Building Advanced Predictive Models - Oil and Gas Case Study
Created:
| Last Modified:
This case study uses JMP and JMP Pro to find optimal geologic and completion parameters in upstream oil and gas processes (identifying, extracting and producing raw materials).
See how to:
- Understand the goal of the models - to determine point of diminishing return for using additional, expensive solid material (proppant) in the process
- Understand the response of interest - gross amount of oil and gas produced from a particular well over one year.
- Understand the study factors
- Controlled factors (20 completion parameters, amount of proppant, well perforation depth, # of completion stages, lateral well length)
- Pseudo-controlled factors (location parameters, county, latitude, longitude)
- Uncontrolled factors (23 geologic parameters, facies, reservoir thickness, porosity & permeability, (TOC) total organic carbon)
- Prepare data for analysis
- Handle missing values using imputation
- Use Predictor Screening to identify significant predictors out of all factors
- Rank all predictors using Bootstrap Forest
- Build model using Fit Model to rapidly develop simple to complex linear models using various fitting techniques, model parameters, and additional settings including random effects
- Construct Standard Least Squares model (JMP)
- Construct Stepwise Regression model (JMP)
- Construct Logistic Regression model (JMP)
- Use Generalized Regression (Pro)to create text, validation and training sets and then model correlated and high-dimensional data
- Use JMP Pro to fit an ensemble model by averaging many decision trees
- See how each split considers a random subset of the predictors
- Use Prediction Profiler to identify point of diminishing returns
