Oil-field-performance

#Random forest codes for estimating recovery factor, maximum oil rate and average depletion rate #Load and Pre-process Data load data

#The code below will convert loaded numerical variables that are categorical into categorical variable. Diagenesis = categorical(Diagenesis); Paleoclimate = categorical(Paleoclimate); GrossDepEnvironment = categorical(GrossDepEnvironment); ProductionStrategy = categorical(ProductionStrategy); StratigraphicHeterogeneity = categorical(StratigraphicHeterogeneity); TrapType = categorical(Traptype); StructuralComplexity = categorical(StructuralComplexity);

#Make a table X containing all the training data using 32 variables X = table(GDE,Depth,averageporosity,averagepermeability,initial pressure,structuralcomplexity,Netgrossratio... Averagedepletionrate,RecoveryFactor);

#Determine the number of Levels in Predictors countLevels = @(x)numel(categories(categorical(x))); numLevels = varfun(countLevels,X(:,1:end-1),'OutputFormat','uniform');

#Grow Robust Random Forest #For the three models we had Mdl1, Mdl2 and Mdl3, with X, Y and Z tables corresponding to recovery factor, maximum reservoir rate and depletion rate models respectively. t = templateTree('NumVariablesToSample','all',... 'PredictorSelection','interaction-curvature','Surrogate','on'); Mdl = fitrensemble(X,'RecoveryFactor','Method','bag','NumLearningCycles',500,... 'Learners',t);

#Estimate the model R2 using out-of-bag predictions. yHat = oobPredict(Mdl); R2 = corr(Mdl.Y,yHat)^2

#Predictor Importance Estimation impOOB1 = oobPermutedPredictorImportance(Mdl1); impOOB2 = oobPermutedPredictorImportance(Mdl2); impOOB3 = oobPermutedPredictorImportance(Mdl3);

#Make prediction on new data

V below is the test dataset

YHat = predict(Mdl,X)