### Systematically Generated Formulas for Spectrum-Based Fault
Localization

*Qusay Idrees Sarhan**,
Tamás Gergely and*
*Árpád Beszédes*
The basic element in Spectrum-Based Fault Localization
(SBFL) are the risk evaluation formulas, which calculate a
suspiciousness score for each program element based on test
coverage and test case outcome information. This score can be used
in debugging to identify the faulty element more efficiently. A
large number of manually crafted formulas have been proposed, but
a line of research tries to generate formulas
(semi-)automatically. Some of these approaches are based on
heuristic search (e.g., genetic algorithms), and researchers
started only recently examining systematic ways to generate
all possible formulas corresponding to a particular class of
formula structures. In a recent work, we explored a very simple
formula template as a proof of concept but this research failed to
find a new formula that outperformed already published ones. In
this paper, we take a next step and investigate a class of formula
templates that are more complex but still feasible to explore
fully. Many of the generated formulas cover some well-known
existing ones, but we also managed to identify two new ones that
are not found in literature and are better than most of the
previously published formulas (evaluated on the Defects4J
dataset).

Keywords:
Spectrum-Based Fault Localization, debugging, suspiciousness score
formulas, systematic search.

Back