TABULAR ARTIFICIAL NEURAL NETWORK IMPLEMENTATION OF RADIAL BASIS FUNCTIONS FOR THE SAMPLES CLASSIFICATION

Abstract

The development and study of a new constructive algorithm for constructing models for sample classification using an artificial neural network with radial basis functions in a Microsoft Excel spreadsheet environment without VBA programming is presented in the subsequent work. The algorithm presented can be considered the most effective method for solving classification problems using artificial neural networks, since a model constructed in this manner is easily expanded and modified, which facilitates its application to solve many similar problems. Creating table models using this algorithm significantly expands the functionality of spreadsheets as a simple and efficient data modeling and visualization tool. The developed table model of an artificial neural network with radial basic functions and the general recommendations about her expansion, modification and application are provided in problems of classification. Results of classification by RBF network of unknown samples based on set educational a vector samples are shown. The tabular model, which is presented in the article, has multiple advantages including its exceptional visibility, which can be effectively used in the educational process for the purpose of studying algorithmic features of neural network operations. Table modeling technology developed for classification algorithms is highly useful for educational purposes, as it provides students with unlimited access to data structures and the algorithms necessary for their processing. Further, it visually displays the intermediate dynamic mode as well as output simulation results. The offered algorithm of creation of models can be also interesting to the experts in subject domain who aren't knowing programming languages.

Keywords

• Аrtificial neural network
• radical basis functions network (RBF-network)
• sample classifications
• clustering algorithm
• tabular modeling