Calibration of atmospheric pollution sensors using Quantum Machine Learning techniques

The need to efficiently process large amounts of data has led information technologies to migrate to the world of Machine Learning, to be able to extract information with a variable degree of accuracy, but still quite high. The use of Quantum Computing for Machine Learning occurs in a similar, but not entirely, scenario, as the amount of data that can be processed tends not to scale easily. The purpose of this thesis is to exploit these new computing paradigms to analyze and calibrate sensors for measuring air pollution, in particular PM-2.5 and PM-10, this is one of the most interesting potential applications of Quantum Technologies: Machine Learning for the purpose of prediction. Furthermore, it is shown that some problems that are generally difficult to calculate can be easily processed by Classical Machine Learning, which is trained on one part of the initial data-set and then is tested on the rest. As a starting example, a case study on the prediction of the price of a house in the classical version is analyzed and then this example is extended to the quantum version. Using classical prediction algorithms such as Linear Regression as a basis, the next intent is to develop a methodology for evaluating the potential quantum advantage in learning and predicting correct values for air pollution detection sensors. The results are promising, compared to classical methodologies, the quantum one is able to calibrate better, with a percentage error almost always below the 10% threshold.