Research On Product Recommendation Algorithm For Network Collaborative Manufacturing

Nowadays,with the rapid development of internet technology,the manufacturing industry is changing to the direction of "intelligent manufacturing" and "network collaborative manufacturing".In the network collaborative manufacturing platform,the research on product production data integration technology is current research hotspot,which is of great significance to guide production and provide decision-making basis for customized production.It is an important research direction of product data integration technology to combine product and user data and recommend products to users.How to build a reasonable data integration model to recommend products and improve user satisfaction has theoretical significance and practical engineering application value.In this paper,based on the network collaborative manufacturing platform supporting mass customization,aiming at improving the accuracy of product recommendation algorithm,and taking the construction of recommendation model combined with multi-source heterogeneous data as the research content,the product recommendation algorithm is studied.In order to solve the problem of single data type of current recommendation algorithm,a LDA-Word2vec+Ridge Regression+Louvain(L-WRL)product recommendation model is proposed,which combines product attribute data,comment information and social relationship information.In this paper,LDA and word2 vec algorithms are used to extract the joint feature vectors of product attributes and comments.The Louvain algorithm is used to cluster the user's friend relationship data.The ridge regression model is used to train the parameters combined with the joint feature vectors and user rating data.After predicting the user's rating of products,the recommended product list is sorted,and the recommendation accuracy is evaluated.Experimental results show that the proposed model can improve the accuracy of recommendation in different sparse data sets.Using the idea of pipeline hybrid recommendation model,and further introducing commodity image data,a multi-source heterogeneous data recommendation model integrating image information is presented.Firstly,histogram image matching algorithm is used to calculate the similarity value between the products in the top-N list of L-WRL model and the user's favorite products,and the image-based product prediction score is calculated combined with the maximum value of product score.Secondly,DBSCAN density clustering algorithm is used to cluster the image-based prediction score,and different weights are given to the score data in different clusters after obtaining multiple clusters.Finally,the L-WRL model product prediction score and image-based score are combined by adjustable weighting factors to get the final product prediction score and evaluate the recommendation accuracy.Experiments show that the proposed model can further improve the accuracy of product recommendation in different sparse data sets.