166results about How to "Solve the sparsity problem" patented technology

The invention relates to a collaborative filtering algorithm, in particular to a collaborative filtering recommendation algorithm based on article sorting and user sorting, which is characterized by including steps: A, article clustering and sorting; B, user clustering and sorting; C, integration of article clustering and user clustering; and D, sorting recommendation. Compared with the prior art, collaborative filtering recommendation algorithm has the advantages that the data clustering is completed by means of modified KMEANS algorithm, the method is simple, and extensibility is improved while the problems of sparsity and cold boot are solved.

The invention belongs to the technical field of data mining and particularly relates to a cluster-based collaborative filtering commodity recommendation method and system. The method comprises obtaining scoring information of users on commodities and type tag information of the commodities through APIs (application program interfaces) of shopping websites; clustering the users through the types of commodities purchased by the users; according to the clustering results, assigning scoring assessment values for default scores in a user-commodity scoring matrix through a scoring assessment formula; calculating the similarity among the commodities in the matrix, predictively scoring the commodities not purchased by the users and recommending the top N commodities with the highest predictive scores to target users. Compared with the prior art, the cluster-based collaborative filtering commodity recommendation method and system has the advantages of solving the problem of data sparsity, reducing the problem of inconsistent scoring scales of various users, enabling the scoring similarity of users of the same type to be the highest and improving the accuracy of default assignment.

The invention provides a clustering collaborative filtering recommendation technology based on a singular value decomposition algorithm. The clustering collaborative filtering recommendation technology based on the singular value decomposition algorithm comprises firstly classifying users by using user attributive character values provided by the clustering collaborative filtering recommendation technology based on the singular value decomposition algorithm, and reducing dimension of a user-commodity grade matrix; improving a singular value decomposition (SVD) algorithm which is frequently used in image processing and natural language processing, and using the improved SVD algorithm in a recommendation system; decomposing a grade matrix in a cluster where users are located, and aggregating the decomposed grade matrix so as to fill predicted scores of non-grade items in the grade matrix, calculating similarity of the users in the same cluster by using the filled grade matrix, calculating final predicted scores of a commodity by applying collaborative filtering technologies based on the users and widely applied in the recommendation system, and carrying out final recommendation. The clustering collaborative filtering recommendation technology based on the singular value decomposition algorithm has the advantages of being capable of improving recommendation efficiency of the recommendation system, solving the problems such as data sparsity of the recommendation system, and meanwhile being capable of improving accuracy rate of recommendation of the recommendation system.

The invention discloses a personalized research direction recommending system and method based on themes. Paper topics read by users and preference of the users for related paper topics can be obtained through the recommending system according to all the papers read by the users and according to the themes of the papers obtained when training is conducted through a theme model training module, therefore, the recommending system can recommend a new research direction for the users to widen the vision of the users. The innovation key of the personalized research direction recommending system and method based on the themes is to construct a three- layer graph model according to the relationship between the users and the papers and the relationship between the papers and the themes, to calculate preference values of the users for the themes according to the three-layer graph model, to obtain a user-theme preference weight matrix, and to calculate similar user set between the users and other users based on the weight matrix. The preference degree of the themes which are not touched by the users is predicted according to the similarity value of the similar users in the similar user set and according to the preference values of the similar users for the themes, and the research direction, namely, the research theme, is recommended for the users according to the prediction result.

The invention discloses a short text characteristic expanding method based on a semantic atlas. The method includes the steps: performing subject modeling by the aid of a training data set of a short text, and extracting subject term distribution; reordering the subject term distribution; building a candidate keyword dictionary and a subject-keyword semantic atlas; calculating comprehensive similarity degree evaluation of candidate keywords and seed keywords based on a link analysis method, and selecting the most similar candidate keywords to finish expanding the short text. Compared with a short text characteristic representation method based on a language model, the method is simple to operate and high in execution efficiency, and semantic correlation information between the keywords is sufficiently used. Compared with a traditional short text characteristic representation method based on a word bag model, the problems of data sparseness and semantic sensitivity are effectively relieved, and the method is independent of external large-scale auxiliary training corpus or a search engine.

