Data Optimization for Industrial IoT-Based Recommendation Systems

Abstract

The most common problems that arise when working with big data for intelligent production are analyzed in the article. The work of recommendation systems for finding the most relevant user information was considered. The features of the singular-value decomposition (SVD) and Funk SVD algorithms for reducing the dimensionality of data and providing quick recommendations were determined. An improvement of the Funk SVD algorithm using a smaller required amount of user data for analysis was proposed. According to the results of the experiments, the proposed modification improves the speed of data processing on average by 50–70% depending on the number of users and allows spending fewer computing resources. As follows, recommendations to users are provided in a shorter period and are more relevant. The faster calculation of modified Funk SVD to exchange the optimal parameters between nodes was proposed. It was determined that execution time can be reduced on average by 75% for using ten nodes exchanging the optimal decomposition parameter compared to using one. Using Spark technology for faster calculation on average by 20% compared to Hadoop was proposed. The architecture of the IIoT system was proposed, which uses a modified Funk SVD algorithm to optimize data on edge devices and monitors the effectiveness of providing recommendations using control centers and cloud resources.

MDBF: Meta-Path-Based Depth and Breadth Feature Fusion for Recommendation in Heterogeneous Network

“…Recommendation systems can help users find items they are interested in among groups of many items [1]. Today, with the development of the Internet, recommendation systems have become more and more important [2].…”