Classification Of Heavy Oil Carbon Dioxide Oil Huff And Puff Effect Based On Adaboost Algorithm

Horizontal well carbon dioxide huff and puff technology is an effective means to improve the development effect of heavy oil reservoirs.However,the traditional carbon dioxide huff and puff well selection and effect evaluation methods need strong professional knowledge,and the modeling is complex and time-consuming.This limits the scale and effectiveness of the technology in the field.In view of the above problems,this paper proposes a new idea of using machine learning algorithm to construct evaluation method.By optimizing the feature selection and classification algorithms suitable for carbon dioxide huff and puff dataset,an improved Adaboost algorithm is proposed,and the superiority of the new algorithm is verified by application.(1)Carbon dioxide huff and puff dataset has the characteristics of high-dimensional small samples,nonlinearity and non-deletable outliers.The most suitable feature selection algorithm is random forest method.This is because decision trees in random forests are parallel and insensitive to outliers.The unbiased estimation model and the random selection of all features balance the generalization ability and classification accuracy of the algorithm.(2)In view of the shortcomings of the feature dimension that is easy to over-fit and the infinite increase of the abnormal point sample weight in Adaboost ensemble algorithm,an improved Adaboost algorithm is established by combining feature selection and abnormal value scaling.The accuracy、precision、recall rate and F1-score of the improved Adaboost algorithm are 0.96、0.97、0.97 and 0.97.Respectively,which are higher than those of the single support vector machine and the traditional Adaboost algorithm.The improved Adaboost algorithm effectively improves the defects of the high dimension of the dataset and the infinite amplification of the noise points.The research results establish an improved Adaboost algorithm for high dimensional small sample datasets of petroleum engineering,which is of great significance to data mining and intelligent decision-making in petroleum engineering.