Information and communication technologies have been explored for several applications in urban environments. The Internet of Things (IoT) demonstrated potential use for collecting helpful information to aid decision-making processes. One sector currently not taking advantage of the technologies to improve its operation is the sector of Municipal Solid Waste Management Systems (MSWMS). Some studies in this strain propose alternatives that apply IoT to improve operations, such as using wireless sensor networks to measure waste levels and optimize collection routes. The studies are focused on exploring hardware development and evaluating the efficiency of their solution on a real scale. However, the strategy used to convert physical measurements into waste levels is not shown by these proposals. Here, we performed a waste-fulfilling experiment in a simulated dumpster equipped with IoT technologies. A node was installed inside a dumpster model, capable to identify the occupancy by measuring the distances from the waste inserted. The controlled waste-fulfilling experiment allowed for collecting information on real waste levels. It was then used to train machine learning models to measure waste levels based on the amount of usage over time. The machine learning models’ efficiency was compared to a linear regression model to validate the strategy by Mean Absolute Error (MAE) metric, which ranges from 0.033 to 0.11.