Comparison Study for Crop Growth Prediction Machine Learning Model for Vertical Farms

Tan, Ze Hong (2022) Comparison Study for Crop Growth Prediction Machine Learning Model for Vertical Farms. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text TAN ZE HONG_Thesis.pdf Restricted to Registered users only Download (3MB)

Abstract

Vertical Farms (VF) have been all the hype in the agricultural field as of late and is sure to be a strong alternative for food production in the near future. Since VF systems are considered as controlled environment agriculture (CEA), it is logical that the next advancement in the field should be to utilise the endless supply of IOT crop growth data to optimise the production of the system. This comparison study aims to probe the feasibility of deploying Machine Learning algorithms that are learning from a limited amount of growth parameter data that are not linear, with fluctuating values from small scale vertical farms. The aim is to compare up to 3 different ML models and evaluate in terms of goodness of fit metric R-squared (R2) and error metrics Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) to determine which ML model works best with limited amount of data. Two crop growth datasets were obtained; one manually while the other from an external source. Multivariate Linear Regression (MLR), Random Forest Regression (RFR) and Artificial Neural Network Regressor (ANN) were modelled using the two datasets. It is easily concluded that the self-obtained dataset was insufficient and lack quality, and the best model for the problem statement was determined to be RFR built using the external dataset with R2 of 0.9841, MSE of 0.9052, RMSE of 0.9514 and MAE of 0.5936. In conclusion the hypothesis of whether a fluctuating monitoring system dataset can be modelled for prediction was successful. With this result, small VF system should be able to deployed said modelled for better automated decision making with the forecast data of their crops available.