Unit:
Κατεύθυνση / ειδίκευση Δικτύωση Υπολογιστών (ΔΙΚ)Πληροφορική
Author:
Koursioumpas Nikolaos
Magoula Vasileia
Supervisors info:
Αθανασία Αλωνιστιώτη, Αναπληρώτρια Καθηγήτρια, Τμήμα Πληροφορικής και Τηλεπικοινωνιών, Εθνικό και Καποδιστριακό Πανεπιστήμιο Αθηνών
Original Title:
Music Recommendation System based on EEG Sentiment Analysis using ML Techniques
Translated title:
Music Recommendation System based on EEG Sentiment Analysis using ML Techniques
Summary:
Over the years, numerous studies have demonstrated that music can produce distinct effects and feelings on people. Although it is relatively easy to name different types of emotions, it remains difficult to relate them to the real emotions experienced by a person. In addition, there are many people who listen to a specific genre of music that they think it is enjoyable when in fact that genre might have a negative effect on them. The current thesis, will try to develop a music recommendation system that will base its output on emotions extracted from Electroencephalography (EEG) data so as to stay as close as possible to the human nature. The system, which is based on Machine Learning techniques, comprises the following features: (a) Processing of EEG data in order to perform various feature extraction methods; (b) perform data augmentation so as to enrich the current dataset; (c) make use of a proper dimensionality reduction method that will find correlations in the data and discard non-critical information; (d) implement classification methods that are able to predict emotion related labels (valence, arousal, dominance, liking); (e) map the predicted emotion related labels into real emotions (excited, happy, angry, sad) and (f) integrate the best models, with the use of a voting method, into a final music recommendation system.
Main subject category:
Technology - Computer science
Keywords:
Music Recommendation System, Electroencephalography (EEG), Sentiment Analysis, Classification Algorithms, Feature Extraction Methods
Number of index pages:
18
