The Research And System Design Of E-commerce Recommendation Technology For Agricultural Products

Agricultural e-commerce recommendation system can improve user stickiness and expand market share.It is of great significance to increase the economic benefit of enterprises and improve farmers' income.At present,many scholars carry out research on the recommendation system of agricultural e-commerce,mainly focusing on content-based,rough set,collaborative filtering,mixed recommendation and other recommendation technologies.The existing recommendation system mainly has the following problems: First,the problem of cold start,the lack of behavioral data of new users,and the recommendation system cannot make accurate recommendation.Secondly,there is the problem of sparse data.Most users rarely evaluate the commodities they have purchased,which results in sparse scoring data and failure to provide necessary data for the recommendation system to calculate user similarity,thus resulting in low recommendation efficiency.Due to the backward recommendation algorithm and low adaptability of recommendation model,the rapid development of agricultural e-commerce platform is seriously affected.In order to solve the existing in the agricultural electric business platform of information overload and cold start and data sparseness problem,in this paper,the existing filtering algorithm was improved,and put forward the basic information fusion user and user preference of user similarity calculation method,through balancing factor and punishment factor to change the recommendation accuracy,complete accurate personalized recommendation.At the same time,in order to obtain the user's key attribute set more accurately,this paper proposes an attribute reduction algorithm that strengthens the positive field,which can divide the key attribute more reasonably and accurately,contain more classification information and have stronger generalization ability.On this basis,the agricultural product recommendation model is constructed,including the construction of user preference model,key attribute acquisition,searching for similar users,agricultural product recommendation and other links.Compared with the existing recommendation algorithms in terms of recall rate,accuracy rate and average absolute error,the results show that the user similarity algorithm integrating user basic information and user preference proposed in this paper has higher recall rate and accuracy,minimum average absolute error and high accuracy when the data set is larger,which can better solve the problems of cold start and sparse data.In addition,compared with the traditional positive domain algorithm in terms of reduction time and the number of remaining attributes,the results show that with the increase of the number of samples and conditional attributes,the algorithm proposed in this paper is better than the traditional positive domain algorithm in both aspects,which shows that the complexity of this algorithm is low and the reduction set is more reasonable.The agricultural product recommendation model constructed in this paper is applied to the agricultural product e-commerce recommendation system,which can not only effectively collect information such as users and agricultural product characteristics and find similar users,but also recommend agricultural products that users may be interested in to users through user preference information.All of those meet the needs of agricultural product recommendation and achieve the expected recommendation effect.In addition,combined with the demand analysis,this paper has completed the design of recommendation process,system function module and database,and developed an e-commerce recommendation platform to provide users with agricultural products recommendation and purchase functions.