ML, IoT, and ERP systems used together will simplify data-driven decision-making. This work replicates an IoT-based data collecting system. CSV files document sensor data including humidity, temperature, and inventory as well as other factors. Saved and under observation is this data. A binary categorization model forecasts, depending on environmental conditions, either high or low inventory levels. Following training, temperature and humidity are gauged by a Random Forest classifier. This stage consists of training and test set manipulation of 100 simulated Internet of Things data by means of scaling and separation techniques. Effectiveness of the model is gauged by accuracy, precision, recall, and F1-score. Visualizations help to evaluate model performance. Among these graphic tools are the error curve, ROC curve, and confusion matrix. The results indicate how real-time IoT data allows machine learning to forecast stock levels, hence improving ERP inventory control. By optimizing supply chains, this approach lowers running inefficiencies and increases decision-making using predictive analytics. The foundation for further studies on scalable machine learning in ERP and Internet of Things integration is set by this work. With a recall of 0.41, an accuracy of 0.41, and an F1-score of 0.41 the trained Random Forest classifier performed as over 200 examples. 41% is the categorization accuracy. In regression analysis, the measurements are a R2 score of − 0.07, a mean absolute error (MAE) of 12.29, and a mean squared error (MSE) of 222.24. Feature significance analysis suggests that temperature and humidity significantly affect inventory projections. Visualizing tools that could be useful when attempting to grasp performance include confusion matrices and feature significance graphs.