The effect of locality based learning on software defect prediction

Abstract

The Effect of Locality Based Learning on Software Defect Prediction Bryan Lemon Software defect prediction poses many problems during classification. A common solution used to improve software defect prediction is to train on similar, or local, data to the testing data. Prior work [12, 64] shows that locality improves the performance of classifiers. This approach has been commonly applied to the field of software defect prediction. In this thesis, we compare the performance of many classifiers, both locality based and non-locality based. We propose a novel classifier called Clump, with the goals of improving classification while providing an explanation as to how the decisions were reached. We also explore the effects of standard clustering and relevancy filtering algorithms. Through experimentation, we show that locality does not improve classification performance when applied to software defect prediction. The performance of the algorithms is impacted more by the datasets used than by the algorithmic choices made. More research is needed to explore locality based learning and the impact of the datasets chosen. I would like to thank Dr. Menzies for introducing me to Data Mining, it was this initial class that set the direction for my education at West Virginia University. You introduced me to the world of research. You taught me the difference between creating something new and doing research. The classes I took from you went beyond teaching the subject matter; you taught not just the "How?", but the "Why?". It was a privilege to be one of your students and one of your research assistants. You not only taught me many things, but you changed how I look at Computer Science. I am thankful to the Lane Department of Computer Science and Electrical Engineering, West Virginia University, and the professors for the education I have received. The skills I have learned here will help me as I further my education, and beyond into the work force. I would like to thank Dr. VanScoy for the classes I took from her. Each lecture was not just informative, but entertaining. Like all the professors I have had the privilege to take courses from here at WVU, the professor did not just "end" at the door to the classroom. I would like to thank my mother for the kind, loving support she has give me throughout my education. She provided the foundation that my entire education is based on. Without her, I would not be where I am today.

HbA1c Reduction in Diabetic Older Blacks and Hispanics: A Study on Mobile Physical Activity Tracking

“…For statistical inference, paired t -test, a nonparametric method (Page test for trend), and one-way fixed effects ANOVA were used when appropriate. All the analyses were conducted in R (24,25).…”

A Work-Pattern Centric Approach to Building a Personal Knowledge Advantage Machine

“…Pseudo code for C4.5[20] function K-Means(data) Centroids = A random subset of K instances from data for each row in data m(row) = the cluster closest to row while m has changed for each centroid in Centroids centroid = the centroid of the current instances assigned to that centroid for each row in data m(row) = the cluster closes to row return Centroids Pseudo code for K-Means[18] …”