论文部分内容阅读
Estor,是一个根据以往的研究收集口头记录提出并精练而成的基于案例的类比推理模型。它结合了五个类比案例问题的求解过程:问题表述,类比案例检索,结果转换,属性映射,和非一一对应的调整。这五个类属过程由有关系家以其专门领域内的知识进行了补允。其结果表现为十五项软件费用估算任务,其中十项曾由有关系家解答,其余五项为新添入任务。为了进行比较,同时要求专家对该五项任务也做出估算。把Estor做出的估算与专家以及由FunctiontPoint和COCOMO对这项计划做出的估算进行了比较,发现由专家(人)和Estor(模型)分别控制而做出的估算之间的差异是明显的。对十五项计划做出的估计(由专家和Estor模型)与实际费用值的相关系数分别为0.98和0.97。此外,这些系数与COCOMO以及Function Points产生的值则相差更远。本文将着重讨论模型和有关专家之间的差异。
Estor, a case-based analogical reasoning model proposed and refined based on previous research that collected verbal records. It combines the process of solving five analogical case problems: problem presentation, analogous case retrieval, result conversion, attribute mapping, and non-one-to-one correspondence adjustment. The five generic processes are compensated by the relevant family with knowledge in their specialized fields. The results showed fifteen software cost estimation tasks, of which ten have been answered by the relevant family, the remaining five new additions to the task. For the sake of comparison, experts are also required to make an assessment of these five tasks as well. Comparisons of estimates made by Estor to those of experts and estimates made by FunctiontPoint and COCOMO for this program revealed significant differences between estimates made by experts and people under Esto (model) control . The correlation coefficients for the fifteen programs (by experts and Estor models) and actual costs were 0.98 and 0.97, respectively. In addition, these coefficients differ further from COCOMO and Function Points. This article will focus on the differences between the model and the experts involved.