Background <p>Stroke is a common cardiovascular complication. The machine learning (ML) algorithms in stroke is a new and noble approach. The objective of the study is to study the practicability of establishing a community-based stroke risk exploratory feature identification using ML and evaluate the effectiveness of individualized interventions in preventing stroke.</p> Methods <p>Retrospective data was retrieved of stroke patients (<i>n</i> = 158) and healthy population (<i>n</i> = 358) who underwent physical examinations in the community. Random forest and Least Absolute Shrinkage and Selection Operator (LASSO) regression were chosen as feature variables. To create a model for predicting the risk of stroke, the multivariate logistic regression analysis was used. Then, a prospective study was conducted with 72 high-risk individuals and divided randomly. Research group received tailored interventions and conventional interventions, while the control group received routine interventions.</p> Results <p>Triglycerides, apolipoprotein B (apoB), serum creatinine, high sensitivity C-reactive protein (hsCRP), and homocysteine were determined as risk factors for stroke by the exploratory feature identification. The prospective study revealed no statistically significant differences in baseline characteristics between the two groups (<i>P</i>&gt;0.05). The incidence of stroke in research group was 5.41%, remarkably lower than the control group (28.6%, <i>P</i>&gt;0.05). The time to stroke occurrence was significantly longer in research group than in control group (<i>P</i>&gt;0.05).</p> Conclusion <p>The community-based stroke risk exploratory feature identification established using ML can effectively screen high-risk populations. Individualized intervention measures can reduce the incidence of stroke significantly.</p>