DESIGN OF GENERATION SYSTEM BASED ON ARTIFICIAL INTELLIGENCE ALGORITHM. PART 1

Abstract

The development of modern tools and methods of artificial intelligent provides a wide range of functionality for various scientific and applied tasks, and also allows obtaining new results. A question arises about how to apply elements of artificial intelligence to obtain new combinations and connections obtained by generation. As a purpose of the article is considered applying a neural network technology for the formation of new, unique recipes for cocktails with the aim of demonstrating the effectiveness of implementation and verification methods and models of artificial intelligence. To solve generation problems according to ingredient recipes, in the article all stages of the technological process are analyzed from data collection, analysis, preprocessing to the selection of a mathematical algorithm and modeling of the trained system. Each stage of the work contains a separate description of the steps necessary to solve the task. The machine learning algorithm has the ability to process thousands of examples to find a certain pattern of combinations of ingredients. And important attention is paid to the formation of a training date.
 At the stage of data collection, individuals and main problems are considered, which are checked during the work of the parser. The important part is the partial use of encrypted technologies when using Internet resources and reading the DOM tree with the code of the HTML page. The paper shows settings for the automated algorithm of information collection. For the parser development, a tool for rapid project development and effective management as Docker and Docker Compose add-ons were used. Completed the stage with the acquisition of the data set and the data for neural network models has been analyzed. The stage consists in preprocessing, decoding and compositions collected data in relevant tables. During the automated data collection process, information with significant noise and redundant elements is processed. Considerable attention is paid to the process of cleaning and preparation of useful information, because the purity of the data and their completeness in the most cases determine the quality of the mathematical model and the modeling process for finding new laws and combinations in generation.