Predicting concentration of metals in aeroponic lettuce using machine learning

The world population will continue to grow until 2080, driving increased urbanization and a reducing available land for agricultural. In response to this challenge, this thesis explores vertical farming, an innovative method that maximises spaces for plant cultivation and optimises the use of resources. The aim of this thesis is the analysis to estimate the metal content of the lettuce during the various weeks of growth, using a machine learning regression model. The purpose of estimating the concentration of metals present in lettuce is to assess how the plant behaves over time in response to various environmental inputs and promote proper development. The experimental analysis was conducted inside an aeroponic greenhouse at the Polytechnic of Turin and in particular, lettuce seedlings were analysed. After an explanation of aeroponics technology, a literature review was conducted to assess the state of the art. Subsequently, data from laboratory analyses were collected, and the available database was cleaned using statistical methods such as the z-score method and interquartile range, in order to identify outliers and prepare an initial dataset. The latter was then used as input for the implemented machine learning model. The model shows that the average accuracy in estimating the concentration of metals within the lettuce is high, with a prediction deviation of approximately 14% from actual values. Indeed, this indicates a reasonable level of reliability, demonstrating that the selected features and model effectively capture the relationships between environmental conditions and metal accumulation, and are in line with the performance of models in the literature.