117results about How to "Good explanatory" patented technology

The invention relates to a Bayesian network (BN)-based rolling bearing fault diagnosis method. According to a common rolling bearing fault diagnosis method, a mathematical model is required to be established, and an initial diagnosis effect is unsatisfactory; problems of the selection of a wavelet base function are unsolved; and the interpretability of a deduction process is low. The method comprises the following steps of: sampling a vibration signal of a bearing, acquiring a sample, performing N-point rapid Fourier transformation processing to convert a time-domain signal into a frequency-domain signal, calculating a fault characteristic vector, discretizing the fault characteristic vector, establishing a fault diagnosis reasoning BN model, setting a fault sample to be diagnosed, acquiring an observational evidence of the bearing, finishing updating the reliability Theta of a fault diagnosis type node Bearing in the BN model, calculating a fault diagnosis type node, and outputting a result. A complex mathematical modeling process for the vibration signal is avoided, an obtained diagnosis reasoning model has the advantages of a few characteristic parameters, prominent fault characteristics, high interpretability and the like, and an effective way for solving the problems of the rolling bearing fault diagnosis is provided.

The invention discloses an event detection method based on Twitter, used to solve the real-time problem of news clue acquisition in the prior art. The event detection method based on the Twitter comprises the following steps: pre-treating tweets by using the natural language processing technology; performing online clustering on the tweets based on short text similarity; performing preliminary judgment on the event attributes of a tweet cluster by means of the characteristics of the tweets in each tweet class; and judging whether the detected event is a first event or not by means of the existing tweet event library, so as to prevent the event from repeating output. The event detection method based on the Twitter has low computational complexity and can be used for detection of sudden hotevents in large-scale flow data; and the event detection method based on the Twitter helps to obtain valuable information from a social media faster and more accurately, and can produce economic benefits directly or indirectly.

The invention discloses a method for recognizing superior shale gas reservoir by adopting conventional logging information. The method comprises the following steps: a, analyzing shale mineral composition; b, qualitatively recognizing the shale gas reservoir; c, quantitatively evaluating the shale gas reservoir; d, comprehensively evaluating the shale gas reservoir. According to the method, the shale gas reservoir is recognized by using a conventional logging curve, and a logging curve combination feature suitable for destination shale gas reservoir is summarized by comprehending mineral rock composition, clay type and deposition environment and analyzing common and personal corresponding features of the logging curve.

The invention disclosers a power grid fault diagnosis method based on topological graphic primitive information fusion; the method comprises the following steps: building a system various-bus, line, protection and breaker topology association matrix according to a power grid topology structure, and building a power grid fault diagnosis model by using said matrix as basic graphic primitive; fusing protection and breaker motion confidence probabilities according to the protection and breaker topology mapping relation; building rules to realize remote backup protection topology mapping according to remote backup protection cooperation logics and an outlet mode; selecting the maximum value to optimize the confidence probability according to the confidence probabilities of different protection inference paths; fusing the fault probability information according to topological graphic primitive probability distribution and port directions, thus obtaining the element integral fault probability; ranking the element fault probability result so as to determine the failure element, and carrying out protection and breaker motion evaluation.

The invention discloses a disease prognosis prediction system based on deep semi-supervised multi-task learning survival analysis. The disease prognosis prediction system comprises a data acquisitionmodule, a data preprocessing module, a prediction model construction module and the like. On the basis of a deep neural network model, a survival analysis problem is converted into a multi-task learning model composed of a semi-supervised learning problem of multi-time-sequence-point survival probability prediction; the model directly models the survival probability, does not depend on proportional risk hypothesis, can fit a time-dependent effect, and has better interpretability; a semi-supervised loss function and a sorting loss function are utilized to fit the data, complete data and deleteddata are fully utilized, and a traditional survival analysis problem and a survival analysis problem considering competition risks can be solved; according to the model, through multi-task learning of multiple time sequence points, data sharing among multiple prediction tasks is achieved, mutual constraint among the multiple prediction tasks is achieved, and the generalization ability of the model is improved.

