Human emotion recognition using artificial intelligence has become a significant research area with applications in emotion analysis, opinion mining and affective computing for understanding users’ emotional states. This study presents an emotion analysis framework based on machine learning and deep learning models trained on a Twitter-derived emotion dataset available on Kaggle. The dataset comprises 20,000 text–label pairs categorized into six emotion classes: Sadness, Joy, Love, Anger, Fear and Surprise. Data preprocessing includes stop-word removal, lowercasing, elimination of URLs and special characters, and TF-IDF–based feature extraction, followed by label encoding for model compatibility. To provide reliable analysis, the dataset is separated into subsets for training, validation and testing. Multiple models Convolutional Neural Networks (CNN), Bidirectional LSTM (BiLSTM), Logistic Regression and a Hybrid CNN–BiLSTM model with Multi-head attention are implemented and compared. According to experimental data, the CNN model outperformed the hybrid, BiLSTM and Logistic Regression models, with accuracy of 91.90%. The results show the effectiveness of deep and hybrid architectures in capturing fine emotional tone from informal textual data. The proposed framework provides an expandable and reliable solution for real-world applications in textual data analytics, human–computer interaction and intelligent emotion monitoring systems, with future scope for transformer-based and multilingual enhancements.