Applications of information theory, genetic algorithms, and neural models to predict oil flow

摘要: This work introduces a new information-theoretic methodology for choosing variables and their time lags in prediction setting, particularly when neural networks are used non-linear modeling. The first contribution of this is the Cross Entropy Function (XEF) proposed to select input order compose vector black-box models. XEF method more appropriate than usually applied Correlation (XCF) relationship among output signals comes from dynamic system. second that minimizes Joint Conditional (JCE) between by means Genetic Algorithm (GA). aim take into account dependence selecting most set inputs problem. In short, theses methods can be assist selection training data have necessary information predict target data. petroleum engineering problem; predicting oil production. Experimental results obtained with real-world dataset presented demonstrating feasibility effectiveness method.