153results about How to "Solve the sparse problem" patented technology

The invention discloses a collaborative filtering recommendation method and a collaborative filtering recommendation system of a convolution neural network integrating an attention model, which relates to the technical field of data mining recommendation, improves feature extraction efficiency and scoring prediction accuracy, reduces operation and maintenance cost, simplifies cost management mode,and is convenient for joint operation and large-scale promotion and application. The invention relates to a collaborative filtering recommendation method of a convolution neural network integrating an attention model, comprising the following steps: step S1, splicing a user feature vector and an item feature vector into a new vector; S2, sending the new vector as an input vector into the multi-layer perceptron to learn and predict the score; The attention model is fused into the object potential vector to obtain the convolution neural network of the object feature vector or the hidden layer of the multi-layer perceptron.

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 topic detection and tracking method based on microblog data. In the method, potential hidden subjects in large-scale social network information are mined. The method comprises the following steps: firstly, partitioning microblog data increasing massively according to time sequence properties, and filtering redundant information; secondly, analyzing and classifying text contents in time windows, returning key subject descriptions with independent semantics after extraction, and extracting topics in different time windows; and lastly, analyzing the inheritance and the identity of topics among the time windows to conclude the variation tendency of microblog topics. According to the method, the dynamic developing process of topic contents, namely, the generation, development, climax and extinction of topics can be shown, and topics are described more accurately and fully.

The invention provides a Chinese and English-named entity identification method and a system based on a conditional random field (CRF). The method comprises the following steps: (101) converting inquiry voice of a user into a text; (102) separating text information into Chinese characters and English letters on the basis of a finite state machine; (103) extracting the characteristics of a text of separated vocabularies; (104) performing entity identification on the text by adopting a training CRF model according to a characteristic extraction result, and marking an entity type, wherein the CRF model is a conditional random field model of a linear chain structure. The step (102) further comprises the following steps: (102-1) performing character separation on Chinese and English; (102-2) identifying English word strings by using the finite state machine, namely, combining adjacent English letters, blank spaces and symbols in English; (102-3) performing word segmentation on the English word strings.

The invention discloses a university library-oriented books personalized recommendation method, and solves the problems of poor large-scale data storage and query, extendibility and recommendation effect in an existing books recommendation algorithm of a university library. According to the basic thought, the method comprises the following steps of firstly, building a graph model by taking readers, books and the like in the library as nodes; secondly, converting operation log files of the readers into a reader-books category preference matrix, calculating similarity between the readers by the reader-books category preference matrix and a reader personal information matrix, and establishing an associated graph spectrum by taking operations and mined information as edges; thirdly, by combining the associated graph spectrum with spectral clustering, proposing a new books personalized recommendation model, and performing calculation to obtain class cluster distribution about the readers; and finally, when books recommendation needs to be carried out, calculating a recommended books list according to a collaborative filtering algorithm in a class cluster corresponding to a reader.

The invention discloses a method for predicting the path destination of a moving object, and mainly solves the problems of sparse data, a low hit rate and poor real-time performance during the prediction of the path destination in the prior art. The method includes the following steps: off-line calculating the history path data in a training data set to obtain a location set, a prior probability of each location, a single-step transition probability matrix and a comprehensive transition probability matrix; calculating the conditional probability and the posterior probability of each location being the destination based on the off-line calculated data; on-line predicting the destination of the path to be predicted based on the posterior probability. Compared with the existing path prediction method, the method provided by the invention has the advantages that a prediction can be performed even when the data in the training data set is sparse; the calculation process of the prediction is optimized, so that the impact of the non-aftereffect property of Markov chain on the prediction accuracy is reduced; the prediction accuracy is improved; the method can be used for pushing location-related targeted advertisements and deploying criminal arrest plans in advance.

The invention relates to an image three-dimensional matching method and a device thereof. The invention is characterized in that the invention comprises the following steps: a graphic identifier is adopted to extract feature points of a detection image one and a candidate matching image two; a multi-dimensional vector identifier makes a multi-parameter description to the feature points of the feature list to acquire the multi-dimensional vectors described by the parameters of the feature points; a data optimizer reduces the multi-dimensional vector described by the parameters of each feature point of the remarkable feature point list of the image one, calculates the similarity of each feature point of the feature point list of the reduced image one and the feature point of the feature point list of the image two to acquire the candidate matching point set of the image one; a classifier is used to classify the comparison result to judge the matching result of the image one and the image two. The invention has higher matching accuracy and efficiency to the three-dimensional image matching.

