Bone mass and bone structure are both diminished during the course of osteoporosis, a degenerative bone disease. For millions of individuals all around the globe, it is the primary source of suffering. In order to intervene and treat osteoporosis in a timely manner, early prognosis of the disease is crucial. The suggested method entails creating a system for assessing the risk of osteoporosis by machine learning (ML). This system uses the Osteoporosis Risk Prediction dataset comprising 14 demographic and clinical variables and 1,958 patient records. Various methods were performed to improve the data and to make it balanced between the classes. Label encoding, Missing value imputation, Procedures for removing duplicates, handling outliers, imputation of missing values, and SMOTE (Sequential Minority Oversampling Technique). For the goal of osteoporosis prediction, four ML models were evaluated: Decision Tree, XGBoost, K-Nearest Neighbours (KNN), LightGBM (LGBM), and Random Forest (RF). In comparison to the other models, the Random Forest model had the best results in terms of accuracy (ACC) (93%), precision (PRE) (92%), recall (REC) (93%), and F1-score (F1) (92%). Characteristic importance analysis also helped identify the key risk variables for osteoporosis prediction. The results indicate that the proposed random forest-based method is accurate and reliable in early osteoporosis diagnosis and it could assist health-care providers to make clinical decisions and diagnose osteoporosis at an early stage.