Analogy-X: Providing Statistical Inference to Analogy-Based Software Cost Estimation.
Abstract
Data-intensive analogy has been proposed as a means of software cost estimation as an alternative to other data-intensive methods such as linear regression. Unfortunately, there are drawbacks to the method. There is no mechanism to assess its appropriateness for a specific data set. In addition, heuristic algorithms are necessary to select the best set of variables and identify abnormal project cases. We introduce a solution to these problems based upon the use of the Mantel's correlation randomization test called Analogy-X. We use the strength of correlation between the distance matrix of project features and the distance matrix of known effort values of the data set. The method is demonstrated using the Desharnais data set and two random data sets, showing 1) the use of Mantel's correlation to identify whether analogy is appropriate, 2) a stepwise procedure for feature selection, as well as 3) the use of a leverage statistic for sensitivity analysis that detects abnormal data points. Analogy-X thus provides a sound statistical basis for analogy, removes the need for heuristic search, and greatly improves its algorithmic performance. Index Terms-Cost estimation, management, software engineering, analogy. Ç 1 INTRODUCTION S OFTWARE project managers require reliable methods for estimating software project costs. For more than 25 years, there has been considerable research effort directed toward software cost estimation. There is still no definitive solution as to the best method of software cost estimation. Among data-intensive algorithmic cost estimation methods, analogy-based software cost estimation (also referred to as Case-Based Reasoning (CBR)) is currently a popular alternative to constructing estimation models using linear regression. However, there are problems inherent in the method: There is no method to measure the appropriateness of the analogy approach for a specific data set. . The existing methods for selecting the best set of variables and project cases to build an analogy is based either on brute-force search or heuristic algorithms (e.g., greedy search or hill climbing). . There is no simple method to identify abnormal projects. These is