656results about How to "Improve recommendation accuracy" patented technology

The present disclosure provides an emoji recommendation method and device, The emoji recommendation method may include: acquiring training sentences corresponding to emojis; clustering the training sentences to categories; according to a number of training sentences in each category, calculating a first probability corresponding to each category; according, to correspondences between the training sentences and the emojis determining emojis included ifs each category, and calculating a second probability corresponding to each emoji in each category; according to occurrences of each word in each category, calculating a third probability corresponding to each word in each category; according to the first probability and the third probability, calculating a fourth probability of a target sentence belonging to each category; according to the fourth probabilities corresponding to the categories, determining a target category for the target sentence; and according to the second probability corresponding to each emoji in the target category, recommending an emoji.

The invention discloses an output method, system and server of recommendation information. The method comprises the following steps of: extracting user action data in a predetermined statistic period from a user action database; classifying the user action data according to commodity identifications to obtain a time sequence about interestingness of each commodity in the statistic period; computing a purchase peak probability of the commodity according to the time sequence about interestingness; when receiving a command of outputting the recommendation information, sorting the purchase peak probability according to an order from high to low, and outputting the recommendation information of the commodity according to a sorting result. The method provided by the invention automatically computes the purchase peak probability of the commodity on time dimension according to the user action data, thereby improving the recommendation accuracy of a recommendation system and reducing the transmission quantity of unnecessary data in the network; since the recommendation information is automatically modified and output by a server, and the computation capability of the server is sufficiently used, the information input quantity and computation resource of the recommendation system can be reduced when the recommendation information is adjusted according to the purchase peak probability.

The invention belongs to the technical field of wireless communication networks and particularly discloses a crowd portrayal system and a crowd portrayal method based on a microblog label. The system comprises two main modules of a microblog label recommendation module and a label theme clustering module; the first module adopts a label recommendation algorithm covering three steps, wherein a first step is homogeneous label recommendation, a second step is co-occurrence label extension, a third step is that a semantic network is built on the basis of a Chinese mapping knowledge domain, the semantic similarity between labels is measured by using a network topology property, the labels with same or similar semantics are thus removed, and the refining property of the label used for portraying a user is ensured. According to the system and the method, the condition that the commercial application value of the label of the microblog user is wide is utilized, and the research direction is indicated for the mining algorithm of labels of internet users and the application of the Chinese mapping knowledge domain.

The invention provides a news recommendation system and method based on an FOLFM model. Based on a content recommendation method, a news content model is expressed abstractly through a latent class model and content characteristics, and an individual latent class preference model is built for each user. Real-time training is carried out on a real-time behavior record of a user to obtain preference, on certain latent class news, of the user, whether the news is recommended to the user is determined through calculation, and a final news recommendation list is obtained after a series of processing processes. The news recommendation system and method based on the FOLFM model deeply excavate the interest of the user, improve recommendation accuracy and satisfaction of the user, avoid a cold starting problem of the news, and guarantee performance under the condition that the recommendation effect is improved as much as possible. The experiment shows that the news recommendation system and method based on the FOLFM model not only guarantee the requirements for high accuracy and high speed, but also realize visual real-time recommendation for the user.

The invention relates to an individual network news recommending method based on multitheme tracing, which comprises the following steps: obtaining a news webpage browsed by a user and dividing the news webpage into a plurality of user interest submodels; dynamically updating a multitheme interest model of a user according to the condition that whether the user reads a news report recommended by the interest submodels or not; judging whether the quantity of the interest submodels exceeds a preset threshold value, if so, searching and deleting the interest submodel which is farthest from the interest of the user; searching the maximal similarity between news to be recommended and all the interest submodels; calculating sequencing values of the news report and sequencing the sequencing values from the large value to the small value; and recommending a sequenced news list to the user. The invention can cover various interest characteristics of the user and has the advantages of high recommending accuracy and lighter following maintenance burden of a system.

