多源头网络用户访问信息自适应识别算法

为了解决传统算法学习规则有效性低、无法保证学习性能、匹配模板不全面、容易出现误识别现象的问题,提出一种改进的反向传播(back propagation,BP)神经网络算法研究多源头网络用户访问信息自适应识别问题。对多源头网络用户访问信息进行数据清洗处理,用多源头网络用户访问矩阵对全部会话集合进行描述;在矩阵中引入网络用户位置信息,将得到的信息保存至数据库,构成信息集。将一段时间内用户访问日志构成用户访问路径数据,依据访问请求抵达顺序,将其保存至相应用户缓冲区。把多源头网络用户访问路径当成隐马尔科夫模型的状态转移序列,将网页中信息集当成状态输出符号集,通过离散隐马尔科夫模型对不同源头网络用户访问信息进行分析,提取其特征。将多源头网络用户访问不同种类信息的概率特征作为输入,建立改进BP神经网络算法,得到的输出结果即为多源头网络用户访问信息自适应识别结果。结果表明:采用的BP神经网络算法学习性能优;所提算法识别准确性高。可见所提算法识别结果可靠。

Research on adaptive identification algorithm of user access information in multi source network

In order to solve the problems of low efficiency of learning rules, insufficient learning performance, incomplete matching templates and easy to misidentify in traditional algorithms, an improved BP neural network algorithm is proposed to study the self-adaptive identification of multi-source network users''access information. The multi-source network user access information is cleaned by data. The multi-source network user access matrix is used to describe all session sets. The location information of network user is introduced into the matrix and saved to the database to form the information set. A period of time the user access log user access path to the data access request, based on the arrival order, save it to the corresponding user buffer. The multi-source network user access path is regarded as the state transition sequence of hidden Markov model, and the information set in the web page is regarded as the state output symbol set. The Discrete Hidden Markov model is used to analyze the access information of different source network users and extract their characteristics. Taking the probabilistic characteristics of multi-source network users''access to different kinds of information as input, an improved BP neural network algorithm is established. The output result is the self-adaptive recognition result of multi-source network users'' access information. The results show that the BP neural network algorithm has excellent learning performance and the algorithm has high recognition accuracy. The result shows that the algorithm is reliable.