网络信息过滤系统的过载处理方法研究

Research on Anti-overload Methods in Network Information Filtering System

针对网络信息过滤系统的过载问题，通过牺牲少量准确率和召回率来提高系统检测能力，并使用采样和规则泛化等近似处理技术保持系统在过载情况下的总体过滤能力。实验结果表明，在准确率或召回率仅降低10％～15％的情况下，能使系统可承受负载提高3倍，性能提升效果显著。

Due to the overload condition for the network information filtering system, this paper proposes the strategy that improves filtering speed paying a little loss of precision and recall, uses approximate methods such as text sampling and pattern generalizing. It can retain most of the filtering capability in overload condition. The experimental results show that the filtering capacity can be improved about three times while losing only less than 15% of precision or recall and has a remarkable performance promotion.