传统的传输控制协议(TCP)中的重传超时值(RTO)设置过于保守,且RTO预测算法对于往返时延(RTT)变化的响应过于缓慢,造成长时间的无谓等待,降低了带宽利用率。为了解决这些问题,提出一种基于最小均方(LMS)滤波的预测算法。为了验证新算法的性能,选择了大量有代表性的网址进行试验,对新旧2种算法的预测结果进行比较,并做了性能分析。试验结果证明,所提出的算法相比于原算法,具有能很快达到稳定,稳定后的预测误差小,对RTT变化的响应快等特点,提高了传输性能。 The retransmission timeout value (RTO) setting in the traditional Transmission Control Protocol (TCP) is too conservative, and the RTO prediction algorithm responds too slowly to the RTT change, resulting in prolonged unnecessary waiting and reducing the bandwidth utilization . In order to solve these problems, a prediction algorithm based on least mean square (LMS) filtering is proposed. In order to verify the performance of the new algorithm, a large number of representative Web sites were chosen for testing. The prediction results of the two old and new algorithms were compared and their performance was analyzed. The experimental results show that compared with the original algorithm, the proposed algorithm has the characteristics of fast stability, stable prediction error, fast response to RTT changes and improved transmission performance.