The Implementation of Artificial Neural Networks and Resolving Efficient Dominating Set for Time Series Forecasting on Vertical Farming Soil Moisture

Abstract

Vertical farming is the practice of growing crops, where the cultivated plants are grown vertically rather than horizontally using soil media. Vertical farming is developed specifically in certain indoor by using Controlled Environment Agriculture (CEA) technology. Since we can cultivate some plants combination in the vertical farming, thus the use of CEA technology is inevitable. The control mechanisms of the vertical farming is more complex, especially related to the control of air temperature, air humidity, light intensity, and soil moisture. In recent times artificial neural networks (ANN) has become popular and helpful model for forecasting, classification, clustering, and forecasting especially in a precession agriculture. ANN is one type of model for machine learning (ML) and has become relatively competitive to conventional regression and statistical models regarding usefulness. In this paper, we will analyze the effectiveness of two artificial neural network architectures in time series forecasting on soil moisture in the vertical farming and analyze it in term of iteration number, mean square error, regression, and their learning rate. The results shows that the best architecture for phase 1 is obtained by Feedforwardnet ANN-567 model, with MSE of 0.3810. The best architecture for phase 2 is obtained by Paternet ANN- 567 model with an MSE of 1,1374x10^(-9). The best architecture for phase 3 is obtained by Cascadeforwardnet AA-567 model with an MSE of 1,1232x10^(-10). Finally, the best architecture for phase 4 is obtained by model Cascadeforwardnet ANN-567 with an MSE of 1,0780x10^(-17).