Research On Cross-border Electronic Commerce Talent Recommendation Model Based On Deep Learning

International trade is playing an increasingly important role in the development of the national economy.Cross-border e-commerce has become an increasingly important part of China ' s cross-border trade,and the training of cross-border e-commerce talents has also caused more and more national Relevant departments pay attention.The country has issued a series of policies to support the training of cross-border e-commerce professionals,but the talent information is huge and redundant,and there is a serious information overload problem.The recommendation system can effectively screen this information.This article mainly studies and develops the following two aspects:(1)Based on the traditional keyword extraction algorithm Text Rank,using co-occurrence words and text mutual information to improve it,reducing the original algorithm's dependence on word frequency,expanding the amount of information in the word map,considering the frequency from a statistical perspective A new keyword extraction algorithm is proposed to extract the self-evaluation text of the rich information-rich talents in the crawled resume information and the job responsibility data in the job information released by the company.These two data texts are longer and are not suitable for direct processing.We need to extract keywords.(2)Construct a hybrid neural network including CNN convolutional neural network and LSTM long-short-term memory network,and then add attention mechanism to make recommendations.First,the original text is processed through the Word2 vec word vector tool.Company position information is input into the convolutional neural network to extract convolutional features.Input the obtained convolutional features and resume information containing time series data of talent work experience into the LSTM network for processing.The network can make good use of the interactive information of the context.By adding the cell state and gate mechanism to the RNN recurrent neural network,it can solve the long-term dependency of the project.Finally,a layer of Attention is added to automatically assign weights to each part,and assign different weights to different parts.The high-impact parts will get high weights to optimize the results.By manually removing the last work experience from the talent resume,the accuracy is judged compared with the work predicted by the model.We have experimented on the data set we constructed by collecting and crawling the job information and resume data published by online companies,and achieved good results.