Search

At present, the construction of digital government has entered a deep water area. The government big data platform, as a data base, supports various government information applications. The security and compliance of its private data has been widely concerned by the industry. Federated learning is an important method to effectively solve data silos, and the application of government big data platforms based on federated learning has high research value. Firstly, the current status of government big data platforms and its federated learning application were introduced. Then this paper analyzed three major management challenges involved in the collection, classification and grading and sharing of privacy data on government big data platforms. Further, the problem-solving methods of federated learning based recommendation algorithms and privacy intersection techniques were explored. Finally, summaries and prospects were made for the future application of privacy data on government big data platforms. [ABSTRACT FROM AUTHOR]

Network representation learning can effectively solve the problem of data sparsity in recommendation. In this paper, LINE and DeepWalk in network representation learning were improved, and a hybrid recommendation algorithm was proposed to be applied to movie recommendation scene. The new algorithm generates three low dimensional feature vectors by learning user preference feature, user aversion feature and similar user feature. Three low dimensional feature vectors are linearly combined to form a user representation vector, and cosine similarity is used as the similarity index to recommend the movies associated with similar users to target users. Experimental results show that, compared with the suboptimal algorithm, the accuracy and Fl index of the proposed algorithm are improved by 12% and 7% respectively on MovieLens dataset, and 16% and 18% respectively on MovieTweetings dataset. The recommendation algorithm based on multi-dimensional feature representation learning proposed in this paper has significant advantages in movie recommendation scenes. [ABSTRACT FROM AUTHOR]

Because heterogeneous information network（HIN） contains rich network structure and semantic information, recommendation systems often use HIN for recommendation. However, the current researches of recommender systems is mainly based on indirect information provided by meta-paths for recommendation, but these researches don’t make full use of direct interactive information. To make full use of this information, this paper proposed a ternary interaction model（AMMRec） that incorporated attention mechanisms and heterogeneous information network meta-paths. This method firstly used the implicit feedback matrix to construct user similarity matrix and item similarity matrix and used the representation learning method of the HIN to obtain the corresponding feature vector embeddings in HIN. Then it used the the attention mechanism to modify the embeddings and designed attention neural network to fuse representation vectors of different meta-paths. Finally it concatenated user embeddings and meta-path embeddings and item embeddings, and generated recommendation results through fully connected neural network. The experimental results on real datasets show that AMMRec improves the recommendation accuracy by up to 9.5%. In addition, AMMRec has good interpretability for the recommendation results. [ABSTRACT FROM AUTHOR]

Focusing on problems of the accuracy recommendation system that the recommended commodity type is relatively single, and commodities are mostly popular goods and lack of freshness, the novelty recommendation is gradually gaining attention. However, current researches don't combine item features when designing algorithms, which make them unable to distinguish and select items with higher novelty for different users. In order to improve the performance of the recommendation system, this paper improved the method based on random walk and designed a new recommendation algorithm by fusing novelty features. This algorithm further analyzed features of items and gave the formal definition of the novelty from perspectives of user interest expansion and prediction. This paper analyzed user demands, constructed new transition probability, generated personalized recommendation lists and improved the novelty of the lists. The experimental results show that the proposed algorithm has less influence on the accuracy than existing methods and has significant improvement on novelty indexes. It concludes that by fusing novel features, this system can improve the recommendation contents effectively while taking into account the accuracy. [ABSTRACT FROM AUTHOR]

When modeling users' preferences, current researches of group recommendation usually ignored the mutual influence between group preference and individual preference and the problem of modeling initialization. To address these issues, this paper proposed a new group recommendation model called ranking based hybrid deep tensor factorization model, namely R-HDTF model. First of all, this paper developed a hybrid deep learning-based initialization method, which utilized deep denoising autoencoder to pretrain the initial values of the parameters for the R-HDTF model. Then, it proposed a pairwise tensor factorization model to capture the correlation among group, individual and item. Finally, it used the Bayesian personalized ranking( BPR) metric to optimize the loss objective function of tensor factorization and learn the parameters of the proposed model. Experimental results on real data sets show that the performance of the proposed algorithm outperforms traditional group recommendation algorithm. [ABSTRACT FROM AUTHOR]

