Analysis of Programming Performance Based on 2-grams of Keystrokes and Mouse Operations

Abstract

Individual differences in programming performance are very large. For objectivity in programming performance, there is a strong need to quantitatively evaluate programming performance of individual developer. In this study, we conduct an experiment to measure programmers’ behaviors including keystrokes and mouse operations of fourteen subjects, and analyze correlations between them and programming performance. The experimental results show that those who check a lot of program operation and those who use shortcut keys frequently had higher programming performance. In addition, subjects who spent a lot of time searching on the web, compiled a lot, and rewrote a lot of code had lower programming performance.