A software project scheduling assistant design and analysis

Abstract

Conventional software development scheduling systems, while doing an adequate job in resource scheduling, have lacked the ability to accurately estimate required implementation effort. Typically the analyst is required to supply values for input parameters which the scheduling system uses to produce required effort. The typical analyst often encounters difficulty in providing accurate values and a poor estimate results. Newer rule based systems correct some of the deficiencies and offer an improvement over the earlier methods. This work proposes that a software development scheduling system operate in the same manner as an experienced analyst who tends to draw from his experience and compare the current project to similar projects developed in the past. The purpose of this research is to design a system which emulates an experienced analyst by matching elements of the current project with similar elements of projects developed in the past. Knowledge bases are designed for the current project, historical projects, and development personnel using semantic nets to represent knowledge about each. A common features selection method is developed which matches processes or modules in a current project knowledge base with similar processes or modules in a historical knowledge base. Experienced analysts were recruited to compile a common features list for program modules. The feasibility of using the common features module matching scheme is investigated using data compiled from completed program assignments from an advanced level programming class. The analysis indicates that the common features match process is valid and can be implemented as part of a scheduling system. The system can determine whether two modules are similar, and, therefore, emulate an analyst using the experience method of estimation. By referencing stored knowledge about actual similar projects previously completed, this system will produce better estimates than those used in the past, and should make the scheduling assistant very valuable to most organizations. This study shows, however, that the system will be very complex and expensive to build.