A Conversational Agent for Smart Schooling A Case Study on K-12 Dropout Risk Assessment

The goal of smart education is to utilize advanced technology in order to improve the teaching experience by establishing a stimulating and interactive atmosphere for learning. Conversational agents emerge as an aid for a smarter education. One of the possibilities to be explored is the building of tools that help predict and prevent student failure or dropout. This case study presents a research project that consists on the creation of a school platform for student interaction, in which a conversational agent, developed using Rasa, communicates with both the students and the class director and is able to assign a risk of academic failure, based on their answers to questionnaires scripted by a team of psychologists. XGBoost outperfomed AdaBoost, Decision Tree and Random Forest algorithms with an accuracy of 97%.