Indoor Microgreen Precision Farming for Optimal Yield Using Machine Learning Regression Method

Yeap, Soon Hao (2023) Indoor Microgreen Precision Farming for Optimal Yield Using Machine Learning Regression Method. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text Full Text -Yeap Soon Hao.pdf Restricted to Registered users only Download (2MB)

Abstract

This paper proposes an innovative approach to indoor microgreen farming using machine learning regression techniques to optimize yield in Malaysia’s climate. Microgreens are young vegetable seedlings that are harvested after a short growth cycle of seven to twenty-one days. They are known for their high nutrient density, containing higher concentrations of antioxidants, vitamins, and minerals than fully developed vegetables. This makes them a beneficial addition to a healthy diet. Due to the uniqueness of the enclosure system used to nurture the microgreen’s growth, the three dimensional (3D) drawing method and the technology of additive printing are implemented. The proposed machine learning (ML) based approach involves collecting data on the microgreens, processing the data to extract insights, and developing ML regression models to predict optimal growth conditions. Results from this research utilises eight different machine learning algorithms which are linear regression, multiple linear regression, multiple predictor variable polynomial with the degree of freedom of two and three, random forest regression, multiple predictor variable random forest regression, gaussian process regression, and multiple predictor variable gaussian process regression. The most promising results among the machine learning regression algorithms implemented are the multiple predictor variable random forest regression and the multiple predictor variable gaussian process regression. The multiple predictor variable random forest regression has the root mean square error (RMSE) of 0.86440, mean square error (MSE) of 0.7471, mean absolute error (MAE) of 0.5030 and the R-Squared with the value of 0.96967. The multiple predictor variable gaussian process variable, on the other hand achieved the regression parameters of RMSE of 0.86719, MSE of 0.7520, MAE of 0.5401, and the R-Squared of 0.96947. This approach has the potential to revolutionize indoor microgreen farming, optimizing growth conditions and maximizing the nutritional value of these plants.