The analysis of spring wheat productivity by the method of principal component

Abstract

The work deals with the use of the factor analysis to reduce the importance of factors affecting spring wheat productivity. To carry out the analysis among eight most affecting on crop productivity factors we used the dataset of 32 years. All the dataset has been pre-normalized, being centered and presented in tabular form. Eight principal components were calculated, and the factor loadings were determined. According to the factor loadings it has been decided to take four principal components describing 84% of general dispersion. Each principal component has been presented as a linear combination of factor loadings and factors. The use of principal components allowed reducing the size of initial dataset from eight factors to four ones. The obtained information has been given in the space of principal components. The new coordinates of the experimental dataset on spring wheat productivity has been estimated by the received interdependences. The initial dataset has been given in a graphic form to search latent interdependences among factors. The number of diagrams in four principal components is six variables in a binary space and four ones in tree-dimensional space. It has been given a diagram of the data according to the first and the fourth principal component. The location of the points shows that the largest value of kernel weight is connected with high indexes of gluten content. It has been constructed a model on the basis of neural network (multiclass perceptron) with one input, one output and one latent layer. The neural network has been preliminary studied according to initial dataset and checked for adequacy