Research On Prediction Of Multi-source Data Based On Time And Space

With the rapid development of the Internet, people’s life and work have become more convenient than before. When individuals are surfing on the Internet, multitudes of network traffic data are generated. Even though most of these data are not harmful to the channel, some data, which are hidden in network traffic, may cause significant threats towards the networks and the hosts on the Internet, such as interactive information. Thus, these problems have attracted a lot of concerns from the corporations and research institutions on the traffic data. Some researchers did a lot of contributions on the network security, including: Some analyzed the potential threats from the flow of data are analyzed and proposed some various detection algorithms; Some discussed the characteristics of traffic data and studied network security, such as the characteristics of network security by means of data analysis techniques such as firewalls, etc.; Others deployed the security measures in advance and improve existing security technologies according to the variation trend predicted by the traffic data. As a result, how to collect, store and analyze the traffic data efficiently becomes a critical problem, which we are going to tackle with.In this thesis, we verify and analyze the multi-source data in the network environment, especially focus on the characteristics of different types of traffic data. Then, we present a comprehensive algorithm to predict the tendency of the data. Finally, to collect the data that are used to analyze, we propose a multi-source data collection and analysis framework. The contributions of this thesis can be shown as follows:First, we analyse the research status of data prediction algorithm and data collection methods. All these include the research status of data prediction based on time series and the methods of the network traffic data collection.Second, we study the characteristics of the multi-source data in the network. Based on the wired traffic data and wireless traffic data, we get the traffic data’s features, such as the length of traffic and the self-similar characteristics.Then we conclude the research of the multi-source data characteristics.Third, the thesis discusses the multi-source data trend prediction model. In this part, FARIMA model is firstly anaylsed. Then based on the model and combined the wavelet transform and time series model, we propose a new algorithm which is used to predict the trend of multi-source data. Depending on some experiments with network traffic on the forcast algorithm, these results of the analysis indicate that the prediction algorithm can improve the accuracy of prediction and also can effectively reduce some calculation.Finally, we propose a collection and analysis framework for collecting and analyzing the multi-source data. Based on Nagios which is a tool to monitor the network, and combined with the data trend prediction algorithm, the framework takes use of the “ative+passive” model so as to collect and save different data from the network. In fact, this framework is used to monitor the network and collect data from different network. After a large number of tests on the framework, we conclud that: with the trend prediction algorithm in the framewok, it can complete a great amount of monitoring tasks and would protect our network well.