Deep Active Learning Multiclass Classifier for the Sentimental Analysis in Imbalanced Unstructured Text Data
M.Ramesh Raja, J. Arunadevi · 2023
This paper presents an active learning framework for multi-class sentiment analysis, focusing on addressing the challenges of class imbalance. The proposed approach combines Convolutional Neural Networks (CNN) as the base classifier, logistic regression as the meta classifier, and alternative meta classifiers including Random Forest, SVM, and Gradient Boosting. The active learning algorithm intelligently selects informative instances for labeling, reducing the need for large amounts of labeled data and manual annotation efforts. The CNN-based classifier automatically extracts relevant features from the text data, enhancing sentiment analysis capabilities. The logistic regression meta-classifiers improve classification performance. Experimental results demonstrate the effectiveness of the proposed approach in handling class imbalance and achieving accurate sentiment classification across different classes. The active learning framework with CNN as the base classifier and logistic regression as the meta-classifier shows promising results, offering a valuable solution for multi-class sentiment analysis tasks where labeled data acquisition is resource-intensive.