Depth Sliding Windows Application on Geophysical Well Log Data

Stochastic processes are the point of convergence of different types of random data that depend on a continuous parameter, be it time, distance or any type of continuous measurement. This has allowed tests to be carried out over the years using methods from one discipline in others, for example, time series analysis can also be applied to well log data, since both represent a nonstationary stochastic process. The sliding window method has been widely used for forecasting time series. The present research develops a proposal for the application of this method to well log data, in order to obtain an adequate data structure to feed supervised learning models. The results showed that it is feasible to sample windows that predict a certain amount of values given a depth interval. Compared to the resampling method, sliding windows provide a better fit for predictions in automated models.