Aplicação de algoritmos de aprendizagem de máquina nos dados de partidas do jogo sério Querida Floresta

Abstract

Serious games are effective teaching tools that can be applied to a wide range of public audiences, including students with intellectual disabilities. For this audience, it is important to have the appropriate level of difficulty for the educational game activities in order to keep them motivated and focused to continue playing. Data learning techniques aid in the process of discovering characteristics of a group of individuals and thus determining a more appropriate level of activities that can be attributed to them. As a result, this work conducted an analysis using machine learning algorithms on the matches of a serious game called Querida Floresta. This analysis was made possible after modifying the game, as it did not include mechanisms to store data about the player's matches. The game was made available for intellectually disabled students from an regional institution during the months of october to december. Through the collected data and the application of the J48, K-Means, and JRip algorithms, a lack of interaction was identified, or delay, for users to perform the first activity (click on the cloud) and the second (dragging an object), but for the third activity it was observed that there was no difficulty for the player to complete it. Thus, suggestions for improvements to Querida Floresta were proposed, based on the results obtained, as well as changes in the way the hint is made available to the player.