On April 26th, Jhonatan Rivera Rivera presented his first scientific contribution as part of the SARAI project. In our work, we applied Machine Learning (ML) and Deep Learning (DL) algorithms to forecast time series of deformation (DTS) associated with subsidence due to groundwater extraction, using a database that contains a binary variable related to groundwater extraction in the elastic Madrid aquifer and previous values of DTS filtered according to Savitzky Golay, 1964.
The filter allows simple and interpretable ML algorithms, such as linear regression, to produce similar results to complex DL algorithms. The tuning of the algorithms indicates that the hyperparameters are similar when the DTS are similar, promoting the clustering of DTS in the optimization of modelling large databases like the one proposed in SARAI.