130 results about "Preference vector" patented technology

An information processing apparatus selects a proper content that well matches preference of a user and recommends it. A matrix calculator acquires M (one or more) feature vectors CCV whose elements are given by weight values assigned to a total of N (two or more) pieces of content meta information and context information. The matrix calculator produces a matrix CCM whose columns are given by the M feature vectors CCV and converts it into an approximate matrix CCM* by modifying the weight values of the respective elements of the M feature vectors CCV such that correlations of elements among the M feature vectors CCV are emphasized. Based on the approximate matrix CCM*, a user preference vector (UPV) generator produces a user preference vector UPV*. A matching unit calculates similarity between the user preference vector UPV* and a feature vector CCV produced from new content meta information or context information.

The invention discloses a generating system and method of user' s interest preference vector, which comprises the following parts: user enroll information storage center, click interactive-journal storage center, topic semantic extracting mode to analyze the interactive journal within certain time in the click interactive-journal storage center to obtain topic classifying mark vector according to the semantic strength sequence, interest preference counting mode to acquire and analyze the static enroll information of user from user enroll information storage center and topic classifying mark vector from the topic semantic extracting mode to obtain the interest preference vector of user. The invention can push information or supply service according to the interest preference vector selectively, which projects the information and service of the user' s interest.

The invention discloses a social network friend recommendation method based on community discovery, belongs to the fields of data mining, social networks and the like, and aims at solving the problem that the social network characteristic is not taken into account in individual recommendation. The method comprises the following steps: acquiring user history data in the social network, modeling hobbies and interests of users to obtain a preference vector set of all users, clustering the users according to the preference vectors representing the hobbies and the interests of the users and the preference vectors representing the hobbies and the interests of friends of the users in the user preference vector set, finding an overlap area of the hobbies and the interests of the users and the friends in the social network, acquiring an initial target user list of friends to be recommended according to the overlap area of the hobbies and the interests of the users and the friends, filtering and sequencing the obtained initial target user list of friends to be recommended, thereby obtaining a final list of friends to be recommended. As friends are recommended according to the hobbies and the interests and the friend relationship of the users, the method is more applicable to social networks.

A system for determining preference, including a client device with feedback controls, a server and addressable URIs; the device and server in communication over an electronic network and the URIs' content retrievable over the network; the server automatically receiving and storing ratings, tracking URIs, using ratings to create a preference model for URIs, and using the preference model to suggest URIs; the stored ratings include a record having a rated item URI, a rater having a unique identification, a rating value provided by the rater, and at least one metadatum for creating subsets of ratings. Also, a method for automatically creating a chimeric preference vector, the method steps including identifying a multiplicity of datasets of rated items; automatically combining the datasets to form a combined dataset; automatically identifying ratings collisions; treating ratings collisions to form a data subset; and generating a chimeric preference vector based on the data subset.

The application relates to a personalized data searching method and device. The device comprises the following steps: searching a data object according to a query word in a search request of a current user; determining a first behavior characteristic of a historic user in the search access process utilizing the query word according to a historic behavior log, and generating an intention vector of the query word; counting a second behavior characteristic of each user group to the data object according to a user attribute, and generating a preference vector of the user group; computing the similarity of the intention vector of the query word and the preference vector of each user group; using the user group corresponding to the condition that the similarity is greater than a set threshold value as a reference group for determining the intention preference of the current user; adjusting the sorting of the data objects searched by the current user through the query word through adoption of the historic behavior characteristics of the reference group. Thus the performance of a search platform is improved and promoted, the accuracy of a search result output to the user is improved, and the result, which is the most reasonable and the best for the search intention, is output for the user.

The invention provides a session sequence recommendation method and system based on a graph convolutional neural network, a directed graph is constructed for each session sequence, and for each directed graph, the directed graph is input to the graph convolutional neural network to obtain the hidden representation vectors of all nodes. Based on the implicit representation vectors of the obtained nodes, a soft attention mechanism network is used for generating a global preference vector and a local click preference vector, wherein the global preference vector and the local click preference vector are both composed of implicit representation vectors of the nodes, and then each session sequence is represented as a combination of the global preference vector and the local click preference vector of the user in the session. And for each session sequence, the probability that each item becomes a next click is predicted according to the combination of the calculated global preference vector and the local click preference vector of the user in the session. Some noises in an original vector space are removed by introducing global and local implicit vector representations, and a more accurate prediction effect is obtained.

