Gernel Lumacad

and 1 more

During the outbreak of COVID - 19 in the last 2020, the traditional face - to - face in - campus has shifted to different learning modalities (synchronous, asynchronous, modified asynchronous, etc.)  among basic educational institutions in the hope of continuing learners’ learning experiences amidst the pandemic. New datasets potential for mining and knowledge discovery emerged during this time. In this study, the learner’s enrollment and survey form (LESF) from the data repository of a private high school in the Philippines is analyzed and mined. The dataset contains learners’ information relating to the present COVID - 19 crisis, which is bound to be utilized for strategic classification of each to student to their corresponding learning modality. The dataset includes major categories of information pertaining to learners’ demographic profile, parent/guardian information, household capacity, school proximity and transportaion information, and access to distance learning.  We present in this paper a machine learning (ML) method called multilayer perceptron neural network (MLP NN) for classifying learner’s learning modality as an alternative to the recommended algorithm for learning modalities delivery (ALDM) by the Department of Education (DepEd). The MLP NN model is trained using learners’ LESF information as input features. Prior to model development, Boruta algorithm (BA) is conducted to determine important features for the learning modality classification problem. We further investigate the sensitivities for each feature with respect to each learning modality using partial derivatives method.