Application of artificial neural networks for the prediction of quality characteristics of potato tubers - innovator variety
Journal Title: Journal of Research and Applications in Agricultural Engineering (ISSN 1642-686X) - Year 2018, Vol 63, Issue 4
Abstract
The aim of the research was to create a model for prediction of tuber dry matter on the basis of underwater weight of tubers (UWW), with the use of neural modelling methods. In order to achieve the aim of the study, data from the years 2011-2017 were collected from the production fields of an individual farm located at the border of Pomeranian and West Pomeranian Voivodeships in Słupski and Sławieński districts. The subject of the research concerned potatoes of the Innovator variety, which were grown for processing purposes - production of French fries. To build a neural model, data from September sampling as well as meteorological and fertilizer data were used. A total of 82 learning cases from the fields covered by the analyses were used, which were divided into two sets. Set 1, for the construction of the neural model consisted of 75 samples. Set 2, which consisted of 7 randomly selected samples, had a validation function and did not participate in the construction of the neural model. For proper model validation, four forecast error measures were used, i.e. relative approximation error (RAE), root mean square error (RMS), mean absolute error (MAE), mean absolute percentage error (MAPE). The model MLP 8:8-12-5-1:1 (BP100,CG31b) was based on eight inputs (meteorological data, fertilization levels) and one output (dry matter of tubers under water). The analysis resulted in a forecast error of 2.81% of MAPE. Moreover, the sensitivity analysis of the neural network showed that the mean air temperature in the period from April to September (T4-9) had the greatest influence on the dry matter of tubers.
Authors and Affiliations
Gniewko Niedbała, Magdalena Piekutowska
Keywords
Related Articles
- EP ID EP515039
- DOI -
- Views 67
- Downloads 0
Assessment of the accuracy of the approximate method used to estimate the heating power demand for single-family houses
Total design heat load as well as the approximate method based on the thermal characteristics of the building were deter-mined in accordance with the PN-EN 12831 standard for 84 randomly selected single-family residentia...
Condition of winter triticale in pure and mixed sowing
The aim of this thesis was to evaluate the impact of a sowing method (pure, 2-species and 3-species sowing) on the health condition of an ear, leaves and a stem base of winter triticale, Borowik variety. A single-factor...
Modeling of carrot root with the use of photoelasticity method
This paper presents elastooptical method for assessment of carrot root displacement under impact load. The main purpose of this study involved an analysis of cross-section model of carrot root performance. The investigat...
Use of lawn grass mixtures with iron sulphate for the renovation of lawn with moss
The experiment was established on the shaded area of the lawn occupied by mosses (Bryophyta sp.) around the Biocentrum building of the University of Life Sciences in Poznań. Lawns with moss were selected for the study, w...
Agglomeration process of post-production tobacco waste
The paper presents the research results of the agglomeration process of post-production tobacco waste arising during the cigarette production process at British-American Tobacco in Augustów. The agglomeration process tes...