一种改进的隐马尔科夫模型对话务量的预测

An Improved HMM models for traffic Load prediction

为了改进隐马尔科夫(HMM)和模糊集(Fuzzy)综合模型(HMM-Fuzzy模型)存在的过学习的问题,提出了将人工神经网络算法(ANN)与HMM-Fuzzy模型相结合的算法,ANN算法具有很强的抗干扰性,也不存在过学习的问题,刚好弥补了HMM-Fuzzy模型的缺陷,并提高了隐马尔科夫的识别能力。并将其运用到无线移动通信话务量的预测中,而且首次将短信数据,用户数据等影响话务量的相关因素考虑进去,结果表明:ANN-HMM-Fuzzy模型在综合考虑多种因素的情况下具有预测精度高,耗时少的特点。

In order to improve the learning problem of HMM - Fuzzy models ANN algorithm combined with HMM - Fuzzy model is presented.. Because ANN algorithm has a strong anti - interference and has not any learning problems, just make up for the deficiencies of the HMM- Fuzzy model, And improves the HMM＇s ability to recognize. It is applied to predict traffic load. And first put SMS data,user data and other related factors affecting traffic into account. Results showed that the ANN- HMM - Fuzzy model in the case of considering a variety of factors have higher predictive accuracy, less time- consuming characteristics.