Cost, Time and Quality Trade-off in Software Engineering
Abdelelah Ghaleb Farhan Saif;
Abstract
Cost, schedule and quality are highly related factors (objectives) in software development. They constitute the three sides of the triangle. It is hard to raise the quality without raising either the cost or schedule or both for the software under development. Similarly, development schedule cannot be decreased without sacrificing software product quality and/or raising development cost.
The Discrete Time, Cost and Quality Trade-off Problem (DTCQTP) is special case of this problem in which there are number of execution modes for each activity (modes can be either bids offered by subcontractors to develop an activity where each bid has duration, cost and quality), and the best execution mode of every activity should be determined in order to optimize the total cost, time and quality of the software project.
This work includes:
1. Introducing an optimized cost-quality model based on dataset in Constructive Cost Model (COCOMO) format by adapting Intelligent Water Drops (IWD) algorithm for optimizing COCOMO II Post Architecture (PA) and Constructive Quality Model (COQUAMO/COQUALMO) models to achieve more accurate estimation/prediction of software development effort and hence cost, time and quality on the project level and to allow making trade-off analysis. For this model, the prediction accuracy of IWD is compared with the original models, genetic algorithm (GA) and Problem Data Based Optimization for Continuous Optimization (PDBO-CO) which is developed for this purpose to enable project manager selects the suitable model. The experiment has been conducted on NASA 93 dataset.
2. Introducing a cost-quality model that optimizes effort and hence cost, time and quality (making trade-off analysis) by finding the best project options on the project level. For this model, a comparison is done among IWD, GA and Problem Data Based Optimization (PDBO) in terms of the stability of selecting the best project options, quality of solution and Processing Time (PT) to enable project manager selects the best one. The experiments are conducted on an imaginary software project with size of 25 Thousand Source Lines of Code (25KSLOC) and on Jet Propulsion Laboratory (JPL) flight software project.
3. Developing an application (solving DTCQTP). The algorithms that are used to solve such problem (DTCQTP) are IWD, GA, Ant Colony Optimization (ACO), Egyptian Vulture Optimization (EVO) and PDBO which is developed for this purpose. By using some examples from the literature, the results of these algorithms are compared in terms of quality of solution and efficiency to enable project manager selects the best one.
4. Finally, comparing the proposed PDBO-CO with Intelligent Water Drops for Continuous Optimization (IWD-CO) for function optimization.
The results are as follow:
For the optimized cost-quality based on NASA 93 dataset, we conclude: Effort and time obtained by PDBO-CO, IWD, and GA are closer to the actual ones than that obtained by the original COCOMO II PA model. By doing a comparison among PDBO-CO, IWD, GA and the original COCOMO II PA model regarding prediction accuracy, the Mean Magnitude of Relative Error (MMRE) of PDBO-CO, and IWD for effort is equal and the best, and for the time, the MMRE of IWD is the smallest. The MMRE of GA for effort and time is larger than that of both PDBO-CO and IWD. Prediction (0.25) (PRED(0.25)) of PDBO-CO and of GA for effort and time is almost equal and the best. PRED(0.25) of IWD for effort and time is smaller than that of PDBO-CO and GA. The MMRE and PRED(0.25) of COCOMO II PA for effort and time is the worst among all. Requirements (Req), Design (Des), Coding (Code) and total defects obtained by IWD algorithm and the original COQUAMO model are closer to the actual ones. By doing a comparison among PDBO-CO, IWD, GA and the original COQUAMO model regarding prediction accuracy, the MMRE and PRED(0.10) of IWD, PDBO-CO and GA are somewhat comparable. The MMRE and PRED(0.10) of original COQUAMO model is the worst, except for Code defects is somewhat better. PT of PDBO-CO is the lowest and PT of GA is the second lowest whereas PT of IWD is the largest.
For the cost-quality model which doesn’t depend on dataset, we conclude: The results for all goals (Faster Better Cheaper FBC, Better Cheaper BC, Better Faster BF and Cheaper Faster CF) are stable for 10 runs of PDBO and are satisfactory, whereas for IWD and GA are unstable. For the two projects, PDBO results generally are the best, whereas IWD results are the worst. GA results generally are the second best. For all goals, PT is recorded in ascending order as follow: PDBO, GA and IWD.
