Enhancing the Performance of Machine Learning Classifiers by Hyperparameter Optimization in Detecting Anxiety Levels of Online Gamers

Fine tuning attribute weighted naïve Bayes model for detecting anxiety disorder levels of online gamers

“…The proposed method in this study tended to have an inordinately higher proportion of performance. Our study suggests that higher accuracy is not associated with the poor performance of the naive Bayes model in previous research using the same dataset [9]. The proposed method may benefit from initializing probabilities and fine-tuning phases without adversely impacting accuracy.…”

“…Meanwhile, a study used a dataset of anxiety disorders in online gamers from Kaggle on nine machine learning classifiers. In this research, multi-layer perceptron provided the best accuracy while gaussian naive Bayes produced the lowest accuracy compared to the other nine classifiers on the dataset used [9]. A different study used deep learning for the classification of anxiety, depression and comorbidities.…”

“…This research concludes that machine learning methods are more suitable for complex datasets [11]. One reliable classification model often used to analyze and identify psychological problems is naïve Bayes [8], [9], [11]- [15]. However, naïve Bayes does not provide the best accuracy for classifying anxiety disorders in gamers compared to the other eight classification algorithms.…”

“…In research on anxiety disorder, a study conducted a comparison of nine classification algorithms, including Gaussian naïve Bayes (GNB), k-nearest neighbor (KNN), decision tree, random forest AdaBoost, support vector machine (SVM), gradient boosting, multi-layer perceptron (MLP) and XGBoost. MLP provides the best accuracy of 99.96% of the nine algorithms, and GNB provides the lowest accuracy of 79.46% [9]. In research on internet addiction by Ioannidis et al [33] naïve Bayes outperformed in crossvalidation settings, but its performance varies more in PR-AUC.…”