Feature Factory

Abstract

We examine the process of engineering features for developing models that improve our understanding of learners' online behavior in MOOCs. Because feature engineering relies so heavily on human insight, we engage the crowd for feature proposals and guidance on how to operationalize them. When we examined our crowd-sourced features in the context of predicting stopout, not only were they impressively nuanced, but they also integrated more than one interaction mode between the learner and platform and described how the learner was relatively performing.

Sample from copula: a COPPY module

“…A direction of improvement for the COPPY module is dependence modeling with Vine copula which has been recently a tool of high interest in the machine learning community see, e.g., [Lopez-Paz et al, 2013, Veeramachaneni et al, 2015, Carrera et al, 2016, Gonçalves et al, 2016 or [Sun et al, 2019]. Therefore, it strengthens the need of dependence modeling with copulae in Python as a not negligeable part of the machine learning community use this language.…”

Tools for Educational Data Mining

“…In EDM, as well as in other areas of data mining and data science, transforming raw and inchoate data streams into meaningful variables is the first major challenge in the process. Often data come in forms (and formats) that are not ready for analysis; the data not only need to be transformed into a more meaningful format, but in addition meaningful variables need to be engineered (see section 3.3 in Baker, 2015, or Veeramachaneni, Adl, & O’Reilly, 2015, for a more thorough discussion of this process). In addition, data often need to be cleaned to remove cases and values that are not simply outliers but actively incorrect (i.e., cases where time stamps have impossible values, instructor test accounts in learning system data, etc.).…”

“…A researcher may be interested in utilizing durations between actions to identify off-task students (e.g., Baker, 2007; Cetintas, Si, Xin, & Hord, 2010) but only have access to raw time stamps. In these situations, new variables must be created in order to conduct the desired analyses, a process termed feature engineering (Baker, 2015; Veeramachaneni et al, 2015).…”