The sentiment analysis uses computational methods that allow automatic detection of sentiment polarity from data published mainly on the web. While machine learning techniques have been employed for sentiment analysis in fields such as medicine, economics, law, tourism, and others; it has led to an increased concern for interpretability and explainability of machine learning systems. In general, the design and development of more interpretable and explainable machine learning models remains a current challenge. This paper briefly describes an explainable sentiment analysis method for restaurant reviews based on an evolutionary algorithm. The proposed method calculates the sentiment polarities of each review by preprocessing each review using natural language processing, calculating the initial positivity of each word using a lexical dictionary, constructing an evolutionary model for sentiment classification, and explaining the results using a strategy based on word polarity scores by highlighting the words that influenced the classification. The feasibility of the proposed method is evaluated using a restaurant domain dataset; and its performance and explainability are compared with other classic methods available in the literature, which indicates promising results in terms of transparency for the end users.