In order to solve the problem that the recommendation model cannot mine the diversity of user interest and capture the order information between the sequence of user behavior, and the interaction occurs at the element level rather than between the feature vectors, etc., based on multiple attention mechanism and location information, this paper proposed the x DMAL FM. Firstly, it extracted the feature depth from different subspaces by using the multi-head attention mechanism, and then location information could capture the sequential relationship between user behavior sequences. Finally, it used three public datasets to conduct comparative experiments and the AUC index to evaluate the results. The experimental results show that the proposed algorithm has better recommendation performance than the xDeepFM model and show the effectiveness and feasibility of the proposed algorithm. [ABSTRACT FROM AUTHOR]

In recent years, as deep learning has achieved good results in many fields, deep learning has also begun to be applied to recommendation systems, such as NFM models and DeepFM models that use deep learning technology to capture high-level feature interactions. However, considering the changes in the external environment and internal perception, the user' s interest should also change dynamically over time, and the combination based on the original features may not necessarily learn effective feature interaction. This paper attempted to build a new model FG_DRFwFm, which could learn the interaction of low-level and high-level features of multiple feature domains and dealt with long-term changes in user interest. It constructed the training features by constructing new features based on the original features and splicing them together, which could better learn effective feature interaction. Finally, the proposed model compared the recommendation effect with multiple advanced CTR algorithms on the MovieLens data set. The experimental results show that the proposed model achieves better results. [ABSTRACT FROM AUTHOR]

In order to solve the problem that users deep interests and preferences cannot excavate in the recommendation. algorithm, which leads to low extraction accuracy and high time complexity of similar users clustering accuracy, this paper proposed a clustering optimization recommendation algorithm (MT-QRPD) based on the evaluation quantization model to optimize differential evolution. Firstly, it used the feature timing of BiGRU network and the strong local feature effectiveness of CNN network to extract the comment depth features, and used the multi-dimensional semantic feature screening of multi-head attention mechanism to mine the comment depth semantic features. Then, through the nonlinear transformation of the multi layer perceptron carried out the multi-feature fusion to achieve accurate quantification. Finally, the algorithm optimized the clustering of similar users with PCA, and completed the project recommendation. The result of experiments shows that the pro posed algorithm can improve the accuracy, time complexity and recommendation performance. [ABSTRACT FROM AUTHOR]

In order to solve the problem of data sparsity in traditional recommendation algorithms, many researchers at home and abroad have proposed corresponding recommendation algorithms. However, most of these algorithms have not achieved good recommendation results in personalized recommendation. Therefore, this paper proposed a recommendation algorithm based on improved filling method and multi-weight similarity. Firstly, the algorithm filled missing values and reduced data dimension by improved filling method, then calculated use r trust degree and user association degree of bipartite graph respectively, and finally used multi-weight factor to combine the two similarities. Based on this, this algorithm obtained neighbor users according to similarity and made recommendation to target users. The experimental results show that the MAE of proposed algorithm is superior to other recommendation methods in the case of sparse data and personalized recommendation. [ABSTRACT FROM AUTHOR]

To solve the problem of lacking user ratings and inaccuracy of user topics model for danmaku video website, this paper proposed a danmaku video recommendation algorithm combing topic model and collaborative filtering ( DRCIT). Firstly, it constructed an implicit rating matrix with user behavior, and obtained the similarity of project rating. Secondly, it established an LDA topic model to generate project topics and calculate the similarity of project topics. Thirdly, it generated a comprehensive project similarity, and calculated the user-project prediction score. Then, through sigmoid penalty, it established the new similarity matrix of project topics and got user-project topic similarity combined with user history behavior. Finally, it took user-project topic similarity as weight to multiply with prediction score to get the final prediction score. The comparison experiments show that the algorithm can get reasonable recommendation results and improve the accuracy of recommendation. [ABSTRACT FROM AUTHOR]

The traditional collaborative filtering recommendation algorithm has the problems of low timing and high computational complexity of similar users with sparse data when there are too many users. According to the above problem, this paper proposed a spectral clustering optimization recommendation algorithm combining the minimum hash signature and temporal model prediction. Firstly, it used the MHS and Levenshtein distance measure to extract similar users from user-item scoring matrix. Then it used AIM-RT prediction algorithm with weight fitting to complete sparse matrix of similar users. Finally,it carried out similar user optimization by combining spectral clustering to find the optimal set of similar users and completed the final recommendation. Experimental analysis and verification show that the proposed recommendation algorithm can improve the overall recommendation performance in terms of computational complexity, scoring prediction accuracy, data missing filling and other aspects. [ABSTRACT FROM AUTHOR]

