Application of Data Mining Technologies for Processing Results of Experimental Studies

The paper discusses the components of data mining technology used to process the results of measurements of wheat biopotentials. Briefly described are the goals of studies of plant biopotentials, the conditions and features of measurements and the measuring installation. The results of compression of measurements are presented, which are subjected to further analysis, the methods of data normalization available in the Python programming environment are listed, the choice of clustering methods is justified and the results of their application are given. The advantages of normalizing aggregated biopotential data by the MaxAbsScaler method are concluded. As for clustering methods, the best results were shown by the Bisecting KMeans and SpectralClustering methods, which are part of the group of methods designed to work with a small number of clusters and data. It is noted that among cluster-scalable clustering methods, the GaussianMixture method has an advantage. The absence (due to sufficiently voluminous result tables) of the analysis of the influence of aggregation methods (more precisely, preliminary processing of measured signals - their centering, low-frequency filtering, differentiation) on the results of clustering is motivated. Methods for evaluating the effectiveness of the experiment plan and the direction of further research are proposed.