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 system, a method and a game platform capable of recommending games in a personalization mode. The recommending system recommends the games according to a user identifier, and comprises a behavioral analysis module, a relative detecting module and a recommending module, wherein the behavioral analysis module is used for collecting historical data in which a user is interested and historical data about different games with which the user plays, obtaining basic preferring value of the user to different games according to the collected data, and obtaining a game preferring degree ranking list of the user to different games. The relative detecting module is used for calculating relevancy degrees of different games in the game preferring degree ranking list according to a relevancy algorithm, and ranking the relevancy degrees to obtain a possibly preferable degree ranking list of the games. The recommending module is used for recommending the possibly preferable degree ranking list of the games to the user. The recommending system analyzes the collected historical data in which the user is interested and the historical data of different games, explores the favor degree of the user to the games, and therefore the recommending system can specifically recommend the games to the users with different personalities.

The invention discloses a music recommending method and system which are used in a mobile terminal. The contextual information of music is obtained first. The method comprises the steps that the contextual information of a user is obtained; and the matching degree of the contextual information of the user and the contextual information of the music is computed, and K pieces of music with the computed large matching degrees are recommended to the user, wherein, K is an integer larger than or equal to 1. According to the method and system, the contextual information of the user and the contextual information of the music are subjected to matching computing, the music with the high matching degrees is recommended to the user, the coupling performance of the music with the user is enhanced, and accordingly music recommending accuracy is improved.

The invention discloses a method and device for recommending cash goods, equipment and a storage medium. The method comprises the steps of obtaining multiple pre-determined positive samples and negative samples composed of cash goods purchased by multiple users; collocating preset user feature information with the positive samples and the negative samples correspondingly and generating a target training set; adopting a binary classifier algorithm, and training a preset mathematical model according to the target training set to obtain a training model; obtaining user feature information of a user to be predicted, inputting the user feature information as an input variable into the training model, and outputting an ordering result of cash goods corresponding to the user to be predicted. According to the method and device for recommending the cash goods, the equipment and the storage medium, the problem is solved that the cash goods recommended for the user are largely deviated from the cashed goods actually purchased by the user, the recommendation accuracy is improved, the profit is increased, and the user experience is improved.

The present invention provides a collaborative filtering recommendation method based on typicality and a trusted network. According to the method, an original sparse scoring matrix is replaced with a dense user typicality matrix and a project typicality matrix, and a user-trusted network is also used to improve a traditional collaborative filtering recommendation algorithm. A sparseness problem referring to little user score data in the traditional collaborative filtering recommendation method is solved by using the project typicality matrix in a project set and the user typicality matrix in a certain project set that the user is interested in, the user-trusted network is also used to further improve the recommendation accuracy, and data dimensionality deduction is realized. The recommendation result can fully use the impact of a user social trust relationship on similar user interests.

The invention discloses a recommendation system optimization method with information of a user and an item and a context attribute integrated. According to the method, the information of the user, the item and the context attribute is integrated in a matrix decomposition model, and recommendation accuracy is improved in a personalized recommendation system. The recommendation system optimization method with the information of the user, the item and the context attribute integrated is characterized in that different influences of the information of the user, the item and the context attribute on overall scores, user interests and item scores are considered, and is applied to calculation of an original matrix decomposition model. The influences of the user, the item and the context attribute on the scores are considered at the same time, and therefore the recommendation accuracy is obviously higher than that of the rectangular decomposition model which only adopts user program two-dimensional score matrix information.

The invention discloses a personalized recommendation method based on bilateral diffusion of a bipartite network, which mainly solves the conflicts among several main arithmetic performance indexes including accuracy, diversity and novelty of a recommendation list with respect to personalized recommendation of Top-N. The method comprises the following steps: (1) constructing a user-article associated bipartite graph network, wherein it is supposed that each target user node has certain allocable resource, which can be allocated to other node objects according to a preset allocation mechanism; (2) establishing an article-article and user-user second-order correlation matrix; (3) implementing a bilateral diffusion process to obtain a user-article interest matrix; and (4) giving a recommendation list of each user with the length N to complete the recommendation of Top-N. Based on the thought of network communication, the effect of the method is obviously better than that of a classical collaborative filtering method, the long tail mining concerned by a personalized recommendation system is better realized, and the method can be used for solving the Top-N problem of personalized recommendation.

