Prediction Model of Greenhouse Tomato Yield Using Data Based on Different Soil Fertility Conditions

Tomato yield prediction plays an important role in agricultural production planning and management, market supply demand balance, risk management. To solve the problems of low accuracy high uncertainty tomato methods solar greenhouses, based on experimental data for water fertilizer consumption by greenhouse tomatoes different regions over many years, this paper investigated models yields under three soil fertility conditions (low, medium, high). Under these conditions, were predicted using neural network model (NN), particle swarm optimization (PSO–NN), adaptive inertia weight algorithm (AIWPSO–NN), improved (IPSO–NN). The results demonstrate that evaluation indexes (mean square error, mean absolute R2) IPSO–NN proposed superior to other (i.e., NN model, AIWPSO–NN model) conditions. Among them, compared with MSE decreased 0.0082, 0.0041, 0.0036; MAE 0.0759, 0.0511, 0.0489; R2 0.8641, 0.9323, 0.9408. These indicated had a higher predictive ability In view technology may be beneficial management decision support.