Exploitation And Description Of Activity Patterns In Software Development

Software process is important for the development of software projects, and it is a economical and reliable method to mine the process from software project. In the era of big data, growing event data in software development contains knowledge, and these knowledge is essential to discover, understand and improve the software process. At present, software development contains a lot of these data, but it is difficult to generate real value. Due to the characteristics of event data, i.e., many types of data, data volume and data which is difficult to be used directly, leading companies do not pay attention to these data. Moreover, the software project development event data has the characteristics like many event record, activities are not clearly demarcated, and single instance. How to mine useful knowledge from these data to guide software development, the issue was placed in front of us.In this paper, mining activity patterns (recurring development tasks) from software development, to better understand the behavior of the various stages of software development. First of all, The method of using folding patterns mine software process activities. As research data, event logs generated from Subversion which is type format version control system, then, classifying the project path, getting the concerned activities, and identifying the atomic activity of event log. Afterwards, for the repeated testing, modifying, improving behavior occurred in the software projects, introducing the tandem repeats and using edit distance analyze activities fragment with a certain degree of similarity to obtain foldable activity fragments. Then, analyzing and folding foldable segments for analysis and folded to obtain activity patterns with a variety of control structures (sequence, selection, parallel). So as to achieve the object of obtaining activity patterns from the single instance project log. By identifying the equivalence classes of activity patterns, improve the effect of folding. Finally, using petri nets to describe activity patterns and get the software development process model.In this paper, the method of mining activity pattern makes that the single instance of the software project log data can also be mined out of the software development process model. Experimental results show that for tandem repeat sequences of activities, the mining method mining results are easy to understand, can get the debugging, modifying and improving behavior in the software project. These make the research of this dissertation has great theoretic and practical meaning.