Optimize Sentiment Detection Model with Hyper-Parameter Tuning

Sentiment detection (SD) is a natural language processing approach for detecting human's opinions, feelings, and attitudes toward a certain objective, including individuals, groups, companies, subjects, activities, services, and goods. It is widely utilized in online and social media platforms, customer-generated content and healthcare materials for marketing, customer service, and clinical medical applications. Several traditional and machine learning models have been investigated in the past literature that do not appropriately optimize SD models due to non-linear information processing. Hence, we propose a hyperparameter tuning and optimizer selection (HT-OS) technique to efficiently optimize and detect human sentiment in this paper. We employ a computation-efficient gate recurrent unit (GRU)-based deep learning model to establish a suitable pipeline for the proposed HT-OS technique. Using an open-source textual mental health dataset, the proposed HT-OS approach with GRU model shows a 66% increase in model error-performance compared to the traditional GRU model and 98.59% predicted confident levels.