Modelo predictivo para la selección de becarios mediante Machine Learning

Abstract

Abstract: This degree work identifies the factors that influence the selection process of scholarship holders through the investigation of several resources found by different universities analyzing the requirements that they propose, their bases and policies, as well as the diverse information of articles that propose predictive models for the selection or different processes. The analysis was performed in tools that facilitate its analysis, through exploration, statistical summaries and graphs where it is possible to identify in a certain way the relationships and patterns involved in such data set, such as the determination of important characteristics for the training of a particular model according to its process. The analysis of the data involves an exploratory treatment (EDA), thorough to obtain the training and test data of each model. Where the model must comply with certain characteristics such as treatment and cleaning, data balancing, standardization, transformation of its variables to the format required by different model to obtain good results a model must be evaluated under various calculations such as precision, sensitivity, accuracy among others that are results of metrics of the right and wrong values of the prediction, through which it is determined which of the developed models are with the objective of the project. The models are evaluated, and the results are shown visually to demonstrate their process and result in the process as an alternative to the process in a short time it is possible to predict a population of students under specific terms that each model determines. In determining the effectiveness and equity in the distribution of resources through the process of the models trained with historical data, all this was done through an iterative methodology and the adaptation of the CRISP-DM methodology.