FRUGAL: Unlocking SSL for Software Analytics

Preprint:

FRUGAL paper

Case Studies:

[1] Static Warning Analysis: Learning to recognize actionable static code warnings (is intrinsically easy)

• Data: originated from Wang et al.'s ''Is there a "golden" feature set for static warning identification?: an experimental evaluation''

[2] Issue Close Time: When SIMPLE is better than complex: A case study on deep learning for predicting Bugzilla issue close time

• Data: originated and used by Lee et al. and Mani et al.

Methodology:

• CLA/CLAFI+ML JAVA implementation from Nam et. al's CLAMI: defect prediction on unlabeled datasets.
• FRUGAL: FRUGAL finds the best combination of unsupervised learners = {CLA, CLA+ML, CLAFI+ML} and 𝐶 = {5% to 95% increments by 5%}.

How To Run:

• ICT:

cd issue_close_time; python clami.py get_CLAGRID

• SWA:

cd intrinsic_dimension; python clami.py get_CLAGRID