Sentiment analysis, a branch of natural language processing, aims to discern the emotional tone behind a piece of text. In recent years, Convolutional Neural Networks (CNNs) have emerged as powerful tools for sentiment analysis due to their ability to capture local dependencies in text data. This paper presents a comprehensive study on sentiment analysis using CNNs, focusing on classification tasks. We explore various architectures, preprocessing techniques, and hyperparameter tuning strategies to optimize performance. Our experiments demonstrate the efficacy of CNNs in accurately classifying sentiment in text data across different domains and languages. Additionally, we compare CNN-based approaches with traditional methods and discuss their advantages and limitations. Overall, this work contributes to the advancement of sentiment analysis techniques, particularly in leveraging deep learning models like CNNs for robust and accurate sentiment classification.