一种半监督学习的代码自动生成性能评估方法

为了提高软件开发的质量和效率,代码自动生成是当前的研究热点,代码自动生成的性能是其中的重要问题.现有代码自动生成的性能分析方法较简单,难以评估代码自动生成过程中程序员与代码自动生成工具各自的特征.本文综合考虑了代码自动生成过程中程序员与代码自动生成工具的作用,提出了一种基于半监督学习的代码自动生成性能评估方法,通过抽取程序员行为与代码自动生成工具行为的重要特征,划分代码自动生成的性能类别,建立了基于深度神经网络的代码自动生成过程性能评估模型,并计算程序员行为特征与代码自动生成工具行为特征对性能的影响程度.实验结果表明,该方法可以有效分析程序员行为与代码自动生成工具行为对代码自动生成过程性能的影响.

Performance Evaluation Method of Code Generation Based on Semi-supervised Learning

In order to improve the quality and efficiency of software development,automatic code generation is the current research hotspot.The performance of automatic code generation is an important issue.The existing performance analysis methods of automatic code generation are relatively simple,so it is difficult to evaluate the characteristics of programmers and automatic code generation tools in the process of automatic code generation.In this paper,the function of the programmer and the automatic code generation tool in the process of automatic code generation is considered.A method of evaluating the performance of automatic code generation based on semi-supervised learning is proposed.By extracting the important characteristics from the behavior of the programmer and the automatic code generation tool,the performance category of automatic code generation is divided.Then the performance evaluation model of automatic code generation process based on Deep Neural Networks is established.Finally,the impact on performance produced by behavior both programmers and automatic code generation tools is calculated.Experimental results show that this method can effectively analyze the impact on the performance from programmer behavior and automatic code generation tool behavior during the process of code generation.