Preparation of the Autumn Brassica napus Yield Map by Using Perceptron Neural Network, Case Study: Sabzevar Township

Brassica napus as an oilseed is a strategic agriculture product in Iran. Knowing the best area for cultivation, helps plan and explore suitable cultivation areas efficiently. At this research, we study 24-Brassica napus farm samples to calculate the actual yields by using GPS set. The independent data are variables introduced such as mean temperature, growing degree-day, mean absolute minimum temperature, mean absolute maximum temperature, mean temperature in sowing time, mean potential evapotranspiration, slope, EC of groundwater, pH of groundwater, and mean relative humidity. We analyzed the consequences of environmental potential on autumn Brassica napus yield using Perceptron neural network by multi-layers structure with 3 hidden layers and feed backward algorithm entered to database and Brassica napus cultivation suitability map was prepared in geographic information system environment. Results show that T-test between actual and prediction values do not have significant difference in 0.05 level. We calculated Pearson correlation coefficient at 0.98, RMSE and MAE. Thus 243 and 101 Kg/h indicate the ability of neural network methods for agricultural product yield predictions.