Neural modelling as a prediction method of starch content in potatoes for post-registration and specific agricultural experimentation

Important role in the production of potatoes in Poland is played by their industrial and nutritional value. Among the cultivated potato varieties those that are characterized by a good culinary qualities and technological parameters for the production of starch, chips, crisps, dried, etc. are distinguished. Knowledge of the most important functional, biological and agronomic characteristic requirements allows for proper selection of varieties adapted to the direction of use, and thus to customer requirements and market outlets. Helpful, in the selection of varieties, can be results obtained from the Post-registration and Specific Agricultural Experimentation. The genotype-environment variability experiments, carried out all around the country, provide a possibility to apply Artificial Neural Network in the above work in order to generate a model enabling to predict the starch content in potato tubers. Based on the collected data, a model of Artificial Neural Network MLP 14 : 70-19-10-1 : 1 was built. The model using 14 independent features enables predicting the starch content in different potato varieties