The embodiment of the invention provides a medical knowledge graph construction method and device. The method comprises: acquiring diagnosis evidence information, clinical manifestation information and disease information; according to the diagnosis evidence information, the clinical manifestation information and the disease information, utilizing a preset diagnosis standard, a clinical standard and a disease diagnosis method to establish an association relationship between the disease and the multi-layer clinical manifestation and an association relationship between the clinical manifestationof each layer and the diagnosis evidence; and according to the association relationship between the disease and the multi-layer clinical manifestation and the association relationship between the clinical manifestation of each layer and the diagnosis evidence, constructing a medical knowledge graph containing the association relationship among the diagnosis evidence entity, the clinical manifestation entity and the disease entity. According to the method, clinical manifestation information is added into the medical knowledge graph, and the association relationship between diagnosis evidences and diseases is established through multi-layer clinical manifestation, so that the medical knowledge graph can fully reflect the differential diagnosis thought of doctors.

The invention provides a redundancy removal feature selection method LLRFC (Locally Linear Representation Fisher Criterion) score+ based on LLRFC and correlation analysis. A DNA (Deoxyribonucleic Acid) microarray technology provides a new direction for clinic tumor diagnosis. Performance of gene expression data corresponding to different kinds of tumor is different; through the analysis on the tumor gene expression data, study personnel can realize the accurate recognition on the tumor and the tumor subtype in the molecular level; and important biological significance is realized on the diagnosis and the treatment of the tumor. The feature genes in LLRFC judging criterion descending sort gene expression data is used to be combined with the dynamic correlation analysis strategy for further eliminating redundant features; an LLRFC score+ algorithm is provided; and the optimum feature gene subset is selected. The feature selection method LLRFC score+ has the advantages that the classification precision of a classifier can be effectively improved; a sample data set does not need to meet the normal distribution; and the method is applicable to data in various distribution types. The feature selection method LLRFC score+ can help people to find the virulence gene of cancer, and the early-stage diagnosis, tumor staging and typing, prognosis treatment and the like of clinic tumor diseases are facilitated.

The invention discloses a rapid ship target detection method, a storage medium and computing equipment, and the method comprises the steps: establishing a feature pyramid full-convolution network anda double-branch module which sequentially comprise an input layer, a feature extraction layer, a feature fusion layer and an output layer, determining a ship data set, inputting a generated training set into the built feature pyramid full-convolution network, and generating a target detection result. According to the method, the ship in the image can be rapidly and accurately detected; and the detection result is accurate and rapid, the requirement on embedded equipment is low, and the method has very high practical application value in various aspects such as military affairs, civil use and the like.

The invention relates to a defensive distillation model construction method and device. The method comprises the steps of obtaining original training data in a training set and a real label of the original training data; Performing knowledge distillation on the deep neural network according to the original training data and the real label, migrating knowledge of the deep neural network to the decision tree model, and generating a soft decision tree model; And performing countermeasure training on the soft decision tree model through the original training data to generate a defensive distillation model. The technical scheme of the invention is adopted; the decision tree model has good interpretability; the limitation of a defensive distillation technology on the framework requirement is broken through; the defensive distillation technology can be applied among different model structures, the application range of defensive distillation is expanded, and the scheme performs adversarial training on the soft decision tree model, so that not only is the problem of overfitting of the soft decision tree model prevented, but also the robustness of the model for adversarial samples can be improved.

The invention provides a machine room sketch identification method and device based on a legend library, and the method is characterized in that the method comprises the following steps: formulating asketch drawing standard, building a standard legend library, and carrying out the regularization constraint of the drawing of a machine room sketch; obtaining sketch data; performing hierarchical segmentation on the whole image in the machine room sketch to obtain a component diagram of the machine room sketch; performing matching identification by utilizing a standard legend library so as to realize detection and identification of the machine room sketch components; carrying out reasonable position constraint on the machine room sketch component; and performing parameterized output of results. The machine room sketch recognition method and device based on the legend library have the advantages that the standard graph recognition result matched with the input sketch can be output quicklyand accurately in real time, therefore a more visual reference basis is provided for designers to conduct quick design after the designers combine a sketch exploration report.

