The incidence of crime is now of great concern globally. The culprits change their tactics on a regular basis. These crimes affect persons, groups, and the government to the extent a whole lot of budgets are allocated to serve as preventive measure to these crimes. The aim of this research is to predict crime based on Twitter hourly sentiments and crime data records. This is because it has been observed that existing crime prediction models that used Twitter data entail some drawbacks in predicting criminal incidents as a result of the unavailability of hourly sentiment polarity and demographic factors. Additionally, SHAP framework was used for the interpretability to rank the feature based on their importance. The xgboost algorithm was utilized with tuning to have an optimal model. The accuracy of 0.81 (81%) was obtained and an Area Under the Receiver Operating Curve (ROC AUC) score of 0.7079 was obtained. The result of this study indicated that crime could be predicted in real-time in contrast to earlier studies on this subject matter. Consequently, it is advised that this work be applied to real-world situations