Comparing MLE Vs. NLR in context of Software Reliability Growth Modes (SRGMs)
-
Upload
rakesh-rana -
Category
Engineering
-
view
63 -
download
3
Transcript of Comparing MLE Vs. NLR in context of Software Reliability Growth Modes (SRGMs)
Comparing between Maximum Likelihood Estimator
and Non-Linear Regression estimation procedures
for Software Reliability Growth Modelling
Rakesh Rana1, Miroslaw Staron1, Christian Berger1, Jörgen Hansson1,
Martin Nilsson2, Fredrik Törner2
1Computer Science and Engineering, Chalmers/ University of Gothenburg 2Volvo Cars Corporation
Software Reliability Growth Models (SRGMs)
• SRGMs are useful for assessing software reliability (quality), Information is useful for:
– Assessing the release readiness; and
– Testing resource allocation decisions
• Two of the widely known and recommended techniques for parameter estimation are Maximum Likelihood Estimation (MLE) and method of least squares (NLR)
• We compare between the two estimation procedures for their usability and applicability in context of SRGMs
Comparing between MLE & NLR
A better Metrics for measuring Predictive Accuracy
𝑃𝑅𝐸 =𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑 − 𝐴𝑐𝑡𝑢𝑎𝑙
𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑
𝐵𝑃𝑅𝐸 =𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑 − 𝐴𝑐𝑡𝑢𝑎𝑙
𝜂 ∗ 𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑 + (1 − 𝜂) 2 ∗ 𝐴𝑐𝑡𝑢𝑎 − 𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑,
𝑷𝒓𝒆𝒅𝒊𝒄𝒕𝒊𝒐𝒏 𝑨𝒄𝒄𝒖𝒓𝒂𝒄𝒚:
𝑤ℎ𝑒𝑟𝑒 𝜂 = 1 𝑖𝑓 𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑>𝐴𝑐𝑡𝑢𝑎𝑙
0 𝑖𝑓 𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑<𝐴𝑐𝑡𝑢𝑎𝑙
𝑀𝑆𝐸 = 1𝑘(𝑎𝑖 − 𝑝𝑖)
2
𝑘 − 𝑞
𝑮𝒐𝒐𝒅𝒏𝒆𝒔𝒔 − 𝒐𝒇 − 𝒇𝒊𝒕:
*PRE provides asymmetric value based on over or under prediction.
Thus we define Balanced Predictive Relative Error, BPRE
Comparing Parameters using MLE & NLR
Table: Comparing parameters with different estimators
Comparing between MLE & NLR
Comparing between MLE & NLR
Thank You
The research presented here is done under the VISEE project which is funded by Vinnova and Volvo Cars jointly under the FFI programme (VISEE, Project No: DIARIENR: 2011-04438).