| The personalized recommendation of agricultural planting technology is a technology that uses a recommendation algorithm to recommend the required planting technology for agricultural users,reduces information overload,and improves production efficiency.It is a vital part in the agricultural information service.In order to improve the quality of planting technology recommendation services,this thesis combines the characteristics of agricultural planting scenes and uses personalized recommendations in the field of data mining to study agricultural users and agricultural planting technologies from three aspects: user characteristic model,personalized recommendation algorithms,and recommendation systems.This study is of reference significance for recommending the required planting technology to farmers.(1)The construction of static and dynamic two-dimensional agricultural user characteristics.Aiming at the lack of behavioral information,Chinese word segmentation was adopted for the basic information of farmers.The static characteristics of farmers are constructed.In response to changes in crops grown by farmers,Multinomial NB classification method is used to establish a model for updating farmers static characteristics.Based on the behavior information,the TF-IDF is used to get the characteristic words and weights that represent planting technology.According to the number of different behaviors of farmers,the weights of different behavior types are obtained by the calculation of information entropy.Combining the time decay function and the number of behaviors to obtain and update user dynamic characteristics tags,an adaptive model of farmers dynamic characteristics is established.Experiments show that the recommendation precision rate based on user characteristics reaches 72.3%.(2)The algorithm design of agricultural planting technology recommendation.Aiming at single consideration factor and low recommendation precision rate,the two similarities are merged to find the nearest user to get Top-N.The Jaccard is used to calculate the user characteristic similarity.The modified cosine is used to calculate the user score similarity.Aiming at the cold start problem,the recommendation based on user characteristics is used to find neighbor users to generate recommendations,which solves the users cold start problem.For matrix sparseness,the Slope One is used to predict unscored items for filling,which alleviates this problem.For poor scalability,K-means clustering is used to find the most similar groups and to screen for strong dependence on geographical factors,which improves scalability.Experimental show that compared with the other two,the precision is improved by1.50% and 6.78%,the recall is improved by 1.03% and 9.02%,and the F1 is 70.63%.(3)Design and implementation of a personalized recommendation system for planting technology.For the recommendation,the agricultural planting technology recommendation algorithm is embedded into the system,and the system is built using the Django framework to show the recommendation results to users.Aiming at the problem of system user characteristic representation,relying on the agricultural user characteristic model,ECharts is used to visually display the constructed and updated user characteristics.The system is easy to promote and can help farmer to obtain information. |
