Variable recommendation method for log records

The invention discloses a variable recommendation method for log records, which comprises the following steps: firstly, extracting graph structure information by using a graph neural network, and then fusing the graph structure information and utilizing semantic information extracted by a pre-training model to recommend log variables. Aiming at the problems that at present, research work for recommending variables in log record statements is less, the effect does not reach an ideal level, and a developer is lack of being helped to compile high-quality log record statements, the method can directly help the developer to compile the high-quality log record statements; the technical problem that log variables recorded in source codes are unreasonable is solved. 本发明公开了一种针对日志记录的变量推荐方法，首先使用图神经网络提取图结构信息，然后融合图结构信息和利用预训练模型提取的语义信息去推荐日志变量。针对目前对日志记录语句中的变量进行推荐的研究工作既少且效果未达到理想水平，并且缺少帮助开发人员编写高质量日志记录语句的问题，本方法能直接帮助开发人员编写高质量日志记录语句，以解决源代码中记录的日志变量不合理的技术问题。