Algorithm 1: Methodology proposed to predict student retention/dropout in HE institutions similar to the Chilean case.

1:Collect and input data about the PSU scores (or equivalent depending on the country), CP index, economic quintiles, and first/second year scores. 2:Standardize attribute names in the input data. 3:Perform cleanup and removal of missing/noisy/duplicate instances in input data. 4:Eliminate data features that do not add value according to [16]. 5:Create a data subset using the PSU scores, CP index, and economic quintiles. 6:Extract data of students with more than three years of academic follow-up. 7:State classes for data as two values considering actives and dropouts. 8:Arrange data for the global model from the full data set considering dropout at any level. 9:Assemble data for the first level model from the full data set considering dropout in the first year only. 10:Dispose data for the second-level model from the first-level data with the first-level retained students plus the first-year scores. 11:Organize data for the third-level model from the second-level data with the second-level retained students plus the second-year scores. 12:Apply ML algorithms to the global model with the data of Step 8, and analyze the results. 13:Employ ML algorithms for the first-level model with the data of Step 9, and state the results. 14:Use ML algorithms for the second-level model with the data of Step 10, and obtain the results. 15:Utilize ML algorithms for third-level model with the data of Step 11, and indicate the results. 16:Establish the performance of the ML algorithms, and propose the best one.