不完全事务行为踪迹标记研究

Research on Tokenizing Behavior Footprints of Incomplete Transaction

研究不完全事务的行为踪迹标记问题,从软件交互行为本身出发,追踪分析事务产生的行为序列,为丢失标记的行为找到归属.本文提出了在分布式环境下事务处理过程满足SMP的系统模型,采用状态划分算法对随机排序的各事务进行"剥离";用偶图匹配方法对分割得到的偶图子系统依MLR分别进行匹配;最后将匹配的结果拼接起来形成完整的行为踪迹序列,提高了对监测信息的使用效率.通过仿真实验证实了本方法在标记不完全事务上的有效性和准确性.

The paper investigates the problem of tokenizing behavior footprints of incomplete transaction,starts from the software interactive behavior itself,tracks and analyses behavior sequences which generated by transactions,and finds the attribution for the behavior which lost token.We propose a system model in which transaction processing satisfied with the semi-Markov process under distributed environment,we use state division algorithm to strip transactions with random order,then,match the bipartite subsystems separately according to the maximum likelihood rule,finally,splice the matched results and form the complete transaction sequences,which improves the efficiency of the monitoring information.The simulation experiment has proved our method has both validity and accuracy in tokenizing incomplete transaction.