The invention discloses a collaborative filtering method based on socialized labels, which comprises the following steps that: (1) first, a tripartite graph is used to build models for three different node, i.e. a user, an article and the socialized label, and random walk algorithm is applied to each user to recommend top-N personalized articles; (2) in order to solve the problem of the sparsity of socialized labels (i.e. articles are always labeled by only a small number of labels) and noise brought by the subjective factors of the user, the invention provides a method for expanding the labeling of the articles with a lasso logistic regression model, i.e. labels related to the semantics are increased for each article, and the labels with noise are removed; and (3) the weight of the labels in the recommendation process is regulated. The collaborative filtering method based on socialized labels organically integrates the semantic information of the socialized labels to the description of the articles, uses the lasso logistic regression model to expand the labels, solves the sparsity and noise problems of the socialized labels so as to greatly improve the precision and the performance of a personalized recommendation system.

The invention relates to a collaborative filtering method for personalized recommendation fusion content and behavior. The method comprises the following steps of, (1) characteristic input, includinga project-attribute matrix representing a project content and a user behavior matrix representing user behaviors; (2) content-based project clustering for calculating the similarity of projects and clustering the projects; (3) score prediction and feature filling including carrying out score prediction on the non-scoring projects, and filling a user-project scoring matrix; (4) behavior-based userclustering including clustering users according to a project clustering result and a user-project scoring matrix; (5) score predication and project recommendation including determining the clusteringcluster where the target users are located, finding a nearest neighbor user set, performing score prediction on the non-scoring projects of the target users, and finally recommending the first N projects with the highest prediction scores to the target users. Compared with the prior art, the collaborative filtering method effectively solves the problems of data sparsity and cold start, and ensureshigh recommendation efficiency.

The invention discloses a knowledge map representation learning method which combines an entity and a relationship description. Text description information of the entity and the relationship is takeninto consideration, two information sources of structured information and the text description information of a triple are well integrated, based on combined extracts of an end-to-end model of neuralnetworks conducted on the entity and the relationship, a balance factor is set to balance the structured information and the text description information, and different score functions are defined according to different prediction objects; a loss function is used for correlating an entity vector and a relationship vector, the loss function is optimized, and when an optimization goal is reached, the vectors of each entity and relationship in the knowledge map and the text description information can be learned. The method solves the sparsity and imbalance of entity and relationship in a knowledge base, accurately and effectively represents the interconnection between the entity and relationship, applies the interconnection to the large-scale knowledge map, and has good practicality.

The invention relates to a webpage training method and system and a webpage prediction method and system. The webpage training method comprises obtaining a prior probability table of classified keywords according to existing data which are associated with the classified keywords; preprocessing a webpage to be trained to obtain a webpage text to be trained; extracting features in the webpage text to be trained according to the prior probability table to obtain an association relation feature vector representation F1 between the webpage to be trained and a specified category; performing model training on the association relation feature vector representation F1 to obtain a classification result of the webpage to be trained. According to the webpage training method and system and the webpage prediction method and system, category systems which are strong in heterogeneity can be simultaneously processed, the large category systems can be processed through less training data, and the problem of data sparseness is largely solved due to the fact that the browsing and search behavior of a user on the whole Internet not just a website is collected.

The invention provides a Bayesian model-based commodity code classification method and system. Aiming at the problems of sparsity and context loss caused by short text features of commodity name information, the method and system adopt a synonym extension technology for word-segmented sparse words, and increase the word synonym degrees through an external search engine technology; and then, a Bayesian model is adopted to calculating condition probabilities between word sequences and code categories of commodity names so as to obtain a Bayesian code classification mode, and a trained model is used for predicting commodity code categories of input commodity names.