The embodiment of the invention provides a product recommendation method and device. The method is used for determining whether to recommend a to-be-recommended product to a target user. The method comprises the steps that multi-field information associated with the target user is acquired, wherein the information comprises purchase data of the target user in the product field of the to-be-recommended product and purchase data in other product fields; a user feature matrix of the target user is constructed according to the multi-field information; for one to-be-recommended product, a user feature matrix of multiple users who purchase the to-be-recommended product is acquired, and a product feature matrix of the to-be-recommended product is obtained based on feature values in the matrix; the user feature matrix and the product feature matrix are input into a machine learning model to obtain user preference vectors and product preference vectors; a selection assessment value between theto-be-recommended product and the target user is obtained according to the user preference vectors and the product preference vectors; and when the selection assessment value is greater than a predetermined recommendation threshold, it is determined that the to-be-recommended product is recommended to the target user.

The invention discloses a job recommending method, and belongs to the technical field of recommending systems. The job recommending method has the advantages that the Matthew effect is avoided, the problem of cold start is solved, and the populations are well utilized to realize personalized recommending. The job recommending method comprises the following steps of obtaining user data and job data; establishing a user preference vector space model and a job vector space model; according to the user preference vector space model and the job vector space model, calculating multi-domain scoring values based on contents, obtaining first scoring values of jobs, and sequencing, so as to obtain a job set; when one job is submitted and belongs to the job set, calculating the scoring valves of the corresponding job based on the similarity of user background information according to the user preference vector space model and the job data, and obtaining second scoring valves of the corresponding job; according to the first scoring valves and the second scoring valves, obtaining the mixed scoring valves of the corresponding job, and sequencing, so as to obtain a recommending list.

The invention discloses a method for a personalized recommendation of a Web text. The method comprises the following steps: performing feature extraction on a plurality of kinds of Web texts generated before a certain time t so as to obtain a feature matrix E of a Web text set, and then performing cluster so as to obtain n categories; besides, according to a time span hj from the time when a Web text oj in a Web text subset relevant to the behavior of a certain user ui before the time t to the time t, calculating out a preference influence degree dj to the user ui so as to obtain a pair of category number-influence degree cj of the Web text oj and generate a dynamic preference vector of the user ui; if the preference influence degree of the user and the category of the Web text to be recommended is found to be higher than or equal to a threshold value tan, recommending the Web text to be recommended to the user. According to the method disclosed by the invention, the dynamic influence changed with the time lapse of current preference from the historic behavior of the user is considered, and the method is more accurate in recommendation, has dynamic performance, and more conforms to the actual condition.

An information processing device includes: a user information obtaining unit configured to obtain information relating to data of content that a user has used; a meta information obtaining unit configured to obtain content meta information corresponding to content that the user has used; a first vector generating unit configured to generate a first user preference vector with each of the obtained content meta information as elements thereof; a second vector generating unit configured to generate a second user preference vector wherein the generated first user preference vector is analyzed and the number of elements of the first user preference vector is compressed; and a user identifying unit configured to identify a user corresponding to a second user preference vector having a high similarity to a second user preference vector determined beforehand from multiple second user preference vectors.

The invention relates to a program recommendation apparatus and method. The apparatus includes: a module configured to extract category information and program abstracts of programs contained in an electronic program guide, extract program-specific terms from the program abstracts by morphological analysis and combine the category information and the program-specific terms to generate category-added terms; a module configured to analyze a history of programs viewed by a user based on the generated category-added terms to generate a preference vector indicating user's preferences for programs; a module analyzing the program abstracts based on the category-added terms to generate broadcast program vectors; a module generating a relevant term model for the category-added terms; a module calculating similarities between the preference vector and each of the broadcast program vectors based on the generated relevant term model; and a module outputting programs having the calculated similarities satisfying a predetermined condition as recommended programs matching with the user's preferences.

The invention discloses a personalized scenic spot recommendation method based on tourist preference modeling, and the method comprises the steps: collecting data, carrying out the preprocessing, and carrying out the numbering of tourists, scenic spots and other objects; converting the display score into an implicit score, and dividing a positive case scenic spot and a negative case scenic spot; constructing a triple and scenic spot knowledge map, and generating a feature vector and a context feature vector of each scenic spot; generating vector representations of historical tourist tour scenic spots and candidate scenic spots through the KCNN; calculating an influence weight of each historical touring scenic spot of the tourist through the attention network to obtain a preference vector of the tourist to the scenic spot; calculating the scenic spot touring probability of the tourists by using the DNN, and generating scenic spot recommendation lists of the tourists according to the probability from small to large. According to the method, when different influences of historical visiting scenic spots of tourists on the candidate scenic spots are depicted and diversification preferences of the tourists are represented, the attention network is used for calculating the influence weights of the historical visiting scenic spots of the tourists on the candidate scenic spots, so that the recommendation result better conforms to the preferences of the tourists.

It is possible to repeatedly recommend a content which attracts a user. Provided is a content recommendation method which includes: a step (S402) which selects some from a plenty of music compositions in accordance with attribute conditions successively generated; a step (S403) which further selects a part or all of the selected music compositions in accordance with the similarity degree between the feature vector of each of the selected music compositions and the user preference vector; and a step (S404) which presents the selected music composition to the user.

