Mood classification is a crucial and complex task in the fields of psychology and Natural Language Processing (NLP). It is particularly challenging to assess an individual’s mood based on their chats or comments on social media platforms. This study introduces a Convolutional Neural Network (CNN) for mood classification and evaluates its efficacy on the ISEAR dataset. The dataset comprises 7,652 short texts labeled with seven different moods including joy, anger, fear, guilt, disgust, sadness, and shame. Our proposed model achieves an impressive F1 score of 94.59%, surpassing existing statistical and deep learning models. The superior performance can be attributed to two key features of our approach. Firstly, our model incorporates the local structure of the text by leveraging relationships between adjacent word embeddings, allowing it to capture the subtle nuances of language. Secondly, it captures character-level details within the input text, allowing for a more comprehensive understanding of the data. Overall, our study demonstrates the effectiveness of the CNN-based approach for mood classification in NLP, with potential implications for developing automated systems that can accurately recognize and respond to people's moods in various applications, such as mental health and customer service.