SOFTWARE EFFORT ESTIMATION USING OBJECT POINT ANALYSIS

SHIMELS, TILAHUN

dc.contributor.author	SHIMELS, TILAHUN
dc.date.accessioned	2024-12-06T11:49:07Z
dc.date.available	2024-12-06T11:49:07Z
dc.date.issued	2023-07
dc.identifier.uri	http://ir.bdu.edu.et/handle/123456789/16320
dc.description.abstract	Software effort estimation is a critical aspect of software project management that involves predicting the energy required to develop the software project from planning to deployment. Delivering high-quality software that meets users' needs depends on accurately predicting software development efforts early. However, previous software projects' effort estimation methods, such as source lines of code (SLOC) and function point size metrics, have limitations in estimating effort at the specification stage. To solve these problems, we propose an object point analysis (OPA) technique for software effort estimation based on a deep neural network (DNN) approach to predict software project effort easily and accurately at the specification stage. However, more historical data is needed for OPA, as it is a new technique. We experimented with a forward approach to convert function points into their equivalent object points and prepared new historical data for OPA as one contribution to this study. To implement this object-point technique, we used an experimental science researcnh design approach. This study conducted a comparative analysis by evaluating three machine learning algorithms, namely Kneighbors regression (KNR), linear regression (LR), and support vector regression (SVR), and a proposed DNN model on two datasets. The Desharnais dataset was evaluated using the function point analysis (FPA) technique, while the new dataset was evaluated using the OPA technique. The R2 values obtained for the algorithms and the proposed DNN model on the new dataset using the OPA technique were 0.97, 0.99, 0.9, and 0.99 for KNR, LR, SVR, and the proposed DNN model, respectively. For the Desharnais dataset using the FPA technique, the R2 values were 0.66, 0.67, 0.34, and 0.75 for KNR, LR, SVR, and the proposed DNN model, respectively. The results showed that all algorithms performed better on the new dataset using the OPA technique than on the Desharnais dataset using the FPA technique. Overall, the study suggests that the proposed DNN model outperforms other machine learning algorithms on both datasets, particularly on the new dataset using the OPA technique. Keywords: Software development effort estimation, object point analysis, deep neural network model, object count estimation, regression analysis	en_US
dc.language.iso	en_US	en_US
dc.subject	Software Engineering	en_US
dc.title	SOFTWARE EFFORT ESTIMATION USING OBJECT POINT ANALYSIS	en_US
dc.type	Thesis	en_US