Economic Evaluation of Waterflood Using Regression and Classification Algorithms

Abstract

Three regression algorithms and three classification algorithms have been applied to forecast economics of waterflood. The three regression algorithms are the regression of support vector machine (R-SVM), the back-propagation neural network (BPNN), and the multiple regression analysis (MRA), while the three classification algorithms are the classification of support vector machine (C-SVM), the naïve Bayesian (NBAY), and the Bayesian successive discrimination (BAYSD). In general, when all these six algorithms are used to solve a real-world problem, they often produce different solution accuracies. In this paper, the solution accuracy is expressed with the total mean absolute relative residual for all samples, R(%), and it is proposed that an algorithm is applicable if R(%) ≤ 10. A case study at the Nebraska Panhandle has been used to validate the proposed approach. This case study consists of two problems: regression and classification. The only difference between these two problems is the predicted variable in regression problem is real number, while the predicted variable in classification problem is integer number. And the integer number is determined from the real number by using proposed convertion rules. For the regression problem, R-SVM, BPNN and MRA are inapplicable because their R(%) values are 140, 51 and 293, respectively. For the classification problem, however, C-SVM, NBAY and BAYSD are all applicable since their R(%) values are all 0. From the case study, it is concluded that: a) For classification problems, the preferable algorithm is C-SVM, NBAY, or BAYSD, and BAYSD can also serve as a promising dimension-reduction tool; b) for regression problems, the preferable algorithm is BPNN, but MRA can serve as a promising dimension-reduction tool only when the studied problems are linear; c) if BPNN is inapplicable for a regression problem because its R(%) > 10, it is proposed to change this problem from regression to classification by reasonable conversion rules, then apply C-SVM, NBAY, or BAYSD; and d) comparing with C-SVM, BAYSD is conditionally better than C-SVM.

Key words: Regression; Classification; Solution accuracy; Conversion rules; Dimensionality reduction; Nebraska Panhandle