Based on the traditional personalized recommendation algorithm, this work presented a random walk recommendation algorithm using multi-weight similarity. In order to make up the absence of the influence among social networks, popular items and common scoring items in traditional collaborative filtering algorithms, this paper introduced a gravity formula, which could calculate user similarity in social networks. In addition, this work used weight factors to combine the gravity formula method with the traditional collaborative filtering algorithms. Furthermore, it also added the random walk algorithm for recommendation functions. The experimental results show that the proposed algorithm has better recommendation performance by comparing with other kinds of recommendation algorithms. [ABSTRACT FROM AUTHOR]

Recommender systems can effectively reduce the information overload by recommending items that users may be interested in. Euclidean-embedding-based collaborative filtering methods map users and items to a unified latent space, which is one of the most important methods to build a recommender system. However, traditional Euclidean-embedding-based collaborative filtering methods only consider the low-order interaction between user latent feature vectors and item latent feature vectors, and cannot efficiently model the complex interaction behavior between users and items in the real world. In this paper, we propose a deep-Euclidean-embedding-based collaborative filtering algorithm, which utilizes deep learning technology to learn the high-order and nonlinear interaction function between user latent feature vectors and item latent feature vectors. This can model the complex interaction behavior between users and items. Experimental results on real-world datasets show that our proposed algorithm outperforms traditional collaborative filtering algorithms. [ABSTRACT FROM AUTHOR]

The current trust-based recommendation algorithms did not fully exploit the trust relationship between users and it lacked reasonable trust transitivity rules, which greatly affected the reliability and accuracy of the recommendation algorithm. Aiming at the above problems, this paper combined user rating data with the user's social relationship to build a trust transitivity model, and proposed a recommendation algorithm based on trust transitivity. Firstly, the algorithm used the score data to calculate the implicit direct trust relationship of the users in the trust transitivity model. Secondly, it solved the indirect trust relationship of multiple trust transitivity chains by solving the ordered weighted average operator. Finally, it converge the user's trust and similarity into comprehensive similarity for predictive recommendation. The experimental results show that the proposed algorithm can effectively improve the recommendation quality of the system. [ABSTRACT FROM AUTHOR]

Aiming at the problem of defects in the accuracy of recommendation and data sparse processing for recommendations made by explicit feedback information. This paper introduced implicit feedback information, designed and implemented a multi-dimensional recommendation algorithm( iMCF) based on implicit feedback. The algorithm included three dimensional information of user, project and implicit feedback. For the first two dimensional information, it used the similarity of cloud model to model. For the information of implicit feedback, it mainly used the probabilistic matrix decomposition model to deal with. Afterwards, it balanced the prediction scores obtained from the three dimensions according to the weights, and obtained the final prediction score and made recommendation. The experimental data shows that the algorithm's performance in terms of recall rate and accuracy is significantly improved compared to other algorithms, and it is suitable for big data environments. [ABSTRACT FROM AUTHOR]

Tradition al personalized recommendation algorithm generally suffers from the problem of data sparseness, which affects the accuracy of recommendation. The Slope One algorithm is simple and efficient, but the algorithm is only based on the user - project score matrix to analyze the data, ignoring the type characteristics of the project and the user's preference for the type of the project. In order to solve the above problems, this paper proposed the LR-Slope One algorithm. Firstly, it constructed the user's preference matrix based on user project score matrix and project type information. Secondly, it calculated the weight of each type by linear regression model and optimized the weight by random gradient descending algorithm. Finally, it predicted the score combined by Slope One, filled the scoring matrix, which improved the quality of recommendation. Experimental results show that the proposed algorithm improves the accuracy of recommendation and alleviates data sparseness effectively. [ABSTRACT FROM AUTHOR]

This paper proposed an approach called RQE-EDKL (recommendation quality evaluation based on empirical distribution and KL divergence) to evaluate the recommendation quality based on empirical distribution and KL divergence. QE-EDKL firstly made use of historical user-item data to produce the historical usage probability distribution of items at different quantities. Secondly, it calculated the KL divergence based on the distributions of the historical usage probability and the usage probability of different recommendations. Thirdly, it regarded the recommendation with the minimum KL divergence as with the best quality and is recommended to the user. Experiments on TalkingData App data sets demonstrate that RQE-EDKL can effectively improve the quality of recommended results of collaborative filtering significantly on both accuracy and diversity. [ABSTRACT FROM AUTHOR]