The invention provides a music recommendation method and system based on group perspective. User characteristics and the favorite categories of songs listened to by users are analyzed according to user history logs, then user sets are divided according to the user characteristics and the favorite categories, user groups with the same general characteristics are built, and the user groups to which the users belong are analyzed; the collaborative filtering fusion algorithm is adopted to generate a recommendation song list, and the recommendation song list is pushed to the users. In the whole process, accurate user group division is carried out for a huge number of users, user interesting multidimensional perspective is achieved by deeply excavating the user data, and the music recommendation precision is improved. In addition, the user groups with the same general characteristics are built, and then the hybrid recommendation algorithm is used, so that music recommendation accuracy and reliability are further improved.

The invention discloses a method for recommending commodities to customers. The method comprises the following steps of: calculating predicted scoring values of target customers on all un-scored commodities according to a set of commodities purchased by the target customers and a corresponding scoring value set, and selecting K commodities with maximum scores as a candidate recommended commodity set by a candidate commodity calculating module; and adding the predicted scoring value of each candidate commodity of the target customers as an actual score of the target customers into a system, calculating the predicted values of the purchased commodities, calculating the evaluation indexes of corresponding prediction accuracy, sequencing the candidate commodity set according to the advantages and the disadvantages of the acquired evaluation indexes, and thus acquiring a final recommending list by a candidate commodity sequencing module. By the method, the candidate commodities with best system evaluation indexes are sequenced in the front of the list, so that the individual requirements of the customers are met, and the recommending accuracy of the whole system is improved.

The invention discloses a collaborative filtering recommending method based on the characteristics and credibility of users. The collaborative filtering recommending method based on the characteristics and credibility of the users comprises the following steps: respectively computing similarity between users according to the historical evaluating scores of users to a project and characteristics of the users; selecting proper weight to combine two kinds of similarity to obtain the finial similarity among the users; computing the nearest neighbor of the user according to KNN (K Nearest Neighbors) method; using the quantity of historical evaluating scores of the users as the credibility, and computing a recommending result by using the nearest neighbor matrix of the user and a predictor formula added with the credibility.

The invention discloses a taxi passenger-searching path recommendation method based on information entropy. The steps include: mining a taxi track, and extracting passenger carrying point data therefrom; mining and extracting passenger-searching points representing passenger gathering places from the passenger carrying points obtained by extraction; measuring attribute values of factors according to which a taxi selects a passenger searching position in a passenger searching process; according to the current position and current time of the taxi, obtaining passenger searching points in a designated range, and building a passenger searching point selection model which includes a set of passenger searching points where the taxi can arrive departing from the current position at the current time and a decision matrix; solving the passenger searching point selection model based on information entropy, and obtaining a designated number of passenger searching points with optimal synthesized attributes; and using the passenger searching point obtained at the previous time as a reference point, repeatedly building and solving the models for a designated number of times, and generating a passenger searching path according to an obtaining level of optimal passenger searching points and recommending the path. The taxi passenger-searching path recommendation method based on information entropy has the advantages of being capable of bring good benefits to a taxi driver and being high in recommendation accuracy.

The application discloses a user involving project recommendation method. The method comprises: acquiring a project set to be recommended to target users; acquiring a user set formed by users that have been involved in projects in the project set to be recommended to the target users; acquiring an influence value of each user in the user set ; according to projects involved in participation behaviors of the users, acquiring similarity degree between each user of the user set and the target users; considering the influence value of each user in the user set and the similarity degree between each user of the user set and the target users, calculating predicted values of degree of interest the target users have on each project in the project set to be recommended to the target users; and considering the predicted values, selecting at least one project from the project set to be recommended to the target users, and recommending the selected at least one project to the target users. According to the user involving project recommendation method and apparatus, the recommendation accuracy and the quality of recommended projects are improved, and the recommended projects are higher in representativeness and effectiveness.

The invention discloses a project-level and feature-level deep collaborative filtering recommendation algorithm based on an attention mechanism. The algorithm comprises the following steps of S1, counting historical project scores of a user; S2, calculating the feature level content representation of the user on the target project according to the historical project score of the user; and S3, calculating a project-level prediction score of the user for the target project according to the historical project score of the user and the technical result of the S2. According to the algorithm, the recommendation precision is improved to a certain extent by combining attention mechanisms on a project level and a feature level, and compared with the prior art, the algorithm has higher interpretability in analysis of historical preferences of users. The extended DACFs, such as recently proposed neural collaborative filtering and discrete collaborative filtering, will also be considered in othercollaborative filtering models, a higher-order characteristic level attention mechanism is explored for future research, and the theoretical basis of research of the recommendation system is further tamped.

