基于一阶马尔可夫链的实验数据序列分类模型

摘要

In order to analysis the sequence data generated by online experimental system, the first-order Markov chain is introduced. It artificially classifies the experimental data into the learning initiative and fraud, and constructs two Markov chain models. It determines by the larger possibility from which model the test data comes. At the end, it discusses the steady state distribution. Experimental results show that the model based on Markov chain has higher classification accuracy.%为了对在线实验系统产生的实验数据序列进行分析，引入一阶马尔可夫链。通过人工分类把实验数据分为学习积极和懒散作弊两类，分别构建马尔可夫链模型。根据输出概率判定测试数据来自哪一个模型的可能性较大。最后讨论了状态的平稳分布情况。实验结果表明，基于马尔可夫链的分类模型具有较高的正确率。