PROJECT TITLE :

An Automated Machine Learning Approach for Smart Waste Management Systems

ABSTRACT:

This study shows how automated machine learning can be used to solve a real-world problem in a Smart Waste Management system. The work focuses on the topic of using sensor data to detect (i.e., binary classification) the emptying of a recycling container. In a realistic situation where most of the occurrences are not true emptying, a variety of data-driven strategies for tackling the problem are studied.

The existing manually constructed model and its modification, as well as traditional machine learning algorithms, are among the strategies studied. When utilizing the best performing solution, machine learning improves the classification accuracy and recall of the existing manually engineered model from 86.8% and 47.9% to 99.1% and 98.2%, respectively, when using the best performing solution.

This solution applies a Random Forest classifier to a set of characteristics depending on the level of filling at various time intervals. Finally, the best performing solution enhances the quality of forecasts for recycling container emptying time when compared to the baseline current manually created model.