The invention mainly discloses a method and device for generating a prediction model for predicting pneumonia severity. The method comprises the following steps: acquiring clinical information of a subject; acquiring image information of the subject; screening N features out from clinical information and image information, wherein N is a natural number; constructing a corresponding classificationmodel by taking at least one of the N features as an input and the pneumonia severity as an output; training corresponding classification models by taking the N features of the plurality of subjects as training samples, and determining the classification model with the optimal fitting precision and parameter number in the plurality of classification models as the prediction model based on a presetcriterion. By adopting the scheme provided by the invention, the prediction model with low model complexity and high prediction accuracy is generated.

The invention provides a vehicle travel energy consumption prediction method and device. The vehicle travel energy consumption prediction method comprises the steps that route condition data of a target vehicle and vehicle condition data, driving behavior data and energy consumption data corresponding to the route condition data are recycled; extracting and training the relationship between the driving behavior data and the route condition data to establish a driving behavior prediction model; extracting and training relationships between the energy consumption data and the vehicle condition data, the route condition data and the driving behavior data to establish an energy consumption prediction model; when a target driving route is selected; acquiring route condition data of the target driving route, the driving behavior prediction model is utilized to predict the vehicle owner driving behavior under the target driving route according to the route condition data of the target drivingroute, and then the energy consumption prediction model is utilized to predict the energy consumption under the target driving route according to the route condition data of the target driving routeand the predicted vehicle owner driving behavior. Therefore, the vehicle travel energy consumption can be accurately predicted, so that a user can reasonably plan the vehicle power mode.

The invention discloses a calculating method and device for traffic transportation capacity based on track analysis, and belongs to the technical field of calculation of traffic transportation capacity. The device comprise a floating vehicle data cleaning module, a floating vehicle track clustering module, a traffic transportation capacity calculating module and a traffic transportation capacity scale evaluation module. Output information of all vehicles passing through an area in some time period is recorded, a track data set composed of single-point data is obtained for the n floating vehicle, and a complete data set is obtained through the floating vehicle data cleaning module; by utilizing the floating vehicle track clustering module, the directed space-time clustering volume meeting the requirements of a specific time sequence and space movement is obtained, a model of the relationship between the transportation capacity and common factors influencing the traffic transportation capacity is constructed, and an association coefficient of the model is extracted to obtain a transportation capacity factor loading matrix; the traffic transportation capacity calculation module isused for solving to obtain the total traffic transportation capacity, and the traffic transportation capacity is graded. The calculating method and device have reliability and scientificity and provide powerful basis for related decisions.

The invention discloses an advertisement putting method applied to the field of artificial intelligence. The method can also be implemented by applying a cloud technology. The method comprises the following steps: obtaining a report conversion number of a target advertisement in a historical time period; obtaining an estimated conversion rate accumulation value of the target advertisement in the historical time period; based on the reported conversion number and the estimated conversion rate accumulation value, determining an estimated conversion number corresponding to the target advertisement through a Bayesian function; determining a sorting result of the target advertisements according to the estimated conversion number; and if the sorting result meets the advertisement putting condition, putting the target advertisement. The invention further provides a related device, equipment and a storage medium. According to the invention, based on the Bayesian function, the estimated conversion rate accumulated value and the reported conversion number are jointly decided to estimate the conversion number; on one hand, the method has better interpretability, and on the other hand, the inherent deviation of the estimated conversion rate accumulated value can be reduced, so that the accuracy of the estimated conversion number is improved, the advertisement sorting accuracy is improved,and the advertisement putting effect is improved.

