Research And Implementation Of Multi-Source Fusion Recommendation System For Agricultural Big Data

In recent years,the country has vigorously promoted the development of smart agriculture to achieve intelligent and precise agricultural production,which has become a hot research topic in smart agriculture.As an important component of smart agriculture,intelligent agricultural recommendation systems obtain beneficial information for agricultural production from agricultural production data and recommend agricultural products or information,thereby effectively guiding agricultural production and decision-making,improving agricultural production efficiency and yield,making agricultural production more intelligent and precise,and has received extensive attention and research from more and more scholars.In response to the problems of single data source and low accuracy of recommendation results in existing agricultural recommendation systems,this thesis designs and implements a multi-source fusion recommendation system for agricultural big data based on the analysis of the characteristics of agricultural big data.The system first uses the regional density-weighted fuzzy clustering technology to reduce the complexity and dimensionality of multi-source agricultural data,and then uses the weighted matrix factorization recommendation technology based on the environmental trust model to improve the accuracy of the recommendation system.The main work of this thesis is as follows:(1)Based on the analysis and research of the characteristics of agricultural big data,a multi-source data fusion technology based on regional density-weighted fuzzy clustering is proposed.Firstly,for the problems of data missing,low accuracy,and high dimensionality of multi-source agricultural data,a multi-source filter is used to preprocess agricultural data,and the quality of raw data is improved through methods such as data accuracy and completeness verification,data filling,and data dimensionality reduction.Then,according to the regional characteristics of agricultural data,the regional density-weighted fuzzy clustering algorithm is used to analyze and cluster multi-source data,reducing the dependency complexity and data scale of multi-source data.Finally,through simulation experiments and result analysis,it is shown that the clustering algorithm in this thesis has high partition accuracy,good compactness,and feature cohesion,which can better meet the needs of the recommendation system.(2)In response to the problems of low accuracy of traditional agricultural recommendation models and poor user satisfaction with recommendation results,a weighted matrix factorization recommendation technology based on the environmental trust model is proposed.This technology first realizes the latent semantic model based on environmental trust by constructing the environmental trust matrix that reflects the hidden feature relationship between agricultural environment and agricultural products,and then weights the environmental feature influence factors and preference feature extraction through the weighted matrix factorization algorithm,eliminating the impact of some special factors on the predicted scoring of recommendation results,thereby improving the accuracy of the recommendation system.Finally,through simulation experiments and result analysis,it is shown that compared with traditional recommendation algorithms,the recommendation algorithm in this thesis has higher accuracy and stability in actual agricultural scenarios.(3)A multi-source fusion recommendation system for agricultural big data is designed and implemented.The system is composed of data source layer,data analysis layer and application layer.In the data source layer,it realizes the collection of multi-source data and the unified format storage of big data in the recommendation system.In the data analysis layer,the multisource data fusion algorithm and multi-source fusion recommendation algorithm are implemented,and the recommendation results are provided to the application layer.In the system application layer,the system implements core modules such as crop recommendation and farming recommendation,as well as auxiliary modules such as basic data management and data query,and provides a friendly WEB visualization interface to display the results.