The invention relates to a method for short-term predicting of photovoltaic generation power on the basis of similar day feature classification and extreme learning machine, belongs to the technical field of photovoltaic station power generation and aims to predict output power of a photovoltaic power generation system. The method includes: firstly, based on meteorological data of a public weather forecast network, subjecting meteorological data and photovoltaic generation system capacity to similar day feature classification according to meteorological features such as season and day type and according to photovoltaic generation power features; secondly, applying a single hidden layer neural network based on extreme learning machine algorithm as a forecast model, and applying selected similar day data as a training sample to train the single hidden layer neural network in the extreme learning machine algorithm; and thirdly, applying known capacity sequence, maximum air temperature and minimum air temperature of similar day forecast periods of most similar forecast days, and maximum air temperature and minimum air temperature of precast periods of the precast days as neural network input, and predicting generating power of a photovoltaic station in future three hours. The requirement of the method for devices is low, a predicting model is highly targeted at regions, and the method is easy to implement and high in precision.

The invention provides an interest point recommendation method based on a graph neural network and user long-term and short-term preferences. The method comprises the following steps: taking a point-of-interest sequence accessed every day in historical sign-in data of a user as a session sequence; creating a directed graph based on the sessions, where each session sequence is regarded as a sub-graph, indicating that each node represents a point of interest, and each directed edge represents that a user accesses a pointed point of interest after accessing a source point of interest of each edge. Based on this graph, relationships between points of interest are captured by a graph neural network and vector representations of the points of interest are accurately generated. Based on the representation vectors of the interest points, the interest points to be accessed in the next step are recommended for the user by combining an attention mechanism. According to the invention, a better geographic information model is fused from the perspectives of users and the interest points. Therefore, the geographic distance between the users and the interest points and the sign-in frequency of theusers on the adjacent interest points are used in the model, and the problem of sign-in data sparseness is solved.

The invention discloses a trajectory data sorting method based on a generative adversarial network. The trajectory data sorting method comprises the following steps: inputting real trajectory data into the generative adversarial network so that the generative adversarial network is trained into a generator so as to generate simulated trajectory data which are the same as the real trajectory data in distribution; then using the generator of the generative adversarial network to generate a plurality of groups of simulated trajectory data; and finally carrying out sorting treatment on the generated groups of simulated trajectory data and the real trajectory data together so as to obtain trajectory user mapping. According to the trajectory data sorting method, the distribution of the real trajectory data can be simulated through the generative adversarial network, generated simulated trajectory data and real trajectory data are together used as data sources for sorting trajectory data, andthe trajectory data are sorted, so that the problem of data sparsity can be solved effectively, and the adverse impact generated by sparse trajectory data on trajectory data sorting is avoided; and due to the fact that corresponding trajectory user mapping also exists in the sparse trajectory data, the sorting for the sparse trajectory data can be realized, and the data sorting effect is improved.

The invention discloses a personalized film recommendation method and system based on feature augmentation. The necessity and feasibility that the personalized recommendation system is transferred to a cloud calculation platform are discoursed by analyzing the problems brought by mass data to the existing personalized recommendation system, and by combining a mixed recommendation model adopting feature augmentation, the method comprises the steps that an original extremely-sparse user item rating matrix is filled for the first time through a collaborative filtering algorithm based on items to generate a pseudo two-dimensional table, then the pseudo two-dimensional table is further filled through a collaborative filtering algorithm based on the user, and lastly the defects of strict object attribute matching are avoided through conversion of quantitative knowledge and qualitative knowledge of a cloud module. According to the method, all rating data of the user is fully utilized, the calculation efficiency is greatly improved and calculation speed is greatly increased through parallel calculation of cloud calculation, therefore, the better experience is provided for the user, and the advantages of an enterprise in competition are achieved.

