Commodity recommendation method based on multi-modal data fusion

Abstract

The invention provides a commodity recommendation method based on multi-modal data fusion, and belongs to the technical field of deep learning. Through ALBERT-TextCNN, the representation capability of word vectors extracted from a customer basic attribute data set is improved, semantic information of different levels of customers is reserved to the maximum extent, and the classification effect of different customer groups is improved; through AlBert-BiLSTM-CRF, the problems that the multi-semantic analysis effect of one word in a text is poor, different contexts of multi-semantic words cannot be processed and other problems cannot be solved by a traditional language processing model are effectively solved, client online and offline transaction data sets are subjected to keywords which better conform to context semantics of a text set, and client tags are constructed in a more targeted manner; similar customers of two label categories are obtained by using a k-dimensional tree method, and recommendation is performed according to the purchase history of the similar customers, so that relatively high recommendation accuracy is realized. According to the method, the accuracy of the prediction data is ensured while the data is efficiently trained.

Description

technical field [0001] The invention belongs to the technical field of deep learning, and relates to a kind of AlBert-TextCNN (A Lite BERT-Sentence Classification-based Convolutional Neural Network), AlBert-BiLSTM-CRF (A Lite BERT-Bidirectional Long Short-Term Memory Network-Conditional Random Field) , Encoder-Decoder (encoder-decoder), and a product recommendation method based on k-dimensional tree nearest neighbor search recommendation. Background technique [0002] The existing customer acquisition methods are basically to acquire customers through various offline promotions, promotional activities, and online advertising. However, there are many pain points in this method. For example, a large number of customer leads cannot be effectively followed up for a long time, resulting in a high churn rate of potential customers; online and offline customer acquisition channels are scattered, the data is not unified, and cannot be effectively managed, resulting in a large amount...

AI Technical Summary

Problems solved by technology

[0007] In order to solve the existing problems such as the difficulty in accurately obtaining target customer groups and the impossibility of precise marketing, the present invention proposes a method based on AlBert-TextCNN, AlBert-BiLSTM-CRF, Encoder-Decoder and k-dimensional tree-based nearest neighbor search Recommended product recommendation method

Images