The invention provides a cloud-edge collaborative industrial control network safety protection method based on a strategy library. The method comprises the steps: generating an abnormal condition recognition rule library through a generative adversarial network (GAN) of a generative model network, namely a long short-term memory (LSTM) network, and a novel discrimination model network, namely a DendriteNetworks DD, at a cloud end; and then deploying the rules to corresponding edge ends, matching a corresponding coping strategy for each rule, and finally forming a cloud-edge collaborative strategy library to meet the low-delay safety protection of the industrial control network. A side end of the method comprises a data collection module and a strategy library module, and a cloud end comprises a data processing module, a generative adversarial network module, a rule generation module and a rule library module. Meanwhile, the rule base and the strategy base further comprise prediction rules generated by time sequence features extracted by the long and short-term memory network and recognition rules generated by the dendritic network. Policy corresponding to different rules can be set by an administrator of the industrial control system.

The invention discloses a real-time risk control variable calculation method based on a distributed streaming calculation engine. The calculation method comprises the steps that a data source configuration module is adopted, wherein a user configures data sources of various calculation dependencies; The variable definition module can be used for developing various variables by using SQL (Structured Query Language); The variable calculation module is used for capturing data from a pre-configured data source according to the calculation request, constructing a memory table, executing all SQL ina related variable pool once, and outputting a calculation result; According to the real-time risk control variable calculation method based on the distributed streaming calculation engine, the variable calculation engine and system researched and developed by the invention are adopted, personnel (such as business personnel and developers) familiar with SQL language can define variables through the system, the variable development period is greatly shortened, and the cost is saved.

The invention discloses a multi-dimensional fusion prospecting prediction method based on multi-source data, and relates to the field of prospecting prediction. Comprising the steps of obtaining multi-dimensional data of a research area; constructing a multi-dimensional prediction model; respectively inputting the pixel-level geological data, the space-level geological data and the structure-level geological data into a multi-dimensional prediction model; respectively predicting the geological data of the three dimensions through a multi-dimension prediction model to obtain prediction results of the three dimensions; corresponding weights are introduced into a multi-dimensional prediction model, different weights are given to prediction results of three dimensions, weighted fusion is carried out, and a final prospecting prediction result is obtained and used for representing a potential mineral resource area. According to the method, the multi-dimensional geological data of the multi-source data are predicted by constructing the multi-dimensional prediction model, and the prediction results are subjected to weighted fusion, so that multi-dimensional feature fusion of the multi-source data is realized, and the prediction precision of mineral product prediction and the reliability of geological interpretation are improved.

The invention relates to a degradation modeling and life prediction method considering performance index clustering in a dynamic environment, and belongs to the technical field of degradation modeling and life prediction. Firstly, test data are collected, then a performance index degradation model is established, unknown parameters in the model are estimated, and finally life and reliability prediction is carried out. The method comprises the following specific steps: 1, collecting test data; 2, establishing a degradation model; 3, estimating parameters, and updating the model in real time; and 4, predicting the service life and the reliability. For the phenomenon that each index of a multi-performance index product shows a clustering phenomenon due to the influence of public potential factors, mathematical modeling is carried out, so that degradation modeling for the product with the phenomenon is more practical, and a better explanatory is provided for the origin of correlation among the indexes.

The embodiment of the invention relates to the field of artificial intelligence, and provides a label classification method of corpora, which comprises the following steps of: carrying out word segmentation on multiple sections of text data of multiple sections of corpus data to obtain corresponding multiple sections of word segmentation results; inputting the multiple sections of word segmentation results into a probability model, and analyzing the word segmentation results by modeling of the probability model to obtain a plurality of K values; calculating the confusion degree of the plurality of K values, and taking the K value with the minimum confusion degree to obtain a corresponding first-level label; and inputting the corresponding multi-segment word segmentation results into a deformed bidirectional encoder representation model corresponding to the first-level label, and obtaining sub-labels under the first-level label through the deformed bidirectional encoder representation model. In addition, the invention further relates to blockchain technology, and the multiple segments of text data can be stored in the blockchain. The invention further provides a label classificationdevice for the corpora, computer equipment and a storage medium. Label classification accuracy of the corpora is improved.

