基于朴素贝叶斯分类提出了一种复杂应用系统的性能预测方法.利用应用系统性能测试的结果作为训练集,引入朴素贝叶斯分类方法训练分类器,再将该分类器包装成预测模块嵌入应用系统,对响应时间等多种性能属性进行预测.与传统方法相比,该方法具有准确度高、构造简单、效率高、鲁棒性强、松耦合等优势.在针对金融报表系统的对比实验中准确率达到65%以上,训练过程的时间开销也明显少于传统方法.%This paper proposes a performance prediction method based on Naive Bayesian classifier for complex application system.In this method, a training set is collected using the result of performance test of application system.Naive Bayes method is introduced to train the classifier, and then the trained classifier is packaged to a prediction module and embedded into the system to predict various performance properties such as the response time, etc.Compared with traditional methods, our method shows a variety of superiorities, including high accuracy, simple structure, high efficiency, strong robustness and loose couple.A comparative experiment pertaining to financial report system shows that its accuracy rate achieves 65％ or higher, the time cost spent in training process is noticeably less than that of traditional methods.
展开▼
