As nowadays Adverse drug reactions (ADRs) are increasing rapidly, which leads to more health complications and higher healthcare costs. So predicting drug side effects has become important. Rather than relying on traditional methods to predict side effects of drugs we used different machine learning algorithms. For this study we used 3 different datasets from SIDER, MIMIC-IV and DrugBank [2]. These datasets together contain around 70000 patients’ drug diagnosis records and 1359 different side effects. Four different machine learning models were used for the prediction: XGBoost, CatBoost, LightGBM and Balanced Random Forest [2, 5]. These 4 different models were trained on the dataset and the accuracy, F1-score, prediction [5] and recall was compared to select the best model. For building the web application we used streamlit which is easy to understand, interactive and easily deployed. This web application provides real-time clinical decision support, and also minimizes the complex backend and frontend work for the developer. After use of all different methods we came to know that gradient boosting algorithms performed better than all other methods especially for the multi class side effect prediction [1, 2, 5]. Basically this work provides the comparison between different types of machine learning models for predicting the drugs side effect with better accuracy, prediction, F1-score and recall value. Also demonstrates how these models can be deployed in the real clinical setting.