The invention relates to the field of animal habitat evaluation, and discloses a method and device for evaluating the suitability of a seal habitat. The method comprises the following steps: acquiringa plurality of evaluation factors influencing the suitability evaluation of the seal habitat and data thereof; performing gridding processing on the data to form a plurality of evaluation factor griddistribution diagrams comprising a plurality of evaluation factor grid units; establishing a seal habitat suitability evaluation model on the basis; calculating according to the model to obtain the snoopard habitat suitability indexes of a plurality of evaluation factor grid units; grading the suitability indexes to obtain a grading result; and evaluating the reasonable degree of the grading result according to the plurality of evaluation indexes. The model established by the method provided by the invention is more reasonable in structure, so that a model evaluation result has a good practical application value, and the protection and planning of a seal habitat are facilitated; and the grading result of the suitability index of the seal habitat is reasonably evaluated according to a plurality of evaluation indexes, so that the comprehensiveness and the reliability of the evaluation can be improved.

The invention provides an intelligent subjective question marking method, system and device based on transfer learning, and the method comprises the steps: constructing a semantic feature extraction network based on an occlusion language model, and carrying out pre-training of the semantic feature extraction network in a self-supervised learning mode; constructing a multi-score evaluation model onthe basis of the pre-trained semantic feature extraction network, selecting examinee data evaluated by experts, and performing supervised fine adjustment training on the multi-score evaluation model;and scoring the examinee data to be scored by using the multi-score-point scoring model after fine adjustment training. According to the method, the problem of small sample learning can be solved, and the intelligent marking system constructed according to the method has high accuracy, high marking efficiency and certain interpretability.

The invention discloses a method and device for doubly judging bad short messages through a pre-training model and a short message address and relates to the technical field of information. The deviceis composed of a classified short message sample set, a pre-training model module, a short message collector, a text processor, an address extractor, a web crawler and a bad short message judger. According to the invention, the problem that traditional machine learning depends on features in bad short message recognition is solved; compared with deep learning, a large number of training sets arenot needed, judgment can be conducted through url short links in the short messages, and the short messages with sparse semantic information can be well recognized; meanwhile, the property of the short message judged by combining the text information and the short message address has better interpretability and more intuitive analysis effect than the property of the short message judged only according to the ip of the short message address.

The invention discloses a public opinion monitoring method and system using an aspect-based sentiment analysis method, and the method comprises the steps: constructing a crawler for crawling news comments, and crawling needed news and corresponding comment information; processing the news and the corresponding comment information, wherein stop words are removed, and word segmentation is conductedon sentences with the stop words removed; performing aspect-based sentiment analysis on the processed information by using a deep learning model, and analyzing sentiment of each aspect in the text; identifying a topic of each news text by using an LDA topic distribution model; and performing statistics on the theme of the news text and the emotion of each aspect in the news text. The emotion analysis can be carried out on the news comment text in a finer-grained manner, the emotion of each aspect in the news comment text can be judged, and the emotion of the news comment text can be judged more accurately.

The invention discloses an identification method and a system based on an integrated tree for false positive alarm signals in a ventilator. The identification method comprises the following steps: S1,data collection: collecting monitoring data of a patient from a ventilator and a monitor in a hospital; S2, data preprocessing: processing missing values and abnormal values in a dataset, conductingstandardization, and generating rules for identifying false positive alarm signals; S3, feature extraction: sorting features by using a random forest, and selecting the features with good recognitioncapability; and S4, false positive alarm signal identification: establishing a false positive alarm signal identification method for the ventilator and the monitor. Experimental results show that themethod of the present invention has excellent identification performance for false positive alarm signals and robust identification effect.

The invention discloses a mixed personalized recommendation method, which comprises the following steps that: obtaining film and television metadata capable of being recommended; carrying out similarfilm and television retrieval to obtain a candidate film and television; through sorting weight, carrying out similar sorting on candidate films and televisions; carrying out user behavior clustering,and generating a mixed recommendation candidate film and television according to a clustering result; and after all candidate films and televisions are sorted for multiple times, obtaining a recommendation result used for displaying. User behaviors are clustered to establish a user portrait, the user portrait obtained by clustering is used for generating the recommendation result, and personalized recommendation is carried out on the basis of the short-term watching historical behavior of the user and a hot broadcast film library, wherein films, television plays, variety shows and the like are mixed in the recommendation result. On the basis of content recommendation, the method is free from the influence of user-film incidence matrix sparseness problems, is explanatory and has a recommendation effect.

