Research on implicit emotion recognition and classification in literary works in the context of machine learning
Yiqian Zhao, Yuanshou Zhang · Alexandria Engineering Journal · 2024
Emotion prediction is crucial in areas such as human–computer interaction, consumer experience, and mental health treatment; nonetheless, effectively forecasting emotions is difficult due to various intricate factors. This study tackles these problems by employing the Boruta algorithm for feature selection, thereby assuring that only pertinent attributes influence the predictions. The curated dataset was further examined utilizing machine learning models: support vector machine (SVM), K-nearest neighbor (KNN) method and artificial neural network (ANN). The results indicate that SVM attains the greatest accuracy of 0.91, succeeded by ANN at 0.89 and KNN at 0.88, underscoring SVM’s appropriateness for this dataset owing to its strong boundary-setting proficiency. Although ANN adeptly accommodates intricate patterns because to its flexibility, KNN’s marginally diminished accuracy may result from its susceptibility to class overlap. All models demonstrate robust predicted accuracy, confirming their dependability for emotion classification tasks. The findings indicate that SVM, specifically, may improve applications in user experience, mental health, and AI-facilitated customer interactions, providing significant assistance for data-informed decision-making across many sectors.