一种动态调度的延迟敏感流网络挖掘算法

Dynamic scheduling latency-sensitive stream network mining algorithms

为了提高延迟敏感数据流的挖掘精度及能量效率,提出一种动态调度的延迟敏感流网络挖掘算法。该算法建立了流挖掘系统模型,对分类器链的选择概率、能量消耗和延迟敏感进行分析。为了控制挖掘系统的延迟时间并节省能量,提出了基于延迟约束的能量最小化组合方程。同时,采用了一个有效的分解定界算法来解决分类器的最佳处理速度选择问题,通过贪婪算法找到组合方程的最小能量边界,实现流挖掘系统在具有更高的分类效率的同时保持较低的能量消耗和延迟。仿真结果表明,该算法相比基于动态时间规整的数据挖掘算法和基于遗传算法优化的数据挖掘算法,能量效率分别提高了39.4%和41.4%,分类精度分别高出11.5%和5.9%,具有更好的节能效果和挖掘精度。

In order to improve the mining accuracy and energy efficiency of delay-sensitive data, a dynamic schedulinglatency-sensitive stream network mining algorithm is proposed. First, the algorithm establishes stream mining systemmodel, and analyses choose probability, energy consumption and latency-sensitive of classifier chain. Then, in order tocontrol the delay time mining system and save energy, constraint equation based on energy minimization combinationdelay is proposed. At the same time, it uses an effective decomposition bound algorithm to solve the optimal processingspeed of the classifier selection problem, and finally finds a combination of minimum energy equation boundary bygreedy algorithm. Stream mining system maintains at a higher efficiency with the classification low energy consumptionand latency. Simulation results show that compared with algorithm based on dynamic time warping algorithm and datamining algorithm based on genetic algorithm optimization of data, energy efficiency is increased by 39.4% and 41.4%,classification accuracy is increased by 11.5% and 5.9%, respectively, with better energy efficiency and excavation accuracy.