This research develops a machine learning-based system to predict maintenance priorities for heritage buildings in Johor Bahru, emphasizing the influence of microclimate conditions such as temperature, relative humidity, precipitation, and wind speed. It evaluates various algorithms, including XGBoost, decision tree, logistic regression, recurrent neural network (RNN), support vector machine (SVM), and linear regression, to identify the most effective model for predicting maintenance needs. The methodology involves collecting data from the Copernicus Climate Data Store and conducting condition assessments of heritage buildings. Performance is evaluated using metrics like MAE, RMSE, and R2. Additionally, a user-friendly Power BI dashboard is developed to display predictive results and monitor building conditions. The research is highly relevant to the field of heritage preservation, offering a novel, data-driven approach to prioritizing maintenance tasks based on microclimate conditions.