Research On COTS Evaluation Methods And Key Techniques

COTS evaluation and selection is the key activity of COTS-based software development (CBD).To provide a decision-making model and quantitive decision support is always the aim of CBD.In this paper, the multiple COTS evaluation and selection method and key techniques in COTS intensive system are well researched.In this paper, a process model for multiple COTS selection called MOTS is presented which consists of two levels. MOTS comprises four interrelated steps. They are system decomposition, alternatives identification, local evaluation and global selection.First the system is decomposed into loosely-coupled subsystems through classes clustering based on scenarios-mapping. During system decomposition, requirements are allocated into subsystems to support concurrent COTS identification and evaluation. Each candidate COTS is evaluated based on its fitness, risk and tailoring effort. With a set of candidates passed into global selection, decision-mkers make a tradeoff among cost, quality and risk through the iterative pre-selection and optimization.This well-defined process model provides a guideline for COTS evaluation and selection activity.The global selection is defined as an optimization problem under given constraints to find a best combination of COTS for maximizing the global fitness. To solve this problem, mathematical programming is used which privides an objective-cost model and quantitative support for multiple COTS selection.Based on gap analysis theory, a requirements-driven local evaluation method is proposed.The gap of functionality between COTS and requirements and corresponding adjustment cost are measured in function point model.The fitness of COTS to requirements is defined as the ratio of the software size of target system to the total "cost" of COTS that is also measured in software size.Multi-level fuzzy synthetical evaluation model is applied to evaluate the nonfunctional attributes with fuzzy characteristics and a new method to determine the membership degree is proposed. Different with the conventional method, linguistic variable is combined with qualifier to express the evaluation result. Each linguistic variable is viewed as a mark on the ruler and the area between two marks is called a fuzzy area. When scoring, qualifier is used to express the membership degree to each fuzzy area. With linguistic variable and qualifier mapped into a number in [0,1], the evaluation result is transferred into a fuzzy number. The fuzzy synthetical evaluation result is transformed into a crisp number based on the center of gravity for fuzzy set. This method is computationally simple, intuitive in nature, easy to implement and with higher semantical resolution.