The Use of Artificial Neural Network and Support Vector Classification for Recovery Factor Prediction

Li Dawei , Guangren Shi


Oil and gas recovery factor is a key parameter for oil industry, and it can be effectively predicted by appropriate data mining algorithms. In this paper, three regression algorithms and three classification algorithms have been applied to forecast recovery factor of oil and gas. The three regression algorithms are the support vector regression (SVR), the artificial neural network (ANN), and the multiple regression analysis (MRA), while the three classification algorithms are the support vector classification (SVC), the naïve Bayesian (NBAY), and the Bayesian successive discrimination (BAYSD). The purpose of this paper is to demonstrate how to select proper algorithms in three algorithms (SVR, ANN, MRA) for recovery factor regression and/or three algorithms (SVC, NBAY, BAYSD) for recovery factor classification. In general, when all these six algorithms are used to solve a real-world problem, they often produce different solution accuracies. Toward this issue, it has been proposed that a) when an algorithm is applied to a real-world problem, its solution accuracy is expressed with the total mean absolute relative residual for all samples, R(%), and b) result availability of a given algorithm application is applicable if R(%)<10, and inapplicable if R(%)≥10. A case study of recovery factor in 39 global oilfields has been used to validate the proposed approaches. This case study consists of two problems: regression and classification. For the regression problem, only ANN is applicable since its R(%) value is 5.89, whereas SVR and MRA are inapplicable because their R(%) values are 68.9 and 38.4, respectively. For the classification problem, only SVC is applicable since its R(%) value is 0, whereas NBAY and BAYSD are inapplicable because their R(%) values are 24.7 and 34.5, respectively. From this case study, it is concluded that the preferable algorithm is ANN for recovery factor regression, while the preferable algorithm is SVC for recovery factor classification.


Data Mining; Support Vector Regression; Artificial Neural Network; Multiple Regression Analysis; Support Vector Classification; Naïve Bayesian; Bayesian Successive Discrimination; Result Availability; Recovery Factor

Full Text:



  • There are currently no refbacks.