Research And Implementation Of Multidimensional Recommend System Lead Into Implicit Feedback

With the rapid popularization of the Internet and the diversification of network applications,the era of big data has arrived.The drastic increase of data makes the number of users and items in the recommend system greatly increase.Relatively speaking,the proportion of available scores will become smaller and smaller,and the scoring matrix will be more sparse,which will lead to the result of traditional recommendation algorithm.The actual deviation is larger and the user experience is poor.And due to the arrival of large amount of data,the traditional recommendation algorithm can no longer meet the needs of computing,and the traditional recommendation algorithm will be faced with a new revolution relying on single-node computing.Therefore,to meet the needs of users in the future,multi-node calculations must be used in combination with multidimensional data to improve the previous algorithms,so as to cope with the impact of big data.This study mainly focuses on the data sparsity and scalability problems of collaborative filtering recommendation algorithms.Firstly,a collaborative filtering recommendation algorithm based on two-dimensional cloud model is proposed.According to the trend in recent years,implicit feedback data is introduced,a multidimensional recommendation algorithm that introduces implicit feedback is proposed,and finally a recommendation system is designed based on this algorithm.The concrete results of this research work are as follows:(1)A collaborative filtering recommendation algorithm based on two-dimensional cloud model is proposed.The algorithm introduces cloud model and hadoop cluster on the basis of two-dimension data.By dynamically determining the weight,the scoring weights of the two dimensions of the user and the project are more reasonable,and the prediction score obtained is more accurate.The experimental data show that the algorithm has greatly improved the performance of MAE(average absolute deviation)relative to other algorithms.Compared with other algorithms,the algorithm verifies that the algorithm can adapt to the big data environment.Due to the use of the cloud model and twodimensional data,the data sparsity problem has also been reasonably resolved and the recommended quality has been improved.(2)A multidimensional recommendation algorithm that introduces implicit feedback is proposed.The algorithm makes full use of the advantages of the MapReduce framework in processing big data.At the same time,the user-item scoring matrix is used to calculate the user dimension and project dimension.Through the processing of user interaction data,the desired implicit feedback interest scoring data is obtained.The final prediction score is obtained by the integration of the three dimensional scores,and the recommendation is based on the score value.The experimental data shows that the algorithm's performance in terms of recall rate and accuracy rate has been significantly improved compared to other algorithms,and the larger the amount of data,the better the recommended quality and the better the accuracy.The comparison between stand-alone and cluster verifies that this algorithm is suitable for big data environments.(3)The prototype system is built with the core of multidimensional collaborative filtering recommendation algorithm that introduces implicit feedback.The prototype system uses data of three dimensions: user,project,and implicit feedback.It uses the data of the first two dimensions to effectively integrate the cloud model,and then combines implicit feedback data to measure three dimensions of data through reasonable weights.Effectively combines the MapReduce computing framework to handle big data problems.