Analogibasert ekspertestimering av software-prosjekter

Abstract

Estimation of new projects is a complex part of software engineering. Project leaders need to understand and complete many different tasks in constructing the actual program. If these tasks are not completely apprehended, acorrect estimate for the software can not be constructed.

This thesis is about using information from historical projects to improve the precision of software estimates for project leaders. Here this is done by constructing a connection between experience reports and a experience database. The idea is that by connecting experience reports to the experience database project leader will have a chance to understand the reasons to the numbers presented in the experience database. By doing this project leaders have an opportunity to make better analogy based expert decision.

One specific solution is constructed after an analysis of experience reports, an experience database, interviews with project leaders and existing estimation techniques. The result from this analysis is a specific framework that can be implemented in the firm who's experience reports and experience database were analyzed. From the specific solution, a general method for constructing experience reports and a experience database are described. The general solution can be used of all firms who want to support the expert estimate project leaders constructs.