For The Telecommunications Industry, Adaptive Intelligent Traffic Forecasting Model And Realization

Along with the prosperity of Chinese economy, the telecommunication industry has got many extraordinary achievements in our country. Each telecommunication corporation has accumulated large amounts of traffic data. Traffic data also has the nature of time series. So we can use the basic theories and the methods of time series to analyze these historical traffic data. Then find out the rules of their changing, and establish some reasonable traffic forecast models to implement auto-adapted traffic forecast for the conditions of common days and festivals. Accurate and reasonable forecast results can be as the basic of decision-making for each telecommunication corporation's network maintenance and market programming ,etc. Therefore traffic forecast has great of significance.This dissertation mainly studies the telecommunication industry's auto-adapted intelligence traffic forecast model and its realization, its main contents include: basic theories of time series and common forecast models, the cycle analysis of traffic time series, traffic forecast model research for the condition of common days, traffic forecast model research for the condition of festivals, the realization of traffic forecast model and so on. The dissertation is a comprehensive summing-up to author's research work.. Its key contents include as follows:(1) Contrasting and analyzing several kinds of common time series forecast models and their applicationThe correlation theories of time series modeling are the academic basic which this dissertation studies. For the question of traffic forecast modeling, the author has contrasted and analyzed several kinds of common time series forecast models, and according to the common criterions of time series models'choice, has made detailed analysis and explanations for their concrete applications.(2) Putting forward a kind of traffic time series cycle analysis algorithm ----- a kind of time series preferable cycle searching algorithmThe author has studied a kind of traffic time series cycle analysis algorithm, it is a kind of time series preferable cycle searching algorithm. This algorithm has overcome some shortcomings of the tradition time series cycle analysis methods, it is also the academic basic of traffic forcast algorithm studied. It has utilized some statistics knowledge,least squares method,and some other mathematics knowledge. It introduces the thought of set division. By reasonable grouping, it can make the overall mean-square deviations which each group of swatchs reach to the smallest, and find out the cycle of traffic time series. Those experiments have proved that this algorithm is easy to implement, its searching result comparing with the actual situation is proper, and the effect is also very good.(3) Putting forward a kind of traffic forecast algorithm for the condition of common days ------- a kind of linear minimum variance auto-adapted traffic forecast algorithmUsing tradition Box-Jenkins (B-J) time series forecast modeling algorithm, carrying on forecast modeling processing. Through detailed analysis and experiments explained that the traditional B-J time series forecast algorithm is not suited for directly using in traffic time series forecast modeling. So the author has studied a kind of traffic forecast algorithm using in the condition of common days, namely a kind of linear minimum variance auto-adapted traffic forecast algorithm. For the nature that traffic time series changes very large when time changes, when doing the estimate of parameter, this forecast algorithm uses the method of least square with forgetting factor, and so on . Through a series of simulative experiment, it is proved that the forecast effect of this algorithm is good in conditions of common days'forcast.(4)Putting forward a kind of festival traffic forecast algorithmThe author has studied a kind of festival auto-adapted traffic forecast algorithm. This algorithm uses the analysis methods of time series level and rate, estimate of parameter by forgetting factor, and method of set division, etc. When use this algorithm to do festival traffic forcast ? It is easy to find out that there are many advantages, for example: adaptitude very strong, high correlation and good forcast effect. Contrasting traffic forcast simulative experiment has explained that this forecast algorithm is proper to ues in festival traffic forcast.(5)Putting forward the method of forcast algorithm designed as softwareThe author has discussed the basic principle of forcast methods designed as software, and put forward the working flow of the forcast software. Also have designed a basic framework of intellective forcast system for Sichuan Mobile Communication Limited Corporation. Finally, the traffic forcast software is designed by the auto-adapted traffic forecast algorithm. Now the forcast software is running in Sichuan Mobile Communication Limited Corporation, and has gotten good effect.