Research On Strengthening Learning Model And Intervention Strategy In Distributed Pair Programming

Pair programming, one of the12practices of agile development methods, is a very simple and intuitive concept which can yield twice the result with half the effort. The current researches focus on theoretical analysis and experimental study about the effect of pair programming, as well as looking for the factors affecting the effect, but lacking quantitative analysis on dynamic mechanism and the intervention of pair programming process.Based on distributed pair programming (DPP for short), this thesis introduces the idea of the utility theory and the learning model, focuses on the utility computing, the prediction by the learning model and the intervention strategy built, and carries out the research along with analyzing quantitatively the communication process by the utility function, forecasting the trend of the communication behavior by the strengthening learning model, and building the intervention strategy. The main work is as follows:Firstly, quantifying the DPP process by the utility function. This thesis selects two quantitative indicators from the communication process, including the communication frequency and the code productivity, and constructs a suitable utility function for the communication process of DPP according to the relationship among the two indicators. Secondly, predicting the communication behavior trend of DPP by using the strengthening learning model. In fact, pair programming is a continuous learning process. Therefore, an analysis and prediction of DPP is made by introducing the strengthening learning model. Besides, a prediction model suitable for DPP is constructed through the comparative analysis of two typical learning models, namely BS (2000) and ARP. And then proposing an intervention strategy for the DPP process according to the conditions provided by the utility function and the learning model. This thesis analyzes the system environment, builds the intervention strategy, and provides the flow chart, which will provide guidance for the further realization of the intervention subsystem. Finally, carrying out simulation experiments. Specially, two experiments determining the utility function and the learning model are done. The former is mainly to determine the function type by function fitting, while the latter is mainly to select the learning model for predicting communication process of DPP through the parameter estimation and the evaluation of the model’s prediction ability.