Based On The Amount Of Video Playback Sarima Model Prediction Software Design And Implementation

With the development and popularization of the computer network, network video services which based on a new generation of network technology, digital technology and information technology have a huge impact on the traditional media. Video sites with the rapid growth of network traffic and the number of users has become one of the main Internet applications. For the video site, the most important advertising sales model is CPM (Cost Per Mille, the fees of advertising a thousand times), and advertising inventory has a linear relationship with video view. It is very important for the video site to establish a suitable model of video view using theory and method of time series analysis and predict the future of video view. With the result we can estimate the amount of ad inventory in the future in order to make the right judgment of advertising sale.Under the guidance of time series relevant theory and methods, through statistically analyzing of a large number of video views, combined with the actual situation, the paper selected SARIMA model as forecasting model for the video views. Combined with Software design functional requirements and time series analysis theory of SARIMA model, the paper put foreword SARIMA modeling and forecasting processes and algorithms which Include pre-processing of data (mainly detection and handling anomaly data), SARIMA model identification, parameter estimates based on the generalized least squares and forecast data based on the principle of best prediction.This article choose the Java language for software development languages, the Eclipse as integrated development environment for Java, MongoDB as database, and call the R software for the relevant statistical calculations to design and achieve a automatic modeling and forecasting software for video views. The software includes the following functional modules:data acquisition module, data extraction module, the data conversion module, the forecasting module, and data storage module. Checking predicted values against actual values, analyzing forecasting results, and diagnosing the prediction model, the paper prove that take SARIMA model and take advantage of the proposed data processing, model identification, parameter estimation and prediction algorithms meet the forecast requirements.