The embodiment of the invention discloses a method and device for classifying voice data. The method for classifying the voice data comprises the steps that first voice data of the category to be recognized are obtained; windowing is carried out on the voice time shaft of the first voice data according to a preset windowing algorithm; an MFCC feature vector is extracted for the voice data in each window; vector quantization is carried out on each MFCC feature vector to obtain a one-dimensional first feature value; all the first feature values are calculated according to a preset histogram drawing algorithm to obtain a first histogram of the first voice data; similarity calculation is carried out on the first histogram and preset histogram feature templates corresponding to the voice data of various voice categories, and a first histogram feature template most similar to the first histogram is obtained; the voice category of the feature template is the voice category of the first voice data. Compared with the prior art, according to the technical scheme, the accuracy and speed of classifying the voice data are improved.

The invention discloses a pulse UWB (Ultra Wide Band) communication system based on a CS (Compressed Sensing) theory. A digital signal (X) transmitted by a transmitting terminal is effectively observed by introducing signal detection in the CS theory through a sparse algorithm in a digital signal processor, combining a random precoding module added at the transmitting terminal and matching with apulse generating module, a UWB channel and a low-speed sampler, and a common recovery reconfiguration algorithm in the CS theory is utilized to recover and reconstruct the digital signal (X) and realize communication. In the communication system provided by the invention, a plurality of parallel correlators and low-speed samplers needed in a first parallel scheme are not needed at a receiving terminal so that the problem of high hardware implementation complexity in the parallel scheme is solved; meanwhile, the low-speed sampler is directly used at the receiving terminal for sampling, and an analogue information converter is not used, so that the influence of the causality of a low-pass filter in the analogue information converter on observation is avoided and the problem of sparse observation matrix caused by the causality in a serial scheme is solved.

The invention provides a method for refining the edge of a laser-point cloud. The laser-point cloud and multiple optical images of a detected object are obtained respectively by adopting a ground laser radar and camera shooting equipment, then an image point cloud is generated by using the optical images, finally an intensity image is generated according to the laser-point cloud, and the intensity image is matched with one of the multiple optical images to obtain at least three matching point pairs; laser scanning points and corresponding points are found from the laser-point cloud and the image point cloud respectively, coordinate transformation parameters are calculated by using homonymy point pairs formed by the laser scanning points and the corresponding points, the image point cloud is converted into the laser-point cloud, and the laser-point cloud with the edge refined is generated. According to the method refining the edge of the laser-point cloud, edge refining can be carried out on data of the laser-point cloud effectively, and the problem that point cloud data of the edge of an entity are missed or sparse is solved.

The invention discloses a hybrid collaborative filtering recommendation algorithm based on project attributes. The algorithm comprises the following steps: step 1, generating a user-project scoring matrix and a project-attribute matrix according to user scoring information and project information; step 2, respectively calculating project similarity and user scoring similarity according to the project-attribute matrix and the user-project scoring matrix; step 3, correcting the user score similarity by utilizing the project similarity to calculate the user similarity; step 4, calculating a common score reward factor, a project attribute preference factor and a user confidence coefficient factor to correct user similarity and obtain user final similarity; step 5, selecting the nearest neighbor of the target user according to the final similarity, and predicting the score of the target user on each project based on the score information of all the users in the nearest neighbor; And step 6,recommending the N projects with the highest scores to the target user. The method can be applied to sparse data, and recommendation accuracy can be improved.

The invention discloses a 3D laser and 2D imaging synchronous scanning device and a scanning method thereof. A laser receiving and sending matrix is installed on a 3D laser radar, the point laser receiving and sending matrix and a line scanning imaging sensor are installed on the same rotatable platform in parallel, the height of the point laser receiving and sending matrix and the line scanning imaging sensor is the same, the positions are close or connected, the rotatable platform can drive the point laser receiving and sending matrix and the line scanning imaging sensor to synchronously rotate, the synchronous scanning device further comprises a control device, the control device comprises a synchronous clock module, the synchronous clock module is connected to the 3D laser radar and the line scanning imaging sensor signal, the control device can synchronously send the signal to the 3D laser radar and the line scanning imaging sensor through a synchronous clock module, so that the 3D laser radar and the line scanning imaging sensor can synchronously scan, and therefore the obtained data has synchronism and facilitates the fusion of radar laser scanning data and line scanning image data.

