Research On Association Rule-based Software Development Recommendation Technology

The demand for software is increasing dramatically and the requirement for thesocialized production is rising continuously, which makes the requirement ofprogrammers for the performance of development tools become higher and higher.Intelligence is one of the research focuses of the software development toolsperformance. Nowadays, software development tools use intellisense technologycommonly to aid users to develop software. However, the role played by this technologyis limited to input auto-completion, which cannot provide more intelligent features. Forexample, it cannot predict which function user will need in the next step. Personalizedrecommendation technology can forecast demand well, which is mainly used in the fieldof Internet but not applied to software development yet.The way that e-commerce sites use the personalized recommendation technologyto predict goods for users is studied in this paper, and the idea of applyingrecommendation technology to software development is proposed here. Those functionsor other resources which supplied by software development tools can be treated asvarious goods, and developing software is a process that developers buy those goods.The software development recommendation system is a system that usesrecommendation technology to assist developers with their developing. By learningcurrent recommendation technologies and taking the characteristics of the softwaredevelopment into account, this paper presents a thorough study of association rulesapplied in the software development recommendation systems. To avoid defects ofApriori algorithm in mining association rules, we propose an improved Apriorialgorithm that improved it from three respects, which are optimizing the transactiondatabase, introducing the recommended result feedback factor, and limiting someparameters for different application. The model for association rules in the softwaredevelopment is set based on study. We also propose a computing strategy with weightedassociation rules for the reason that common code and proprietary code have differenteffects on users.An intelligent software development recommendation system for LabVIEW isdesigned and implemented based on the improved Apriori algorithm and the computingstrategy with weighted association rules. Test results shows that the recommendationsystem can predict the requirement developers needed in the next step well. The systemgives consideration to both accuracy and personalized features, and greatly increases thedevelopment efficiency of LabVIEW engineer. What is more, it resolves the problem that LabVIEW development tool is not intelligent enough. Meanwhile, it is also provedthat this recommendation technology has high accuracy and good practicality.In further research, to compute faster and get recommended result more accuracy,the association rule mining algorithm and recommended result computing method willbe continuously researched, and will better assist software developers to developsoftware.