Combination predication model based on SVR and its application

Aimed at the problems that single-production prediction models of oilfield development have low generalization ability and poor accuracy for median or long term prediction, a combination predication model based on support vector regression (SVR) is presented. The model is built based on small samples and has powerful generalization ability, which combines the applicable conditions and advantages of different single-prediction models, and is applied to predict oilfield production perfectly that obtains just a few experimental data. The structure designing and algorithm implementing of a combination predication model based on SVR is given, and is applied to process the data of oilfield actual production, which obtained accurate predication results and verified the effectiveness of the model and method.