The invention provides a cross-domain heterogeneous large data multi-view clustering method and device based on tensor. The method includes: constructing sample tensor according to fused cross-domainheterogeneous feature space, and constructing feature space combination vector according to different context; accumulating the sample tensor to obtain a combined tensor; normalizing the order corresponding to each characteristic space of the merging tensor to obtain a connection tensor; obtaining the feature space scoring vectors according to the connection tensor, and obtaining a scoring tensorby outer product of the feature space scoring vectors; introducing the feature space combination vector and the scoring tensor into a high-dimensional space tensor distance to construct a combined scoring tensor distance; calculating sample similarity according to the combined scoring tensor distance, and constructing a view matrix according to the feature space combination vector; obtaining the multi-view clustering results under different views according to the view matrix. The invention solves the technical problem that in the prior art, different clustering results cannot be generated according to the requirements of different applications in different contexts, thereby providing high-quality clustering services for different applications.

The invention discloses a knowledge graph-based complex problem semantic understanding method, which comprises the following steps of: firstly, carrying out entity recognition and relationship extraction on a complex problem to obtain entity characteristics and relationship characteristics among entities, and generating a semantic graph of the complex problem; using the query graph prediction model to predict the semantic graph, and generating a query graph corresponding to the semantic graph; traversing in the knowledge graph according to the query graph, obtaining a query path, and obtaining an answer to the complex question according to logic calculation of the complex question. According to the method, input questions are converted into corresponding query structures, the method has excellent interpretability and accuracy, and complex questions can be accurately answered in combination with knowledge graph search.

The invention relates to the technical field of data classification prediction, and discloses a sandy soil earthquake liquefaction potential rapid evaluation method, which comprises the following steps: S1, selecting data from a data set commonly used in the field of sandy soil liquefaction as a first data set; s2, constructing a sandy soil earthquake liquefaction potential evaluation model; s3, inputting the data of the first data set into the model, performing domain adaptive training by using the maximum mean value difference, correcting the model, and performing data prediction on the second data set by using the corrected model; and S4, comparing the predicted sandy soil liquefaction condition with the actual sandy soil liquefaction condition, and carrying out validity verification on the sandy soil seismic liquefaction potential. The scheme improves the adaptability of the model on different data sets, solves the problems of low prediction efficiency, easy occurrence of an overfitting phenomenon of the model and the like in an existing scheme, and improves the prediction accuracy of the sandy soil seismic liquefaction potential. And the method shows higher accuracy and reliability in sandy soil earthquake liquefaction potential evaluation, so that the method has better interpretability and applicability.

The invention discloses a machine learning interpretable method and apparatus, and a computer readable storage medium. The method comprises the steps of firstly obtaining a to-be-interpreted sample; sampling the to-be-interpreted sample based on the feature correlation of the to-be-interpreted sample to obtain a sampling set comprising a plurality of sampling values; carrying out model training byutilizing the plurality of sampling values to obtain an interpretable model; and finally, explaining the to-be-explained sample by utilizing the explainable model to obtain an explaining result.

The invention discloses a color face recognition method based on quaternion non-convex penalty sparse principal component analysis, and belongs to the field of mode recognition and artificial intelligence. The method comprises the following steps: firstly, encoding red, green and blue channels of a color image by utilizing quaternion, constructing a complex representation form of a quaternion matrix, then calculating a quaternion covariance matrix, and carrying out characteristic decomposition on the quaternion covariance matrix to obtain a principal component vector; and introducing a non-convex penalty L1 / 2 norm as a sparse constraint term to obtain a new quaternion non-convex penalty sparse principal component analysis (QHSPCA) optimization model, and solving a sparse solution of the model by adopting a coordinate descent method and an immobile point iteration method, and finally realizing face recognition by using a nearest neighbor classifier. Experiments on a Geoga Tech face database show that the QHSPCA method provided by the invention has relatively good recognition performance, and the calculation efficiency is also improved.