An Enhancement of the Gaussian Naive Bayes Algorithm Applied to Air Quality Classification

The Gaussian Naive Bayes Algorithm is a machine learning technique based upon the Bayes Theorem. It is commonly used for classification tasks to calculate the likelihood of events. This study developed an enhanced GNB algorithm to classify the air quality in Pamantasan ng Lungsod ng Maynila. The enhancement made in this study sought to increase the classification performance of the traditional GNB against zero frequency issues. The zero-frequency problem is an inherent limitation of the conventional GNB due to the algorithm’s reliance on multiplying probabilities. It occurs when a feature value is absent from the training data. The Parzen-Rosenblatt Window method was applied to address the issue and increase the algorithm’s stability against the problem. OpenWeather-AQI and USA-AQI datasets were used to evaluate the algorithm. The algorithm's accuracy improved from 71.77% to 74.16% (2.39%) in the OpenWeather-AQI dataset. In comparison, the other dataset showed a 5.26% improvement, increasing from 59.33% to 64.59%. These results showcase how the enhanced GNB algorithm outperforms the traditional one. Thus, the Enhanced GNB Algorithm effectively addresses the zero-frequency problem, increasing classification accuracy and demonstrating its potential as a reliable method for assessing air quality.