Converting Dates for Use in Model to Predict Oil & Gas Drill Timing?

MissWickes

Hi There! I recently completed MIT's "No-Code AI & Machine Learning" certificate where we used RapidMiner on some more simple models. I'm now trying to put what I've learned into use in my business. I am trying to use certain "key drilling indicators" to predict when a drilling rig will show up on a given tract of land. The attributes (aka key drilling indicators) would include nominal attributes like "operator" aka company and "state." Then we will also have date (numeric) attributes like "Permit Date," "Leasing Date," as well as relevant dates. The label ID will be "Drilling Date." I have a spreadsheet with a few hundred rows of historical information and dates on wells that have been drilled. The goal is to find correlations among the Key Drilling Indicators and be able to predict future development timing on new wells, based on the KDIs at hand (i.e. use the information we receive to predict the future drilling date - could just be a month - for a given undrilled well location). I'm not sure how to create a predictive model using these dates. Any help would be much appreciated! I would also be willing to pay for a "tutor" to help me build the model. 

BalazsBarany

Hi!

Dates can be expressed as numbers, e. g. days after an epoch. The operators Date to Numerical and Numerical to Date help you convert between these representations.

Aside from this detail your use case is a simple regression task. You can try linear regression, Generalized Linear Model, Neural Network, Support Vector Machine, even decision trees for it. Most of these will expect you to recode the nominal attributes to numeric indicators (try Nominal to Numerical with dummy coding for that).

Creating a regression model e. g. with Linear Regression, GLM or Decision Tree will help you understand the correlations between the attributes and the label.

Regards,
Balázs

MissWickes

Thank you! I will give that a shot.