Analysis of Combining Software Estimation Techniques

Abstract

In this paper, analysis of the result obtained fromexperiment withthree projects is described. Combinationof different estimationtechniques helps estimator toreduce error and keep control overthe deviation ofestimates away from actual. As a response tosurvey fromestimators working in software industry, It isrevealedthat, according to stages, estimates are termedasbudgetary, initial, progressive and closure. As abudgetaryestimates are dominated by analogy basedestimation techniques,It can be complemented by COCOMOII ApplicationComposition Model. The Initial estimatesis average ofCOCOMO II Early Design model with Objectpoint sizing, whichcan be complemented by Functionpoint and Usecase point basedestimation. TheProgressive estimates are calculated by averagingofCOCOMO II Post Architecture Model and Class pointbasedestimates. It is observed that effort estimatesare more accuratethan using only COCOMO II.