An intelligent business location method based on neural collaborative filtering

Abstract

The invention provides an intelligent business location selection method based on neural collaborative filtering. By collecting restaurant data and geographic information data, the interaction matrixdata of restaurant category and address block is obtained, and the restaurant category preference is simulated by singular value decomposition (SVD), multi-layer perceptron (MLP), linear and nonlinear, respectively, and the two preference values are connected to the final predicted value, which is used to recommend the address list of restaurant category. (Neural collaborative filtering is used topredict missing values in the interaction matrix from which to recommend address lists for restaurant categories.) The invention utilizes the following principle: the missing value in the interactionmatrix can be predicted by the neural collaborative filtering, and then the predicted value is recommended to the restaurant category in some way. It can effectively solve the problem of matrix factorization linear simulation interaction, avoid the cold start problem, and add a multi-layer perceptron this depth learning method, can learn any function from the data, increase the accuracy of the address recommendation.

Description

technical field [0001] The present invention relates to the field of business location selection and recommendation system based on deep learning, in particular to a method of business address recommendation based on neural collaborative filtering. Background technique [0002] Big data-driven business location selection is one of the innovative applications in the new retail era. At present, there have been some studies on commercial location selection, most of which are based on multiple heterogeneous information such as geographic information services and POI data to select the best location. Specifically, they first extract information from multivariate data to obtain features in various aspects: geographic features (density, competitiveness, etc.), mobility features (regional popularity, inflow, etc.), and then use these features for linear regression analysis , Matrix Factorization (MF) or transfer learning to obtain the optimal position. Although these methods can p...

AI Technical Summary

Problems solved by technology

If the feature extraction is not comprehensive, good results may not be obtained; in addition, both linear regression and MF methods are linear models. If there is a certain relationship between features, this model cannot express these features well.

Images