Comparative Analysis of Gaussian Naïve Bayes with Bernoulli Naïve Bayes for Guitar Chord Classification

Naïve Bayes Algorithm is extensively used in machine learning for classification problems. Guitar chord classification is a trivial problem in music recognition where Naïve bayes algorithm is exploited. To further investigate the performance of Naïve bayes, its types can be experimented. In this paper, we have attempted to investigate two Naïve bayes techniques viz, Gaussian Naïve Bayes and Bernoulli’s Naïve bayes which are tested on guitar chord dataset with major and minor chords. Guitar chord classification problem has been explored and experimented in more than one dimension for music recognition and recommendation models. The preprocessing of guitar chords audio into numerical data is computationally extensive. In order to test the performance of variants of Naïve Bayes Algorithm we have considered the harmonics generated during the preprocessing and data transformation stage and analyzed the effect of higher order harmonics. In this paper we have evaluated and compared two versions of Naïve bayes algorithm for guitar chord classification problem for two sets of harmonics count. Gaussian Naïve bayes is compared with Bernoulli’s Naïve bayes for datasets of guitar chords with harmonics count from 4<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">th</sup> harmonic to 9<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">th</sup> harmonic. This paper has interesting findings that at lower harmonic counts Gaussian Naïve bayes performs better and if the feature count of harmonics is higher, in that case Bernoulli’s Naïve bayes exhibits better performance. All the results are elaborately discussed and visualized in the result section of the paper. Further, the classification problem and the experimental results are obtained, presented and compared with the existing work in the results and discussion section.