Sugarcane yield prediction is important for making accurate management decisions during sugarcane growth. Genetic algorithm (GA) optimized neural network can improve the prediction efficiency and prediction accuracy, and find the optimal solution quickly by high-speed calculation. Based on the meteorological factors (atmospheric humidity, atmospheric temperature, rainfall), field hydrothermal factors and sugarcane yield obtained from the field IOT at Zhanjiang Observation and Experimental Station during 2011—2020, BP neural network and GA-BP neural network models were used to predict and correlate the sugarcane yield in the selected areas. The results showed that the correlation coefficients of Pearson and Spearman showed that sugarcane yield was highly significantly correlated with monthly maximum soil temperature, monthly minimum soil temperature, monthly average soil temperature, monthly maximum atmospheric temperature, monthly average atmospheric temperature, monthly average atmospheric humidity with correlation coefficients higher than 0.7, significantly correlated with monthly average soil water content, monthly rainfall, and weakly correlated with monthly minimum atmospheric temperature. Under the GA-BP neural network model, the prediction accuracy of sugarcane yield was significantly higher than that of the BP neural network model, with R2 reaching 0.8428, MAPE of only 0.90%, and RMSE of 1.10t/hm2. The degree of fit between the predicted and experimental values was high, and the V-cross validation results showed that the model prediction results were accurate and stable. Therefore, GA-BP prediction can predict sugarcane yield more scientifically and rationally, which was an important guiding significance for sugarcane field management measures and coordinated allocation.