DATAMINING CAPABILITIES FOR CLUSTERING CONCRETE MIX FORMULATIONS

Abstract

The article discusses the possibilities of using DataMining technology for clustering concrete mixtures. In practice, it is often necessary to face tasks in which it is necessary to choose the formulations closest in quality characteristics from a large number of formulations of concrete mixtures. The distribution of formulations of concrete mixtures by classes is provided on the basis of specified criteria such as strength, as well as the composition of the ingredients of the concrete mixture. Earlier in work [1], clustering of concrete mix formulations was carried out with the help of the program "Comprehensive quality assessment and classification of multidimensional objects", which made it possible to distribute formulations into classes with the closest characteristics and collect the highest quality concrete mix formulations into the appropriate classes. The result of using Data Mining technology for clustering concrete mix formulations allowed us to create classes in which the distribution using the DBSCAN algorithm is quite high-quality, however, there is a need for more detailed training of this algorithm, since clustering using the program "Integrated Quality Assessment and classification of multidimensional objects" turned out to be more optimal.