Project Overview
Predicting fault causes from fault reports. Using machine learning and natural language processing to make predictions about the solutions for faults using prior fault data.
Progress Timeline
-
Date: 04/10/2023
Inital project description report written, describing my inital thoughts on the project.
Download the report (PDF) -
Date: 10/10/2023
Created this website to track progress and keep relevant people updated.
-
Date: 11/10/2023
Wasn't sure where to start on the whole project, so as I'm quite unfamiliar with NLP and ML in general I thought I should probably review some literature and try learn some lessons and find out what's actually involved. This paper discussed how they used "Text Case-Based reasoning" to diagnose faults. While there was a lot of esoteric stuff about "rough set theory". There were a few lessons that could be learned about pre-processing. Wrote up these findings in overleaf.
-
Date: 14/10/2023
I found another paper which looked to predict fault types for textual bug reports. They took the approach of trying to classify text based bug reports into certain categories. Where the final ML classifiers would try to classify bug reports into these categories. This made it easier to compare the ML models with humans. Perhaps I could do the same thing with fault categorisation, but obviously with different fault categories based off of Estates data. Again, wrote my findings in more detail in overleaf.