Análisis de técnicas de aprendizaje automático para la clasificación de palabras en un curso virtual de la plataforma MOODLE

Abstract

The present work aims to try some of the machine learning techniques; focused on the use of neural networks and Bayesian networks, to obtain the most accurate algorithm with which the text classification can be perform. Some datasets and tools were implement to apply the algorithms with actual MOODLE messages, previously its necessary to made the natural language processing through labor "Adaptation of a processing tool for natural language labeling feelings and analysis of language used in Spanish "focused on feelings. Now days there are many machine learning techniques which help us to implement programs that can improve their performance through experience at the time we are getting results. A comparison of the results of the tools and techniques were performed; so Bayesian Networks was selected to obtain better result by testing in both instances correctly and accurately classified. The algorithm selected was "Multinomial Naive Bayes" and validation by "Percentage Split" for best results. We worked with the "Weka" tool that allows us to use as many of these techniques algorithm and better visualization of results; and the tests were performed with different dataset to end with real-world messages of MOODLE.