The invention relates to the technical field of recommendation, in particular to a collaborative filtering method and system based on similarity propagation. The method comprises the steps as follows: traversing historical behavior data of all users to obtain a relationship vector describing the preferences of all the users on articles; setting a threshold value by the relationship vector and calculating the similarities among the users; calculating the similarity between a target user and the other user, with a similarity value of zero in the matrix, by a similarity propagation calculation principle; obtaining an estimation value of the preference degree of a current user on an unselected article according to the preference degree of the other user most similar to the target user on an article unselected by the target user; and screening the prediction results of all the users to generate recommended articles for all the users. The system comprises a data relationship vector module, a threshold value judgment module, a similarity propagation calculation module, a preference degree estimation module and a prediction screening module.

The invention discloses a stream-oriented recommended engine, recommendation system and recommendation method based on clustering. According to the stream-oriented recommended engine, the recommendation system and the recommendation method based on clustering, an incremental clustering method is constructed in cooperation with advantages of a clustering structure and a collaborative filtering method, users and commodities are classified through the clustering structure, and the incidence relation between each user and the corresponding commodity is excavated based on the collaborative filtering method. By means of the stream-oriented recommended engine, the recommendation system and the recommendation method, the recommendation accuracy rate can be guaranteed, meanwhile the field characteristic and correlation characteristic of the recommendation result are improved, and recommendation accuracy is improved.

The invention discloses a heterogeneous information network based content providing method. When users subscribe to a recommendation service, content is recommended to the users through an optimal prediction matrix, a similarity matrix of the users and a similarity matrix of projects are obtained by an element path based similarity calculation method according to a heterogeneous information network through the optimal prediction matrix, the fusion is performed on two or three of the user similarity matrix, a user and project evaluation matrix and the project similarity matrix to obtain the internal relation between the users and the projects, the prediction is performed by a collaborative filtering based matrix decomposition prediction method, and results are combined to obtain the optimal prediction matrix. According to the heterogeneous information network based content providing method, the recommendation accuracy is effectively improved, the results which are accord with the will of the users can be recommended to the users through limited time calculation on the basis of the existing data, the cold start problem is partially solved, and the interpretability of the recommended results is improved.

The invention discloses a recommending system and method used for search input, and relates to the field of search engines. The system includes a keyword acquiring unit suitable for acquiring search keywords as per user input, a search tree storage unit suitable for storing Chinese characters in form of a tree data structure, a recommended word bank suitable for storing recommended words, an address acquiring unit suitable for inquiring the search tree storage unit and obtaining address information of the recommended words according to the search keywords, and a recommending unit suitable for obtaining the recommended words by inquiring the recommended word bank as per the address information and recommending the recommended words to the user, wherein each data node in the search tree storage unit stores one Chinese character and contains address information of recommended words of the character. The system and the method enable each data node in the search tree storage unit to contain the address information of all recommended words of the node character, thereby improving the speed for inquiring the recommended words and overcoming the problem about lowered inquiring speed due to enlarged capacity of the recommended word bank.

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 provides a method and system for recommending a commercial tenant. The method comprises: a query request and a position of a user are obtained; commercial tenant searching is carried out according to the query request to obtain a searching result, and extraction is carried out on the searching result according to the position to obtain a commercial tenant candidate list; a commercial tenant association recommending list is extracted according to the commercial tenant recorded in the candidate list, wherein the commercial tenant association recommending list is generated based on a similar user browsing record of the user; and the candidate list and the association recommending list are screened to obtain a commercial tenant recommendation result. The system includes an input obtaining module, a request processing module, an association extraction module, and a screening module. With the method and system, recommendation accuracy can be effectively improved.

