates a portfolio of data analytics techniques (e.g. analogical reasoning, multi-objective optimization, classification, process control, interactive search) to gain better insight from the frequent release iterations and strong stakeholder involvements. PMI integrates the characteristics of iterative development and applies it towards an operational platform to make better data driven decisions. Initial empirical evaluation of PMI further supports applicability and benefits of the framework in real world iterative software development.
Description
Keywords
Data analytics, Decision Support, Release management
Citation
Didar Al Alam, S. M. (2018). Data analytics for decision support in software release management (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/31856