Research On Agricultural Books Recommendation Algorithm Based On Deep Learning

As an important carrier for spreading agricultural knowledge and promoting agricultural technological achievements,agricultural books were deeply loved by the majority of farmers and readers interested in agriculture.However,due to the degree of these readers informatization is not high,agricultural practitioners in our country have insufficient accurate retrieval ability how to catch the key information in the current large amount of book information,so it is difficult to select the right books and documents efficiently in the massive information.Recommendation system can save time and energy for users in querying relevant information,and provided them with guidance and assistance.Deep learning is stronger than traditional models in terms of its expressive ability.It has become the most popular research direction in the last ten years.However,there are few recommendation algorithms for the field of agricultural documents and books,and there are few agricultural book recommendation methods based on deep learning.This paper mainly studied the recommendation algorithm based on deep learning to improve the accuracy and efficiency of the recommendation for agricultural books.The data in this thesis comes from the real data with a university library past five years.The data includes the information of agricultural books,the reader information of agricultural books and the borrowing records of agricultural books.At the same time,the textbook was introduced into the historical records of readers to alleviate the cold start problem of new users when new readers lack historical behavior.This thesis contained following work.Firstly,for the rich text features of books,this paper used natural language technology to propose an improved model based on BERT to generate vector representation of semantic features such as the book title and the abstract for subsequent recommendation.At the same time,in order to fully tap readers’ interests,this paper used the Bi-LSTM to model readers’ borrowing records,analyzed readers’ borrowing preferences,and finally introduced the Attention to effectively interact with readers and dynamically mine readers’ reading interests.Secondly,in order to study the effect of the recommended model proposed in this paper,multiple groups of controlled experiments were designed.Both the traditional recommendation algorithm and the deep learning model were selected for the comparative experiment.In the summary,analysis showed the result of experimental data of university library,the new model proposed in this paper has better recommendation effect than other models in agricultural book recommendation.In the summary,The results of this paper provided specific solutions for the actual application scenario of agricultural informatization,the research of this paper have certain practical value.