Development of a Model for Conflict Resolution in the Requirements Engineering process of Software Systems.
Loading...
Date
2016
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Obafemi Awolowo University
Abstract
In this research a framework for conflict resolution was formulated, a model was designed, implemented and evaluated in the requirement engineering process of software development. This was with the view of identifying and resolving conflicts among multiple stakeholders’ expectations that often arise during requirements analysis.
The research employed both qualitative and quantitative research approaches. Requirements were elicited from stakeholders using interview, case study, observation and secondary data. Ten (10) staff from the Pharmacy Department of Obafemi Awolowo University Teaching Hospitals Complex (OAUTHC) and five (5) software developers from software industry were the respondents. The requirements were analysed using the requirement filtering technique. Delphi method was used to resolve the inputs from the requirements. Clustering Algorithm was used to formulate the framework for conflict resolution. The model was specified in Unified Modeling Language, and the system was implemented using Java programming language. The system was tested for completeness and consistency using recall and precision as parameters.
The result showed the Kendal Coefficient of Concordance (KcoC) W was 0.000115598 0.00. This indicated the existence of disagreement among the stakeholders’ subjective views. With the clustering analysis, the result showed 5 clusters and their corresponding centroids and standard deviation. The analysis carried out indicated that cluster 3 is more reliable with the lowest standard deviation. This is because the data were clustered closely around the centroids. Cluster 3 had an average standard deviation of 0.61, while clusters 1, 2, 4 and 5 had high standard deviation of 0.95, 0.78, 0.86 and 1.31 respectively. Further analysis of the result revealed that 81.19% of all the attributes with the lowest standard deviation were in cluster 3, and 18.81% of the attributes with the lowest standard deviation were distributed in clusters 1, 2, 4, and 5. This also makes cluster 3 more effective. Based on the final resolution on cluster 3, a generated list of ordered requirements was produced. The output indicated the order of priorities finally assigned to each requirement. Seventy-seven (77) requirements were seen to be “very high”, making it 76.24% of the entire sets of requirements, while twenty-four (24) requirements were finally seen to be “high”, making it 23.76% of the entire sets of requirements. The model evaluation revealed that clusters 3 and 4 had 100% recall while clusters 1, 2 and 5 had 50%, 66.7% and 80% respectively. The model evaluation also indicated 100% precision in clusters 2, 3 and 5 respectively, while custers 1 and 4 had 66.7% and 78.95% respectively. The F-measure showed the harmonic mean of precision and recall to be 0.57, 0.80, 1.00, 0.88 and 0.89 for clusters 1, 2, 3, 4 and 5 respectively . Consequently, the F-Measure of cluster 3 with the value of 1.00 (100%) is most effective and reliable. This implies that all the instances originally belonging to cluster 3 were correctly clustered and all clustering results also belong to the original instances in cluster 3. The total sum of false negative (FN) in the model was 6.0 (14.29 %). This means the model is reliable with an accuracy level of 85.71%.
In conclusion, the study opined that resolving conflicting views of stakeholders’ in software development is crucial and vital in order to reduce the cost of software development.
Description
xix,288 Pages
Keywords
Conflict Resolution, Engineering Process, Software Systems, Pharmacy Department, Obafemi Awolowo University Teaching Hospitals Complex
Citation
Gambo,I.P.(2016).Development of a model for conflict resolution in the requirements engineering process of software systems.Obafemi Awolowo University