@MASTERSTHESIS{pgi2020005,
	author = "M. Mihaylov",
	supervisor = "M. Roper",
	title = "Predicting the Resolution Time and Priority of Bug Reports: A Deep Learning Approach",
	school = "Department of Computer and Information Sciences, University of Strathclyde",
	year = "2019",
	abstract = "Debugging is a crucial component of the software development lifecycle. Most of the 
debugging done, once a system has been released, is based on bug reports. However, 
working on those bug reports could take developers a lot of time and resources and there 
are not a lot of available tools to help with that, which identifies a need to develop such 
tools and guidelines to analyze the information contained in bug reports. 
This paper introduces the usage of neural networks when performing bug report analysis. 
With the help of ensemble configurations of Multi-Layer Perceptron (MLP), Convolutional 
Neural Network (CNN) and Long Short-Term Memory (LSTM) and using sentiment and 
textual analysis, an experiment has been developed to predict the resolution time of bug 
reports and classify the priority. Additionally, the experiment analyses the importance of 
both textual and numerical factors when analysing bug reports. Based on the models 
developed, it is shown that using both textual and numerical data improves the results of 
standard text only neural networks. The obtained results show that determining the bug
fix time based on the information supplied in bug reports is achievable and also show 
success when trying to classify the priority of a bug report. 
Further evaluation and optimization of the best models, leads to the discovery of 
inequality in the distribution of the labels of the dataset, which leads to not very well 
trained models. With the use of another experiment an attempt has been made to 
improve the results by reducing the dataset, to contain similar counts of labels, which 
proves unsuccessful.",
}