The invention discloses a social recommending method based on classification. The social recommending method comprises the steps that an evaluation matrix of a user on classification is constructed; an initial user classification matrix is constructed according to evaluation data of the user on a project and classification information of the project; normalization processing is carried out on the initial user classification matrix, and the user classification matrix is reconstructed through a matrix decomposition method; the credibility between the friends of the user is constructed through the friend information of the user; the parameters of an obtained model are learnt and obtained through a stochastic gradient descent method according to the evaluation of the user on the classification of the project in the user classification matrix, the evaluation of the user on the project predicted and obtained through a socialization model and the credibility between the friends of the user, and therefore the final evaluation of the user on the project is predicted. According to the method, project classification information is guided into the social recommending method for the first time, the socialization information of the user and the classification information of the project are integrated on the basis of original collaborative filtering recommendation, recommending precision is improved, and the problems of data sparseness and cold starting in a recommending system are solved.

The invention discloses a laser radar rotation control method for target detection. The laser radar rotation control method comprises the steps that S1, a laser radar is installed on an operating platform to form a detection system, the change trends acquired at different operating speeds of the operating platform that point cloud distributions change with the scanning speed of the laser radar areobtained; S2, when the detection system performs target detection, the operating speed of the operating platform is acquired in real time, the change trend of corresponding point distribution is obtained according to the operating speed acquired in real time, and the current required laser radar scanning speed is determined according to the change trend of the point distribution; S3, the laser radar is controlled to rotate according to the scanning speed of the laser radar determined in the step S2 so as to complete target detection. The laser radar rotation control method has the advantagesof being simple to achieve, small in working blind zone and large in effective detection distance and making scanned point cloud data rich.

The embodiment of the invention discloses a knowledge representation learning method and device, equipment and a storage medium, and relates to the technical field of natural language processing, deeplearning and knowledge graphs. A specific embodiment of the method comprises the following steps: sampling a knowledge graph sub-graph from a knowledge base; serializing the knowledge graph sub-graphto obtain a serialized text; and reading the serialized text according to the sequence on the knowledge graph sub-graph by using a pre-trained language model, and learning to obtain knowledge representation of each character in the serialized text. According to the embodiment, knowledge representation learning is oriented to entity and relationship representation learning in the knowledge base, semantic association of entities and relationships can be efficiently calculated in a low-dimensional space, the problem of data sparsity is effectively solved, and the knowledge acquisition, fusion and reasoning performances are remarkably improved. By utilizing the strong knowledge acquisition capability and context analysis capability of the pre-trained language model, the knowledge representation learned by the pre-trained language model can better represent the complex relationship in the knowledge base.

The invention discloses a collaborative filtering method based on a coupling topic model. The collaborative filtering method based on the coupling topic model is used for effectively combining historical grading information and user-generated content which is UGC in short in a recommendation system and performing effective recommendation through grading prediction. The method includes the following steps that firstly, a record of the user-generated content is obtained; secondly, a grading record on a product by the user is obtained; thirdly, a user file and a product file are extracted; fourthly, the coupling topic model is used for learning user feature vectors and product feature vectors; fifthly, the grade on different products by a target user is calculated according to the feature vectors, and corresponding product recommendation is performed. The analysis on user-generated content information is introduced into the collaborative filtering method, the user interest and product properties can be directly and explicitly found, the sparse problem of a grading matrix is effectively solved, and the effect more accurate than the prediction based on user grading information is obtained.

The invention discloses a position recommendation method for recommending positions related to a target position. The method is executed in computing equipment and comprises the steps that the co-occurrence frequencies of the target position and other positions are computed according to behavior logs of multiple users; a first number of positions are sequentially selected according to the sequence that the co-occurrence frequency values are arranged from high to low to generate a recommendation candidate set of the target position; the text similarity of the target position and each position in the recommendation candidate set is computed; and the positions of which the text similarity values are larger than a threshold value are selected to be recommended. The invention further discloses the corresponding computing equipment.