Integration between Information Science and Data Science:

content analysis applied to variable selection for predictive modeling

Abstract

The variables used in predictive models are fundamental to the quality of the results obtained, making their identification and selection relevant steps in the modeling process. This article aims to present the application of content analysis, a technique widely used in Information Science, in the exploration and selection of variables for predicting high school dropout among students in the Santa Catarina State public school system. The research is characterized as applied, with exploratory objectives, a predominantly qualitative approach, and documentary research and a case study as its technical procedures. The corpus consisted of 16 studies selected through a bibliographic survey, from which 269 variables were identified, subsequently grouped into 114 groupings based on semantic and conceptual similarity and classified according to the characteristics of Availability, Relevance, and Dimension. The results demonstrate that content analysis made it possible to organize and systematize the variables in a contextualized and structured manner, supporting their selection for predictive modeling. It is concluded that the application of content analysis fosters the integration of Information Science principles and Data Science methods, providing an initial systematization of the variables identified in the literature that may support future research on variable selection and predictive modeling of school dropout.