The invention discloses a collaborative filtering recommendation algorithm based on a graph convolution attention mechanism. The method comprises the steps of firstly collecting and processing data and dividing a data set, secondly constructing a GACF model, and finally training the model and recommending by predicting an association score between a user and an item. According to the graph convolution attention mechanism collaborative filtering model provided by the invention, firstly, user-project interaction information is mapped to a vector space by using a graph embedding technology, then, the embedding expression of a user-project interaction graph is learned through a graph convolution network, and then, different weights are allocated to neighbor nodes by using an attention mechanism. By aggregating the feature information of the neighbor nodes, the weight between the neighbor nodes can only depend on the feature expression between the nodes, so that the generalization ability of the model is improved, and finally, a plurality of embedded vectors learned by a graph convolution layer are weighted and aggregated to obtain the association score between the user and the project.

The invention discloses a recommending method of a knowledge map based on the location service field, which includes: extracting a location entity and obtaining an entity set as a seed set of the knowledge map, corresponding the seed set with the entities in the knowledge map, forming an entity correspondence table, in a knowledge map triple in that knowledge map, using Word2Vec model to embed vocabulary into n-dimensional space, generating corresponding vectors, obtaining a position or domain entity vector set E and a relation vector set R, translating the entity vector set E and the relationvector set R by using a TransE algorithm, and obtaining a triple vector set capable of quickly calculating semantic similarity between entities; according to the location or domain entity vector setE, calculating respectively the semantic similarity simA, B (A, B) between the searching locations or domains to generate the semantic similarity matrix of the tourism location, using Semantic Similarity Matrix for Top-K Recommendation List, clustering the recommendation list according to machine learning clustering algorithm, and recommending the clustering result to the user. The method has highprecision and can solve the problems of cold start and sparsity.

The invention discloses a personalized POI recommendation method based on multi-influence embedment. By conducting associative embedded learning on seven bipartite graphs (a user-user graph, a user-time frame graph, a POI-time frame graph, a POI-region hierarchy graph, a POI-category hierarchy graph, a user-gender graph and a user-POI graph) and a check-in sequence, influences on the aspects of social contact, time, geography, semantics, user gender, user preference and sequence are integrated; meanwhile, certain expansibility is achieved, influences on other aspects can be conveniently integrated, so that the problems of data sparsity and cold start are effectively solved, and high-quality POI recommendation is provided for users.

A short text topic model mining method based on a word network to extend characteristics comprises a weighted word network construction step, a short text characteristics extending step, and a topic mining step. The weighted word network construction step comprises preprocessing a text, performing Chinese words segmentation on the text in a short text corpus, and deleting stop words; establishing a weighted word network from a document after the Chinese words segmentation is performed, wherein nodes in the weighted word network are words, each edge between the nodes is cooccurrence relation of two words in the same document, and the weight of the edge is the cooccurrence time of the two words in the whole corpus; and ending. The short text characteristics extending step comprises using the word nodes included by each short text after the Chinese words segmentation is performed as a community of the established weighted word network. According to the short text characteristics sparsity solution method based on word network community module degree, the problem that the effect of applying an LDA topic model to the short text is poor is solved. Accuracy of a short text topic model is increased.

The invention provides a recommending method for mining depth user similarity based on interactive sequence data. The recommending method comprises the following steps: pre-processing the original recorded data of the interaction between the user and the article to obtain interactive sequence data of the user and the article, and generating the user according to the interactive sequence data; themethod comprises the steps of: mining the depth user similarity based on the interactive sequence data; mining the depth user similarity; mining the depth user similarity based on the interactive sequence data. Item Interaction Matrix and Users-Gram matrix; In accordance with that user -Item interaction matrix and the user-gram matrix constructs a recommendation model and trains the recommendationmodel; Based on the trained recommendation model, each user's preference for all items is calculated, and all items are sorted from high to low according to the preference value. The items with the highest preference value among the remaining items are recommended to the corresponding users as the result of personalized recommendation by removing the items that the user has interacted with. The invention solves the problem of data sparseness and improves the recommendation accuracy by applying the preference of the user to the article and the similarity among the users to the recommendation method.

