Sentiment analysis is a crucial task in natural language processing, with significant implications for applications such as customer feedback and product reviews. This study investigates whether integrating cross-validation techniques (repeated \(k-\) fold and stratified \(k-\) fold) with pool-based and stream-based active learning can improve the performance of an active learning model for sentiment classification.Results show that the incorporation of cross-validation into active learning significantly increases model performance, achieving an average Area Under the Curve gain of around 2%. This improvement demonstrates the effectiveness of cross-validation as a reliable validation technique, leading to better generalisation and robustness. However, this boost comes with a trade-off: while it elevates model performance, it requires additional computing power and time due to the iterative process of training and validating multiple times during cross-validation. The findings highlight the importance of incorporating cross-validation into active learning methods for sentiment analysis, despite the Area Under the Curve gain/computational time trade-off. As datasets grow in size, pool-based active learning becomes increasingly computationally burdensome. In such scenarios, stream-based active learning emerges as a more practical and efficient approach.