This paper presents a novel approach for sentence-level sentiment classification using an extended pinball loss function-based twin support vector machine with deep learning (EPLF-TSVM-DL). The method involves data preprocessing, word embedding, CNN-based feature extraction, and EPLF-TSVM-DL for classification. The proposed method outperforms existing classifiers in terms of time consumption, convergence, complexity, stability, and classification accuracy. The study uses two datasets: Sentiment140 and COVID-19_Sentiments. The results show that the EPLF-TSVM-DL achieves high accuracy, precision, recall, and F1-score for both datasets. The method is effective in handling imbalanced data and improves the classification of sentiment sentences. The study also highlights the importance of feature extraction and the benefits of using deep learning techniques for sentiment analysis. The proposed approach is a robust solution for sentence-level sentiment classification, with potential applications in social media monitoring, customer feedback analysis, and other areas requiring sentiment analysis.This paper presents a novel approach for sentence-level sentiment classification using an extended pinball loss function-based twin support vector machine with deep learning (EPLF-TSVM-DL). The method involves data preprocessing, word embedding, CNN-based feature extraction, and EPLF-TSVM-DL for classification. The proposed method outperforms existing classifiers in terms of time consumption, convergence, complexity, stability, and classification accuracy. The study uses two datasets: Sentiment140 and COVID-19_Sentiments. The results show that the EPLF-TSVM-DL achieves high accuracy, precision, recall, and F1-score for both datasets. The method is effective in handling imbalanced data and improves the classification of sentiment sentences. The study also highlights the importance of feature extraction and the benefits of using deep learning techniques for sentiment analysis. The proposed approach is a robust solution for sentence-level sentiment classification, with potential applications in social media monitoring, customer feedback analysis, and other areas requiring sentiment analysis.