The invention discloses a GCN-based text classification method. The method comprises the steps of obtaining a text classification corpus set, wherein the corpus set comprises a plurality of samples, and each sample comprises a title and a chapter; preprocessing the corpus set, and dividing the preprocessed corpus set into a training set, a verification set and a test set; processing the chapters through space, and extracting a graph relationship between words; embedding each word into a low-dimensional real value vector space of the matrix according to the graph relationship; constructing a bidirectional LSTM according to the vector representation of the word, and obtaining sentence representation; reconstructing sentence representation based on a self-attention mechanism, inputting the sentence representation into a GCN neural network, and training a semantic classification model; inputting the text word vector of the verification set into the model, and recording and storing model parameters when the effect on the verification set is optimal; and testing the test set based on the optimal model obtained by the verification set to obtain a classification result. According to the method, a more accurate class result is finally obtained by applying LSTM and GCN in combination with an attention mechanism.

The invention belongs to the technical field of big data and data mining and recommendation systems, and particularly relates to a system of mobile-phone application recommendation of multi-label classification and a method thereof. The system of the invention includes five modules of a data acquisition module, a data processing module, a data storage module, a service logic module and a display module. A multi-label classification algorithm based on a random walk algorithm is provided in the invention, and multi-label data are mapped into a multi-label random walk graph. A multi-label randomwalk graph series is established when unclassified data are input; and then each node in the graph series is randomly walked to obtain probability distribution of traversing each vertex, and the probability distribution of the vertex is transformed into probability distribution of each label. According to the system and the method, a problem of user interest diversity recommendation and a problemof recommendation operation complexity increasing brought by constant user interest changes are solved, recommendation technology which is more flexible than traditional recommendation technology is obtained, and recommendation quality is improved.

The invention discloses a method for completing continuous missing data of dam deformation monitoring, which solves the problem of completing continuous missing data of dam deformation monitoring. Firstly, the continuity loss of deformation monitoring data is pretreated. Secondly, from the angles of global space, global time, local space, local time and semantics, the spatio-temporal similarity and functional similarity of deformation monitoring points are calculated respectively, and the missing data in deformation monitoring data are interpolated globally and locally in space, time and semantics. Finally, a depth neural network model is constructed to complete the continuous missing data in dam deformation monitoring. The preliminary results of the missing data are used as input, and thenonlinear fusion is realized by using the depth neural network representation ability.

The invention discloses an order-preserving submatrix (OPSM) and frequent sequence mining based emotion classification method for e-commerce comments. The method comprises the following steps: (1) performing pretreatment and Chinese word segmentation on the e-commerce comments; calculating to obtain a TF-IDF weight vector of synonyms; and then mining a local mode in the weight vector through a biclustering algorithm based on OPSM; (3) mining classification frequent phrase characteristics through an improved PrefixSpan algorithm, and meanwhile, improving the capacity for distinguishing emotion tendency by the frequent phrases through limitation such as word intervals; and (4) converting the characteristics mined in steps (2) and (3) into a 0/ 1 vector to be used as an input of a classifier, and thus obtaining the emotion classification result of the e-commerce comments. With the adoption of the method, the emotion classification characteristics of the e-commerce comments can be accurately mined, so that potential customers can know the goods evaluation information before buying, and meanwhile, the businessman can fully know the suggestions of the customers and accordingly improve the service quality.

The invention discloses a collaborative filtering-based content recommendation system and method. The system comprises a data preprocessing module, an algorithm mixing module, and a result generation module, wherein the algorithm mixing module further comprises an algorithm selection unit, a weighting similarity-based collaborative recommendation algorithm unit, a balance score prediction mechanism-based collaborative recommendation algorithm unit, a score filling-based mixed recommendation algorithm unit, and a collaborative recommendation algorithm unit using score time characteristics; and the algorithm mixing module inputs preprocessed data to the weighting similarity-based collaborative recommendation algorithm unit, the balance score prediction mechanism-based collaborative recommendation algorithm unit, the score filling-based mixed recommendation algorithm unit, and the collaborative recommendation algorithm unit using the score time characteristics, and outputs an algorithm result to the result generation module. According to the system and the method, the sparsity problem and the concept drift problem in the recommendation system are better solved.