Systems and methods for building a search index for query recommendation and ad matching are disclosed. The system accesses a query-URL graph and extracts a subgraph related to an ad campaign. The subgraph is annotated according to desired criteria. The sub graph is reversed and the reversed annotated subgraph is ranked to find nodes of importance. The nodes of importance are then used to build a preference vector which is used to find a stationary distribution of the sub graph. A plurality of random walks of the sub graph is performed to build a corpus of words. The corpus of words are input into a language model to learn associations, from which the top query terms associated with an ad campaign are found and indexed. The index is then inverted for recommending ads for received query terms.

The invention discloses a commodity recommendation method combining an attention network and user emotion. The method comprises the steps: 1) extracting the score and comment data of a user for a commodity, carrying out the preprocessing, and constructing a sample set T; 2) continuously training to obtain an attribute matrix W of a corresponding field by adopting an unsupervised learning model andutilizing text data commented in the T; 3) constructing a neural network structure C based on attention; constructing a user preference vector U and a commodity feature vector I by using a W-based memory network and a recurrent neural network as bases and using the predicted emotion score as a weight, calculating a prediction value of a missing score by using the U and the I, and calculating attribute vectors of a current user and a commodity for final recommendation explanation; and 4) performing descending sort according to the predicted scores, recommending the first N commodities to the user, and providing explanation of attribute levels for a recommendation result according to the attribute matrix and the attribute vectors, thereby solving the problems of lack of explanation, difficulty in processing large-scale data and the like of a traditional score prediction recommendation method.

The invention provides a dynamic news recommendation method based on a hierarchical attention network. The adopted modules comprise the hierarchical attention network, a convolution layer and a full connection layer. The hierarchical attention network comprises a sentence level attention network and a news level attention network. A news sequence and candidate news of the user are subjected to thesentence level attention network to obtain attention weights of sentences in the news sequence; a weighted sum of the sentence content vectors is calculated to obtain a news content vector; the embedded representation of the news sequence is connected with the content vector to obtain the integral representation of the news, and the integral representation passes through the attention network ofthe news level to obtain the attention weight of the news in the news sequence and obtain the final representation of the news; in the convolutional layer, the final representations of the historicalnews are stacked in sequence to obtain a matrix, and the matrix is input into the convolutional layer to learn a user sequence reading mode to obtain a sequence preference vector; and in the full connection layer, the sequence preference vector, the candidate news overall representation and the user embedded representation are connected to obtain the probability that the user clicks the candidatenews.

The invention discloses a recommendation method and system for local low rank matrix approximation based on implicit feedback information, and belongs to the technical field of recommendation. The invention starts from the implicit feedback information that is ignored in the user information, uses an implicit feedback information optimization feature extraction model to extract the single-dimensional multi-dimensional preference vectors of a user and an item according to the local low rank matrix approximation principle, and gives a more accurate user rating recommendation manner on the basisof preference information of the user and the item. The recommendation system comprises a data engine module, a multi-dimensional feature extraction module, a score prediction module, a collaborativerecommendation module and an information update module; through the collaborative work of all the modules, the multi-dimensional recessive factors of the user in different environments are extracted,and the accuracy and personalization of the whole system recommendation result is improved; the universality is high; and the method and the system are suitable for most of the existing recommended scenarios.

Disclosed are a recommendation system construction method and device, being applied to a system comprising a plurality of users, the scoring information of objects of the users, the user preference vectors of the users, and the object feature vectors of the users are stored at the user terminals of any user, and the plurality of user terminals are trained in a collaborative mode to realize matrixdecomposition.

The invention discloses a comment-driven deep sequence recommendation method, which comprises the following steps of: establishing a vocabulary for a user comment text, and endowing each word with a randomly initialized word vector; constructing a document word vector expression matrix for each document; obtaining an aspect-perceived document expression tensor and a plurality of feature maps; calculating the long-term preference vector of the user and the vector representation of the commodity; calculating user short-term preference vectors of a joint level and an individual level; performingweighted addition on the two levels to obtain a final user short-term preference vector; multiplying the short-term preference vector of the user by a reduction coefficient, adding the short-term preference vector of the user to the long-term preference vector of the user to obtain vector representation of the user, and calculating a preference score of the user for the commodity; training and obtaining an RNS model; and applying the trained RNS model to an online sequence recommendation scene. The comment-driven sequence recommendation problem is well solved, the method has the advantages ofbeing high in training speed and short in test time, and it is shown that the method has wide practical significance and commercial value.

