Features of building recommendation systems based on neural network technology using multithreading

Abstract

The article is devoted to the creation of a recommendation system for tourists regarding hotels using a neural network based on a multi- layer perceptron. The work uses the mechanism of parallelization of the training sample of the neural network. To check the quality of the provided recommendations, the average absolute and root mean square errors, accuracy and completeness were used. The results of the experiments showed that when analyzing 10 html pages with descriptions of hotels, the metrics of root mean square error and accuracy gave the best results at 500,000 epochs of neural network training when using 8 processors.