The invention relates to a continuous interest point recommendation method based on check-in time interval mode, belonging to the recommendation system field. According to the check-in data of each user, synthesizing personalized preference, geographic distance preference and check-in time interval preference to form the user's comprehensive preference for the next interest point to be visited, and adopting the third-order tensor model to model the continuous check-in behavior; a probabilistic model is constructed to learn the comprehensive preference degree of the user to the points of interest by taking the check-in interval preference as a potential variable. In the parameter learning phase, the expectation maximization algorithm is designed to optimize the parameters of the probabilitymodel, and finally the task of recommending the next point of interest for users to visit is realized. The tensor/matrix factorization algorithm is used to compensate the missing information in the tensor and matrix. Compared with the prior art, the method of the invention effectively solves the problem that the method of the invention effectively solves the sparsity problem of the user-point ofinterest sign-in matrix, and provides the user with an accurate and efficient continuous interest point recommendation service.

The invention relates to a personalized recommendation method based on dynamic neighboring point spectral clustering. A user-store bipartite network is set up according to user sign-in information; the user-store bipartite network is projected to a user-user one-aside network and a store-store one-aside network, and a node2vec algorithm is used for projecting the two weighted one-aside networks totwo different vector spaces; the spectral clustering algorithm based on dynamic neighboring points is called to cluster user vectors and store vectors obtained above to obtain multiple user clustersand multiple store clusters; the sign-in information existing between single users is converted into a cluster network among the user clusters and the store clusters; a K-means algorithm is used for dividing the one-dimension vector into two classes, the store clusters in the class with the larger sign-in average number are recommended to the user clusters; personalized recommendation is conductedaccording to each user cluster and the recommended store clusters. The accuracy of the recommendation method is improved effectively.

The invention discloses an intelligent recommendation system based on big data, and belongs to the data processing technology field. The system comprises a client used for collecting the log information and the real-time operation attributes of a user, writing into a server and displaying the recommendation result information fed back by the server; the server used for collecting the user log information and the real-time operation attributes, adopting a deep learning technology for data analysis and establishing a user interest model; and a cloud big database which is used for screening out the products to be recommended in real time by utilizing a recommendation algorithm according to the user interest model and returning the products to be recommended to the client for display through the server, wherein the client is connected with the server, and the server is connected with the cloud big database. According to the present invention, the problems that a traditional recommendationsystem is inaccurate in recommendation information, does not have real-time performance and is greatly limited by the environment, are solved.

The invention relates to an LBSN supernetwork link prediction method based on time-space relationship and belongs to the field of data mining. The method comprises the following steps: S1, acquiring adata source; S2, constructing a supernetwork model; S3, defining and quantizing a supernetwork edge weight value; and S4, based on the model, firstly constructing multiple types of weighting hyperedge structures, mining different semantic relations among users by virtue of different structures, finally training model parameters by virtue of a gradient descent method, and then predicting link relations in network. The method provided by the invention has the advantages that multiple incidence relations among nodes can be effectively mined, the sparsity problem in the network can be solved, anti-noise property and stability of the model also can be improved, and predictive accuracy of the model is also greatly improved.

The invention belongs to the technical field of big data and a data mining recommendation system, and particularly relates to a friend recommendation system and method based on community detection. The system comprises five modules, including a data acquisition module, a data processing module, a data storage module, a business logic module and a display module. The invention also puts forward a calculation method based on community detection. The method comprises the following steps that: according to the signed geographic position information and the common friend number of a user, calculating a user distance similarity and an acquaintance degree; adding into an LMF (Latent Factor Model) algorithm to carry out community division; in a generated friend community, according to user discrete sign in time, adopting a kernel density estimation algorithm to calculate the 24-h integral sign in probability density of the user; and then, calculating the site similarity of the user sign in site, and establishing a user space-time similarity implementation model. By use of the system, the problem that recommendation operation complexity is increased due to a user interest diversity recommendation problem and constantly changing user interests can be solved. Compared with the traditional recommendation technology, the system is characterized in that the technology is more flexibly recommended, and recommendation quality is improved.