Estimating Final Software Development Hours for DoD Program Cost Evaluation using Regression Analysis and Machine Learning

This praxis develops a decision support tool for evaluating a DoD program’s final software development hours using statutorily required software attributes reported by DoD contractors at the beginning of a project. The tool’s inputs consider attributes such as estimates for initial software hours, number of lines of code, number of peak staff, schedule, and the software application domain. The praxis describes common software estimation techniques for different software development methods and analyzes a dataset of over three hundred DoD software efforts; exploring the relationship between initially reported software attribute estimates in a program’s proposal phase to actual software development hours collected at the end of a program. Estimations utilize regression analysis and machine learning algorithms applied to the software program datasets. Analysis shows that machine learning algorithms produce prediction models with less error and better accuracy compared to typical regression methods and models. Using final software development hours as a surrogate for software cost, the estimation tools will enable DoD decision makers and cost estimators to infer a program’s final most probable cost as part of contract source selection.

Author

• Schonhoff, Shane

Language

• en

Keyword

• Software Cost Estimation
• machine learning
• DoD
• SRDR

Date created

• 2021

Type of Work

• Dissertation

License

• All rights reserved

Rights statement

• In Copyright

GW Unit

• Engineering Management

Degree

• D.Engr.

Advisor

• Holzer, Thomas

Committee Member(s)

• Eveleigh, Timothy
• Etemadi, Amir

Persistent URL

•  https://scholarspace.library.gwu.edu/etd/v405sb146