Aiming at the problem of ignores the users interest in the key words and the data sparseness in traditional collaborative filtering algorithm. This paper proposed a collaborative filtering recommendation algorithm combined with the user interest degree clustering, and used user ratings for projects and extracting keywords from item attributes. It proposed a new rating frequency-inverse item frequency algorithm. According to the target users' scoring frequency for a key word and the frequency of the keyword being evaluated by all users, it got users' preferences for keywords, formed user preference matrix, and clustered on the basis of this matrix. Then it used logistic function to get users' interest in projects, cleared user preferences and found similar users of target users in the clusters. Then extracted N items from neighbors' preferences, and recommended users. Experimental results show that the algorithm accuracy rate is always better than the traditional algorithm. It's more accurate to judge the user interest, alleviating the problem of data sparseness, and effectively improves the accuracy and efficiency of recommendation. [ABSTRACT FROM AUTHOR]

There are some shortcomings in user-based collaborative filtering, this paper proposed a new collaborative filtering recommendation algorithms, named user's gravitation based collaborative filtering (UGBCF) recommendation algorithm, it used a new method of similarity measure to improve the user-based collaborative recommendation algorithms. This paper thought that the social tags used by user can reflect user’s preference and how much the preference, so it used those social tags to build user's preference object model. It computed the gravitation between preference objects, viewed the gravitation as the similarity of users. According to the similarity, the neighbor user of the target user could be gotten, and the prediction score of his unselected items could be calculated by aggregated the neighbor users' score. The results of experiment show that UGBCF can provide better recommendation quality than other collaborative filtering recommendation. [ABSTRACT FROM AUTHOR]

In order to solve the problem of cold start and data sparsity for traditional collaborative filtering recommendation algorithm, and the accuracy of similarity measurement, this paper proposed a matrix decomposition recommendation algorithm based on the LDA theme model. Firstly, it used the improved LDA algorithm to output the project-topic distribution, using the perplexity as the modified function of the subject number. Secondly, it calculated the similarity matrix of the project based on the cosine similarity and the KL divergence, combining the obtained similarity matrix with the original scoring training set to output the pre score, and then filled the preliminary score to the training set. Finally, it input the training set to ALS matrix decomposition algorithm to get the recommended results. The experimental results of the MovieLens data set show that the proposed algorithm can get a smaller MAE values than the traditional ALS algorithm under different implicit parameter settings and it greater than traditional recommdation algorithm. The experiment shows that the results of the ALS algorithm are better than other algorithms by integrating the LDA theme model. [ABSTRACT FROM AUTHOR]

This paper proposed a linear regression algorithm to integrate collaborative filtering based on the data sparse influence of the traditional collaborative filtering algorithm. Firstly, it built a similarity matrix between the user and the project based on the user’s rating of the project, as well as the user and the project’s own characteristics. Secondly, based on the similarity matrix, it selected the user and project nearest neighbor set. It predicted the score that the users had graded respectively by the way of collaborative filtering algorithms based on the user and the project. And it would take the difference between predicted scores and the real scores as features to generate new training data, and regarded the new training data as the input of the linear regression model. Finally, according to the training model, it could predict the unknown score, and used the top-N algorithm to generate the recommended list. It conducted the experiment on the MovieLens data set. The experimental result shows that the proposed accuracy of the new algorithm improves compared with the traditional collaborative filtering algorithm. [ABSTRACT FROM AUTHOR]

Aiming at the unreasonable allocation of resource in the recommendation algorithm of bipartite network, and in order to enrich the diversity of recommended results and promote the recommendation of unpopular items, this paper proposed the algorithm for bipartite graph network structure, which used differential path weight to change resource delivery. It changed the first stage of resource transfer rules by user similarity to construct path weights and made the resources flow more towards the user nodes similar to the target users. Through the consideration of constructing path weight of the articles attributes’ similarity, the second stage resources were more likely to flow to the objects which had similar attributes with the target users’ purchased products. Experimental results show that the proposed algorithm improves the combination property of recommendation, and better solves the related problems in recommendation, which outperforms other popular network structure algorithms. [ABSTRACT FROM AUTHOR]

