NASA Analogy Software Costing Tool: A Web-based …csse.usc.edu/new/wp-content/uploads/2016/05/2016... · NASA Analogy Software Costing Tool: A Web-based Tool or ... Number of Deployables
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Jairus Hihn & Michael SaingSystems Analysis, Modeling & ArchitectureJet Propulsion Laboratory, California Institute of Technology
NASA Analogy Software Costing Tool:A Web-based Tool
All captured in on-line help as well as a user guide
NC State University
10
Reminder - What We Learned from Methodology
ª There are a variety of models whose performance are hard to distinguish (given currently available data) but some models are better than others
ª If one has sufficient data to run COCOMO or a comparable parametric model then the best model is the parametric model
ª When insufficient information exists then a model using only system parameters can be used to estimate software costs with relatively small reduction in accuracy. The main weakness is the possibility of occasional very large estimation errors which the parametric model does not exhibit.
ª A major strength of the nearest neighbor and spectral clustering methods is the ability to work with a combination of symbolic and numerical data
ª While a nearest neighbor model performs as well or better as spectral clustering based on MMRE, spectral clustering handles outliers better and provides a structured model with more capability
ª Web-based tool run from a serverª Has access control
ª Written Pythonª Uses a number of Python Packages
ª MySQL databaseª Django Web and Restª Javascript. ª The tool was deployed on the nextgen servers, which run
CentosOS. ª The graphing library used is plotly.js, which is open source
and fully local.
NC State University
16
ASCoT -1
Inputs
Estimate
On-Line help windows
NC State University
17
ASCoT – 2- Cluster Parameter Variation
Not in family
In family
NC State University
18
ASCOT -3 – Cluster Effort Variation
Where estimate falls
Data value popups
NC State University
19
ASCOT – 4 – 3D
SW Dev Cost = 3227 + 0.04273*Total_SC_Cost + 0.05347*(Num_of_Instr)
NC State University
20
Next Steps
ª Plan is to deliver ASCoT Rev 1 in December • Add Simple regression, COCOMO II, Nearest Neighbor• Improved clustering algorithm• Add LOC estimates and statistics• Add MRE performance statistics • Add data export feature
ª COMPACT, the NASA CubeSat cost model, will be delivered in the same web-based environment