Abstract:

Predicting the number of online audio-visual users can provide valuable information to help manufacturers get more profits. Based on time series analysis, support vector regression is used to make accurate prediction with adjusted feature. The time series is first modeled and predicted, a linear regression model used to make further improvement, and then, by combining time and real-life characteristics, adding a new feature. Samples of the new feature are trained with support vector regression. Optimal parameters of the radial basis function are sought using the social cognitive optimization. A good prediction result can be obtained using the proposed method.

Key words: auto-regressive and moving average (ARMA), linear regression, support vector regression