Unsupervised Internet of Things product recommendation method, system and device and storage medium

Abstract

The invention discloses an unsupervised Internet of Things product recommendation method, system and device and a storage medium, and can be widely applied to the technical field of Internet of Things. An encoder adopted by the invention is located in a neural network model and is sequentially connected with a linear representation structure consisting of an unbiased full-connection layer and a decoder in the neural network model, and the full-connection layer adopts a first matrix with a diagonal element being zero to operate the output of the encoder; the working parameters of the encoder are adjusted in combination with the decoder, so that when the encoder performs dimension reduction on the customer feature matrix, the customer feature matrix can be mapped to a low-dimensional linearly separable space, the accuracy of the clustering result of the feature vectors is improved, and more accurate products can be recommended to customers.

Description

technical field [0001] The invention relates to the technical field of the Internet of Things, in particular to a method, system, device and storage medium for recommending unsupervised Internet of Things products. Background technique [0002] At present, the main process of the generally used product recommendation system for IoT customers is to construct the customer feature matrix, use the linear dimensionality reduction PCA in the dimensionality reduction method to extract the first k principal components, and then use the K in the iterative solution clustering method - The Means algorithm clusters the principal components to perform unsupervised classification of customers and mutual product recommendation for customers within the class. This product recommendation method, because the PCA dimensionality reduction algorithm uses linear operations for dimensionality reduction, and the data corresponding to the customer feature matrix has the characteristics of high compl...

AI Technical Summary

Problems solved by technology

This product recommendation method, because the PCA dimensionality reduction algorithm uses linear operations for dimensionality reduction, and the data corresponding to the customer feature matrix has the characteristics of high complexity and high sparseness, the dimensionality reduction of linear operations may not be able to map the customer feature matrix into a linearly separable space, which may lead to failure to recommend realistic products to customers

Images