Up to now, researchers developed many recommendation algorithms to solve information overloading. Mass diffusion and heat conduction are two popular algorithms which receive widely attention in recent years. However, there are still some defects for the two algorithms. For example, though the mass diffusion algorithm is of high accuracy, the diversity of the recommendation results is not good. Ileal conduction algorithm is just the opposite, which is high in diversity, but low in accuracy. In order to solve the contradiction, this paper proposed a hybrid algorithm based on mass diffusion and heat conduction. The new algorithm not only mixed the advantages of two traditional algorithms, but also added an adjusting parameter to restrain the excessive recommendation of popular items. The experimental results show that the new algorithm does better in both indicators when proper parameter is set. Meanwhile, the system can reach a better balance between precision and diversity. [ABSTRACT FROM AUTHOR]

The efficiency of traditional user-based collaborative filtering (user-based CF) recommendation algorithm is reduced with data increasing. This paper proposes a recommendation algorithm based on community detection (RACD) in bipartite networks by introducing bipartite network community detection theory into user-based CF recommendation algorithm. Firstly， the user-item rating matrix is mapped into user-item bipartite network. Then， the community information of each user is obtained by using RACD to divide the user-item network. Finally， the items are recommended to the target user according to other users in the same community. Experiments on real-world classic network datasets show that the RACD can effectively improve real-time recommendation efficiency of the recommendation system. [ABSTRACT FROM AUTHOR]

To solve the problem of an obvious rise in error as the decrease of the number of training samples in the recommendation algorithm based on latent factor model (LFM), this paper proposed a new training algorithm based on continuous actions learning automata (CALA). The new algorithm took use of the superior property of continuous learning automata on optimizing parameters in a more random environment. Then the new algorithm insteaded of the original gradient descent algorithm, to perform the singular value decomposition (SVD), which could achieve a smaller error between the reconstructed matrix and the original matrix, and then it reduced the forecast error. To analysis the optimization performance of the new algorithm, this paper made a test based on a real movie rating data set,and compared the forecast result with that of the original algorithm. The results indicate that the CALA-based training algorithm can behave better than the original algorithm in a more random environment and achieve a more accurate forecast effect. [ABSTRACT FROM AUTHOR]

Time series data is a high-dimensional data type in which the attributes change over time,and it reflects the dyna- mic change of user interest. Recommender system based on the time series data is able to improve the efficiency of recommendation, but it is not suitable for large scale data sets,matrix factorization is a commonly used method to reduce the dimension of high dimensional data set. This paper proposed a method to combine the time-series model with matrix factorization. Firstly,this method extracted the characteristics of the original time series data by using the matrix factorization, then it extracted the trend of the feature by the sequential model. Finally,it obtained the prediction results according to the predicted characteristics. The experiments demonstrate that the proposed method performs better on root mean square error (RMSE) and mean average precision ( MAP) than the conventional recommender algorithms, and it is also appropriate for the prediction task of large-scale data. [ABSTRACT FROM AUTHOR]

Due to the development of e-commerce, traditional stand-alone model is difficult to meet the needs of massive data in real-time recommendation. And collaborative filtering-based recommender system has become increasingly evident. Therefore, this paper proposed a distributed recommendation method based on Spark computing model, which the theory was based on spectral clustering and Naive Bayes. In addition, the hybrid method used the increment update schemes to refresh the ratings and improved the precision of the system, without all the re-training model. The experimental results demonstrate that to be compared with traditional stand-alone mode recommendation algorithm, the distributed recommendation algorithm overcomes sparsity and scalability problem to a certain extent and has higher scalability and reduces the response time of the system. [ABSTRACT FROM AUTHOR]

College used book trade is different from the general online shopping. Recommendation engine applied common recommendation algorithm cannot be very good to recommend. In order to make students ’ used book shopping site better to recommend books to users, to improve the accuracy of the results and the efficiency of recommendation, based on the original recommendation module, the paper proposed a suitable mode for shopping website of used books recommendation engine with an optimal recommendation algorithm module. The new algorithm module has the following characteristics ： (1) Based on the items of the recommendation algorithm，books ’ scores and " old and new" factors are taken into consideration to optimize recommendation results ；( 2 ) For the convenience of users purchase, a new packaged recommendation function is added, so users can buy the required books by pushing one key ； (3) According to the fact that the needs of students can be predicted, the recommendation algorithm was designed based on time nodes. [ABSTRACT FROM AUTHOR]

