Abstract
Summary This work focuses on a data-driven approach to speed up this process. A comprehensive literature review has been conducted, encompassing more than 200 research papers related to SI. After thorough examination, the data of over 400 SI tests have been collected, and machine learning (ML) models have been trained and tested on this data. The dataset of 400 data points is referred to as the combined one, which is further separated into two subgroups: secondary and tertiary SI datasets. Secondary imbibition refers to the SI experiment, where the fluid in the Amott cell was fixed through the experiment without a change in salinity. Tertiary imbibition relates to the tests, where the salinity of the fluid in the Amott cell has been changed several times ( Austad et al., 2010 ). Thus, the models have been trained and tested on three different datasets. Based on all the metrics combined, it was concluded that the GB model outperforms all other ML models. It was decided that R2 is the primary evaluation metric, and the R2 of the GB model was the largest.