The invention discloses a recommendation method, device and equipment based on a self-adaptive attention mechanism and a computer readable storage medium. The method includes obtaining a short-term preference vector of the current shopping cart and a long-term preference vector among all historical shopping cart vectors based on a self-adaptive attention mechanism;reflecting the short-term preference diversity and the long-term preference diversity of the user respectively according to the short-term preference vector and the long-term preference vector, and finally, inputting the combinationof the short-term preference diversity and the long-term preference diversity into the recurrent neural network LSTM to better learn the sequence behavior of the user, so that interested commodities are accurately recommended to the user.

The invention discloses a detection method for user group behavior rules based on stay places in a mobile trajectory. The method comprises the following steps: firstly, receiving multi-user trajectory data from a data collector; secondarily, extracting stay place sequences from the trajectory data, and transforming the stay place sequences into place preference vectors which represent user schedule behaviors; then extracting schedule behavior vectors, and using a weighted average method to calculate schedule behavior pattern vectors; finally, calculating similarities among users based on schedule behavior patterns, and using a Kmeans clustering algorithm to detect behavior rule patterns of user groups. According to the invention, the similarities among the users are calculated by using the schedule behavior patterns based on the stay places to solve the problem of nonoverlapping of user stay areas, so that the behavior rule patterns of the user groups can be effectively detected.

A collaborative recommendation model construction method based on knowledge graph preference propagation comprises the following steps: (1) performing knowledge modeling by using an ontology modeling tool, extracting entities and relationships to obtain a triple, and constructing a domain knowledge graph; (2) obtaining multi-layer preferences of the user through the preference propagation model, and calculating vector representation of the user preferences according to the multi-layer preferences of the user; (3) learning the vector representation of the article based on the user-article interaction matrix, and performing dot product operation on the preference vector representation of the user and the implicit vector representation of the article to calculate the click probability of the user on the article; and (4) recalling articles interested by the user according to the click probability, comparing a recall result with a user interaction list, and performing descending sorting after excluding the articles operated by the user to obtain a recommendation list. According to the recommendation model provided by the invention, the path structure information of the knowledge graph can be fully utilized, and the interpretability and diversity of recommendation results are improved.

If content having an attribute different from that of content usually used by a user is selected, the present invention allows proper content recommendation in accordance with the selection. When notified from a control block of a selected piece of music, a short-term preference vector generation block gets music vector of that piece of music from a music vector storage block and stores obtained music vector into a short-term preference vector storage block as short-term preference vector. A recommendation block computes a difference between short-term preference vector and long-term preference vector and, by use of a weight based on the computed difference, determines a piece of music to be recommended to a user, thereby recommending the determined piece of music to the user.

A multi-preference high dimension object optimization method based on co-evolution comprises the following steps: 1, initializing parameters; 2, updating; 3, using an ASF expansion function to map theideal solution in an evolution population into a target space, using same as a preference vector to guide the population to evolve in a reference direction, using the preference area selection strategy to obtain two temporary reference points and further builds a decision maker interested area ROI, determining the upper and lower boundary scopes in preference vector set generation, and using a co-evolution mechanism to guide the population to converge to the preference area; 4, using the co-evolution mechanism to optimize the target function, and selecting N candidate solutions with the maximum adaptation value from the candidate solutions in the ROI area scope and entering the next generation evolution; 5, determining whether termination conditions are met or not, if not, returning to step 2; if yes, outputting the optimal solution set, and ending the algorithm operation. The solution set obtained by the method has better convergence, has strong transplantability, and has better implementations.

The invention discloses a content distribution method based on D2D and service unloading, which is applied to a fifth-generation mobile communication system, and belongs to the field of wireless communication. The method comprises the following steps of (1) constructing a popular content set; (2) constructing a user preference vector; (3) constructing a user cluster preference vector; (4) randomlyselecting a user cluster; (5) constructing a trade-off model of an unloading rate and energy consumption; (6) distributing popular contents; (7) calculating remaining storage space size; (8) judgingwhether the remaining storage space size is larger than 0 or not; (9) selecting a to-be-distributed content cluster; (10) judging whether the content distribution of all user clusters is completed ornot; and (11) completing the content distribution of all users. Compared with the traditional content distribution method, the method has the advantages that the high service unloading rate of a basestation is guaranteed while the energy consumption is considered, and the user experience quality is improved.

The invention discloses a user use preference based application recommendation method. According to the invention, use preference of different users on N seed application program and grades corresponding to the user preference are acquired and preference values of the users are generated through preference vectors. When a random user performs a preset operation on an application program in a system, target users are determined and a message recommending the application program is sent to the target users. The invention discloses a user use preference based application recommendation system. The system includes a storage module, an acquisition module, a calculation module and a recommendation module. By adopting the method and system provided by the invention, application programs needed by system users can be recommended to the users more efficiency and accurately. On one hand, a complicated process of trying a large quantity of different programs by the users is saved. On the other hand, high quality application programs become available to target customers easier and possibility of acquiring more applications is increased.