Research Of Box Office Forecasting Based On Rough Set And Support Vector Machine

Box office forecasting problem has recently become a focus of the western film industry. Among the numerous results gained from this research, it is evident that scientific theories and guidances are insufficient in the industrialization of china's film industry with regard to a complextion that the investigation of the film industry in china is still rough and original. How to conduct the scientific research in the film industry and how to make box office forecasting a real powerful tool for the analysis and forecasting in this field both need to be taken into serious consideration.The current research of box office forecasting begins at 80th of the 20 century, it is a comprehensive method combined by many subjects such as cinema, transmition, economics and management.This method analyses the factors related to the success of the box office (for example the originality, publication, sales quality and so on) and builds the mathematical model in order to achieve the goal of box office forecasting.Support vector machine (SVM) is a highly effective classifying recognition method developed on the foundation of statistical learning theory (SLT) in the middle of 90's. SVM is powerful for the problem characterized by small sample, nonlinearity, high dimension and local minima. It has been widely applied into many areas, such as pattern recognition, signal processing and data mining. Nowadays, SVM has become more and more popular in the machine learning field, but when processeing the bigger data set, it will meet the problems that the processing speed becomes slow and the training time turns to be excessively long. These problems will affect the classifying performance. For this, Rough Set (RS) method is introduced. The rough set theory is a mathematical approach to statistical analysis on incomplete, inaccurate and inconsistent information. In order to improve the classification capacity it can process large mounts of high-dimensioned data by attribute reduction without any prior information as the classification capability remains unchanged.In this paper, a SVM classifier is designed by combining the data processing function of rough set and the principle of support vector machine algorithm. This model employs rough set as the pre-processor and optimizes the input variables by attribute reduction in order to improve the capacity of classifier. Apart from that, regarding to the situation that there exists a lack of scientific and theoretic guidance of domestic box office forecasting, the model is applied into box office forecasting. During this course, the factors that make contributions to the box office are selected to be the input variables while the categories of the box office incomes to be the output variables. The results gained from the data processing and data testing course prove that the classifier solves the box office forecasting problem effectively and the method is superior to the multilayer perceptron classifier conducted by Ramesh S. and Dursun D. The accuracy is less than 10% which well meets the industry requirement. It shows great classification ability.