The Discrete Time, Cost and Quality Trade-off Problem (DTCQTP) is special case of this problem in which there are number of execution modes for each activity (modes can be either bids offered by subcontractors to develop an activity where each bid has duration, cost and quality), and the best execution mode of every activity should be determined in order to optimize the total cost, time and quality of the software project.
This work includes:
1. Introducing an optimized cost-quality model based on dataset in Constructive Cost Model (COCOMO) format by adapting Intelligent Water Drops (IWD) algorithm for optimizing COCOMO II Post Architecture (PA) and Constructive Quality Model (COQUAMO/COQUALMO) models to achieve more accurate estimation/prediction of software development effort and hence cost, time and quality on the project level and to allow making trade-off analysis. For this model, the prediction accuracy of IWD is compared with the original models, genetic algorithm (GA) and Problem Data Based Optimization for Continuous Optimization (PDBO-CO) which is developed for this purpose to enable project manager selects the suitable model. The experiment has been conducted on NASA 93 dataset.
2. Introducing a cost-quality model that optimizes effort and hence cost, time and quality (making trade-off analysis) by finding the best project options on the project level. For this model, a comparison is done among IWD, GA and Problem Data Based Optimization (PDBO) in terms of the stability of selecting the best project options, quality of solution and Processing Time (PT) to enable project manager selects the best one. The experiments are conducted on an imaginary software project with size of 25 Thousand Source Lines of Code (25KSLOC) and on Jet Propulsion Laboratory (JPL) flight software project.
3. Developing an application (solving DTCQTP). The algorithms that are used to solve such problem (DTCQTP) are IWD, GA, Ant Colony Optimization (ACO), Egyptian Vulture Optimization (EVO) and PDBO which is developed for this purpose. By using some examples from the literature, the results of these algorithms are compared in terms of quality of solution and efficiency to enable project manager selects the best one.
4. Finally, comparing the proposed PDBO-CO with Intelligent Water Drops for Continuous Optimization (IWD-CO) for function optimization.
The results are as follow:
For the optimized cost-quality based on NASA 93 dataset, we conclude: Effort and time obtained by PDBO-CO, IWD, and GA are closer to the actual ones than that obtained by the original COCOMO II PA model. By doing a comparison among PDBO-CO, IWD, GA and the original COCOMO II PA model regarding prediction accuracy, the Mean Magnitude of Relative Error (MMRE) of PDBO-CO, and IWD for effort is equal and the best, and for the time, the MMRE of IWD is the smallest. The MMRE of GA for effort and time is larger than that of both PDBO-CO and IWD. Prediction (0.25) (PRED(0.25)) of PDBO-CO and of GA for effort and time is almost equal and the best. PRED(0.25) of IWD for effort and time is smaller than that of PDBO-CO and GA. The MMRE and PRED(0.25) of COCOMO II PA for effort and time is the worst among all. Requirements (Req), Design (Des), Coding (Code) and total defects obtained by IWD algorithm and the original COQUAMO model are closer to the actual ones. By doing a comparison among PDBO-CO, IWD, GA and the original COQUAMO model regarding prediction accuracy, the MMRE and PRED(0.10) of IWD, PDBO-CO and GA are somewhat comparable. The MMRE and PRED(0.10) of original COQUAMO model is the worst, except for Code defects is somewhat better. PT of PDBO-CO is the lowest and PT of GA is the second lowest whereas PT of IWD is the largest.
For the cost-quality model which doesn’t depend on dataset, we conclude: The results for all goals (Faster Better Cheaper FBC, Better Cheaper BC, Better Faster BF and Cheaper Faster CF) are stable for 10 runs of PDBO and are satisfactory, whereas for IWD and GA are unstable. For the two projects, PDBO results generally are the best, whereas IWD results are the worst. GA results generally are the second best. For all goals, PT is recorded in ascending order as follow: PDBO, GA and IWD.
Other data
| Title | Cost, Time and Quality Trade-off in Software Engineering | Other Titles | التوازن بين الكلفة و الزمن والجودة في هندسة البرمجيات | Authors | Abdelelah Ghaleb Farhan Saif | Issue Date | 2017 |
Recommend this item
Similar Items from Core Recommender Database
Items in Ain Shams Scholar are protected by copyright, with all rights reserved, unless otherwise indicated.