Tag-based recommendation algorithms have become a hot research topic. Existing related research used tags to improve collaborative-filtering recommendation algorithms or content-based recommendation algorithms，few researchers used tags to improve matrix factorization recommendation algorithm which was one kind of more advanced recommendation algorithm. How ever， existing matrix factorization recommendation algorithm mainly used commodities' types to model users' preferences or commodities' characteristics， which would limit its improvement in prediction accuracy. This paper made use of tags to build factor vector， and proposed a novel tag-based matrix factorization recommendation algorithm. Results of experiments on real data sets demonstrate the proposed recommendation algorithm performs much better than traditional type-based matrix factorization recommendation algorithm. [ABSTRACT FROM AUTHOR]

Collaborative filtering-based recommender algorithm have become extremely popular in recent years, due to the development of E-commerce. By the way, it has some limitations such as sparsity, scalability. This paper developed a hybird recommendation method to overcome the limitations of CF. Firstly,the hybrid method used spectral clustering to transform the cluster to segment,and found similar segments. Secondly,the hybrid method applied the extended logistic regression on naive Bayes to build prediction model. Finally,the hybrid method used the increment update schemes to refresh the ratings and improved the precision of the system. The experimental results demonstrate the proposed approach overcomes sparsity and scalability problem to a certain extent,has higher accuracy,lower complexity,scalability than traditional collaborative filtering. [ABSTRACT FROM AUTHOR]

This paper studied the personalized recommendation algorithm for social tags systems, and proposed a new recom-mendation algorithm based on gravitation and random walk(GARW).Aiming at the problem that the present recommendation algorithm lacked of physical explanation and heavy reliance on users'score, this new algorithm introduced into the universal law of gravitation innovatively, also gave the definition of item gravitation and the calculation method, and considered the strength of item gravitation as the similarity of items, then got the correlation graph of items. After that, using random walk method, it spreaded users'preferences on the correlation graph of items, got its'steady-state probability on nodes of the graph which reflec-ted the degree of correlation between users'preference and items. The items with higher steady-state probability were recommen-ded to the user. Experimental results show the new algorithm performs better than the other algorithm compared to. [ABSTRACT FROM AUTHOR]

In modern information society, recommendation systems had already been widely used. Traditional recommendation algorithm didn't take users' emotion tendencies into consideration. To aim at the problem of improving the performance of the recommendation system, this paper proposed an improved recommendation algorithm on the basis of traditional collaborative filtering algorithm. This algorithm combined with emotion analysis. As the user's reviews more directly reflected their real fee-lings, it was feasible to generate the corresponding scores by analyzing the emotional tendencies of user reviews to replace the scores which was given by users. The improved algorithm has certain originality. The experiment shows that this recommendation algorithm improves the performance of the recommendation system, with positive significance for the development of the recommendation system. [ABSTRACT FROM AUTHOR]

A collaborative recommendation algorithm based on KNN didn't take in count the neighbours that blindly followed others, ideas, which made the neighbours useless to predicte the rates of the target users to the unknown items. To address this problem, this paper put forward the contribution factor, thought over the non-common evaluation items, considered the contribution of neighbour recommendation, and selected the neighbours with the traditional similarity. And it recalculated the neighbour prediction weights to the unknown projects to advance the performance of recommendation. The experiments show that this algorithm improves the degree of accuracy. [ABSTRACT FROM AUTHOR]

Similarity used single mark and sparse data lead to inaccurate neighbors and poor recommender quality for the traditional recommendation algorithm. This paper proposed a collaborative filtering algorithm measured by compound multiple factors. The algorithm defined a similarity computing function constrained with visit times, access time, rating, which avoided the problem of similarity measured by single rating and improved the accuracy of similarity compute. To solve the problem of inade-quate data for computing similarity, the algorithm created item's and user trust model based on item category, items visi-ted by target user, users who had visited target item, exploiting sequence and used trust degree instead of similar degree when data was very sparse. The experiment results show that the algorithm can significantly improve the neighbor accuracy and recommendation quality. [ABSTRACT FROM AUTHOR]

Aiming at recommender system's problems that poor performance was caused by compute without considering the mass data and similarities based on history information couldn't reflect the change of user interests, this paper proposed a compute adaptive fast recommendation algorithm meet user interests drift. The algorithm dynamically adjusted window's length of the target users according to the computing resources usage rate such as CPU and allocates computation time in accordance with categories and access popularity of items, it achieved that computing automatically adapts the priority, similarity computation difficulty of items and target users, improved the computation efficiency and response speed. Moreover, the thesis created an interest measure function by access number changing over time to adapt interest drift, which improved the quality of the recommended. The experimental results show that the algorithm has faster speed, more accurate recommender precision and a better user experience. [ABSTRACT FROM AUTHOR]