The invention discloses a neighborhood recommendation method based on differential privacy protection. The neighborhood recommendation method comprises the steps that firstly, collected article evaluations or preferences of users are converted into a user-scoring matrix serving as a training set of a recommendation method model at a training stage; secondly, a scoring prediction model is established by utilizing a neighborhood-based recommendation method, and the article scoring situations of the users are predicted. In the neighborhood-based recommendation method, and an average value under the differential privacy protection, user bias items and article bias items are calculated; at a scoring prediction stage, a differential privacy protection method based on an index mechanism is utilized to select neighbors; the local sensitivity of similarity is utilized to add Lapras noise for differential privacy protection; finally, the scoring prediction model and trained differential privacyprotection model parameters are utilized to predict articles scores of the users. By adopting the neighborhood recommendation method, differential privacy protection can be conducted on the information of the users when recommendation results are provided, and the method has higher recommendation accuracy.

The invention discloses a mobile phone game recommendation method based on a binary decision tree, mainly trains a binary decision tree classification model to determine whether a user is interested in a mobile phone game to be recommended or not so as to transform influence on recommendation by characteristic weight into the automatic prediction of a learning model from subjective assignment. The mobile phone game recommendation method has the characteristics that the binary decision tree is used for determining the characteristic selection and weight measurement problems in a game recommendation scene, and a corresponding recommendation model is given. The mobile phone game recommendation method can be used for favorably guaranteeing the accuracy of a recommendation result and user preference, and meanwhile, the diversity of the recommendation result can be guaranteed on the premise that the recommendation result does not need to be additionally subjected to second pickup.

The invention provides a sequence recommendation method based on a self-attention auto-encoder, and the method comprises the steps: obtaining a user commodity sequence and a scoring matrix, and carrying out the preprocessing of the commodity sequence; training the commodity sequence by using a self-attention model, and predicting a relevance score of the user and the commodity; reconstructing thescoring matrix by using an auto-encoder, and calculating a user preference index; and in combination with the relevance score of the user and the commodity and the user preference index, obtaining a high-score commodity to preferentially recommend to the user. According to the sequence recommendation method based on the self-attention auto-encoder, the article browsing sequence of the user is converted into a low-dimensional dense vector by using a word embedding method; the position codes are combined and input into the self-attention model, then the self-encoder is used for fitting and reconstructing the scoring matrix, the user preference index is calculated, finally, the final prediction score is obtained, recommendation is made for the user, factors such as long-term and short-term preferences of the user are considered at the same time, and the recommendation precision is effectively improved.

The invention provides a highly accurate code snippet recommendation method. The method comprises the steps of splitting and dividing a project, and generating class files, a code snippet , a code snippet method body and a code snippet note; generating a code snippet method body vector and a code snippet note vector based on the code snippet method body and the code snippet note; compiling and decompiling each class file, and extracting an instruction sequence of the code snippet from the decompiled file; generating an inquiry vector based on an inquiry input by a user, and screening out a group of primary recommendation results with highest text similarity according to a text similarity of the inquiry vector and a code snippet text characteristic; and according to statement similarity ofall code snippet s in the primary recommendation results, screening the primary recommendation results for the second time and then re-ranking so as to form a final recommendation result. The method overcomes the defect that the existing method is low in recommendation accuracy due to unique use characteristic and failure in filtering of recommendation results. The method is applicable to the field of recommendation of open source code snippets.

The invention relates to a random walk and diversity graph sorting-based personalized service recommendation method. Firstly the defect in calculating user similarity by a conventional Pearson correlation coefficient is analyzed, and then a similarity relationship between users is subjected to multi-time transmission through a WRW model, so that more similar neighbors are found for a target user,and the data sparsity problem is effectively solved; based on predicted QoS values of all the similar neighbors, a service graph model is built by adopting an SGMC algorithm to filter a large amount of candidate services with excessively low performance, reduce an algorithm optimization space, and ensure quick convergence of a subsequent diversity graph sorting algorithm; and finally, an optimal node set is searched for on the service graph model by adopting an ONCS policy, and k nodes contained in the set are a service list recommended to the user and having recommendation accuracy and functional diversity. The method has relatively high recommendation accuracy and relatively good functional diversity, and can meet potential functional demands of the users to the maximum extent while ensuring service recommendation quality.