Research into the incidence of stroke has seen as many failures as successes. The risk of stroke was one of the leading causes of mortality and disability worldwide, requiring prediction, early identification, anticipation and prevention. The healthcare sector has been limited by a lack of resources, understanding, transparency and better interpretability, which encouraged medical experts to adopt measures. Predicting and identifying the factors responsible for stroke risk remains a very complex task. To overcome these problems, the integration of artificial intelligence was the solution envisaged. Three machine learning models, such as Random Forests, K-Nearest Neighbors (K-NN) and a Logistic Regression model applied to Lime, were constructed, evaluated and compared on a Japanese dataset from the Okayama region comprising 12 parameters: ID, gender, age, hypertension, heart disease, marital status, smoking status, type of work, type of residence, glucose level, body mass index and stroke. The results were satisfactory and demonstrated that the approach combining Logistic Regression with the acronym Lime stood out thanks to its excellent results, which surpassed those of other models with an accuracy of 99.9%, sensitivity (100%) and F1-Score (99.0%). In addition, many factors responsible for this incidence were very well classified: age, hypertension and blood glucose levels were the main ones. These factors were strongly correlated and were more prevalent in older people.