38results about How to "Short forecast time" patented technology

The invention discloses a neural network method for performing short-term earthquake prediction by using earthquake parameters, which is a method for predicting earthquakes by using regression analysis and neural network combined technology and comprises: (1) determining a region to be studied; 2) collecting and preprocessing earthquake catalogues; 3) acquiring the earthquake parameters and time varying parameters by using the earthquake catalogues; 4) sorting information; 5) performing regression analysis; 6) performing neural network simulation; and 7) computing earthquake magnitudes of predicted earthquakes. The method has the advantages that: regional short-term earthquake prediction can be realized and the prediction time is no more than 6 months; earthquake magnitude quantization prediction can be realized, the computing result is stable and the prediction precision is high; the maneuverability and practicality are high; and the regional short-term earthquake prediction result can be used to provide service for 'disaster prevention and reduction', and the social and economic benefits are obvious.

The invention provides an automatic electrocardiosignal classification method based on single channel. The method comprises the steps that a limb channel electrocardiosignal is received through Web socket; a wavelet median threshold method is adopted to perform electrocardiosignal denoising; a Pan-Tompkins method is used for conducting R wave detection, the electrocardiosignal is segmented based on R wave to resolve RR period characteristics, corresponding characteristics vectors are obtained by sequentially conducting empirical mode decomposition, Gaussian random projection matrix, polynomialfitting and interval extremum on obtained electrocardiogram segments; normalization is conducted on obtained characteristic features to allow the characteristic features to be conformed to standardized normal distribution, the standardized characteristic vectors are input to a trained XGboost model, and a corresponding detecting value is output. By means of the method, the problem of personal electrocardiosignal specificity is solved, the method is run at a server end, the pressure on a client is relieved, and the method has higher reference value on N and V type arrhythmias detection results.

The invention discloses a vehicle position information prediction method based on the Online-WSVR algorithm. The method includes the steps that 1, driving state information of a vehicle is collected in real time through an integrated navigation system in the vehicle; 2, modeling is carried out with the Online-WSVR algorithm according to the driving state information, in a previous period of time, of the vehicle, and weights are distributed to data at all time points so that the data can make different contributions to the modeling coefficient and accuracy can be improved; 3, position information, at the next moment, of the vehicle is predicted in real time with an Online-WSVR modeling function according to the current driving state information, wherein if current GPS signals are valid, the predicted longitude and latitude information at the next moment is erased, and if the current GPS signals are invalid, the longitude and latitude information, at the next moment, of the vehicle is predicted with the Online-WSVR algorithm and written into a training set for modeling to serve as a modeling sample for subsequent prediction. The method has the advantages that the principle is simple, the application range is wide, the positioning and prediction precision is high, the cost is low, portability is achieve, and reliability is good.

A method of building a set of experimental prediction model that requires fewer experimental frequency, shorter prediction time and higher prediction accuracy by using the advantages of combining the experimental data of Taguchi method and neural network learning is disclosed. The error between the experimentally measured result of photolithography and the simulated result of the theoretical model of near field photolithography is set as an objective function of an inverse method for back calculating fiber probe aperture size, which is adopted in the following Taguchi experiment. The analytical result of Taguchi neural network model of the present invention proves that the Taguchi neural network model can provide more accurate prediction result than the conventional Taguchi network model, and at the same time, improve the demerit of requiring massive training examples of the conventional neural network.

The invention discloses a disease predicting model construction method based on a gradient iterative tree. The disease predicting model construction method comprises the steps of preprocessing collected clinical data, adopting basic information and blood routine examination indexes to construct features; constructing a first predicting model based on a GBDT algorithm, labeling a data set of the first predicting model, adopting a training set to train the first predicting model, adopting grid search to adjust and optimize the parameters, and optimizing the first predicting model, wherein the first predicting model is used for predicting diseases and health conduction; constructing a second predicting model based on the GBDT algorithm, labeling a data set of the second predicting model, adopting the training set to train the second predicting model, adopting grid search to adjust and optimize the parameters, and optimizing the second predicting model, wherein the second predicting modelis used for predicting specific disease categories. By the adoption of the disease predicting model construction method, data can be rapidly labeled, the obtained disease predicting models have high predicting accuracy rate, and the predicting time is short.

The embodiment of the invention provides a terminal space-time movement predication method and device. The method includes: obtaining movement information of a terminal, wherein the movement information includes a cell to which a terminal is switched in a movement process, a switching time used for switching to the cell and a duration for staying in the cell; according to the obtained movement information of the terminal, generating a historical movement sequence of the terminal and updating a historical movement sequence set of the terminal according to the historical movement sequence; and at a to-be-predicted time point, based on a cell which the terminal is currently in and a current time point of the terminal, according to the corresponding cell in the historical movement sequence set and the switching time and duration of the cell, predicting subsequent movement information of the terminal. The terminal space-time movement predication method and device are used for optimizing a method for predicting a user-terminal movement path which changes with time.

The invention discloses a power short-term load predicating method based on fast periodic component extraction. According to the method, training data signals are subjected to spectral analysis, periodic and aperiodic components of the signals are sequentially extracted, then, the periodic components are subjected to cyclic predication, the aperiodic components are subjected to difference autoregressive moving average model predication, and the load condition of one day (the predication day) is obtained. The invention provides a novel fast detecting method for the short-term load predication of a power system, and solves the problem of the load short-term predication of a nonlinear system with complicated network structure and unstable parameters in the existing power system.

The invention relates to the field of precise prediction of dynamic trajectories, in particular to a method for predicting a dynamic trajectory of a moving object under a fixed air route task. The method comprises the following steps of in an offline state, defining a trajectory deviation sequence with a position label, constructing a two-dimensional container sequence based on the trajectory deviation sequence, and storing historical trajectory deviation data of the moving object under the same course task in the two-dimensional container sequence; in an online state, retrieving a forward known trajectory deviation sequence of the prediction object in the two-dimensional container sequence to obtain a sample set; adopting an online ISO algorithm, using the sample set, and establishing a trajectory deviation prediction model of the moving object online based on an RBF neural network structure; predicting the future trajectory of the moving object by using the trajectory deviation prediction model of the moving object; and repeating the steps 2 and 3 until the task is completed. The problem that an existing moving object trajectory prediction model obtained offline fails when the environment dynamically changes can be solved, and meanwhile, the trajectory prediction precision is improved.

The invention relates to the field of imaging medicine, in particular to a new coronal pneumonia frosted glass focus radiography analysis method and system and a storage medium, and the method comprises the following steps: S1, carrying out frosted glass shadow sketching based on imaging omics; S2, predicting the new coronal pneumonia probability based on deep learning; S21, preprocessing the frosted glass shadow obtained in the step S1, and improving the convergence rate and prediction speed of the convolutional neural network; S22, building a neural network based on the WRN and the mixed domain attention module; and S23, stacking the frosted glass shadow pictures preprocessed in the step S21 as input, and putting the frosted glass shadow pictures into the neural network in the step S22 to obtain the probability of suffering from the new coronal pneumonia. According to the method, a small convolutional neural network uses an attention module of a mixed domain, the training speed can be increased, and the prediction accuracy can be improved.

The invention discloses a city taxi demand prediction method. The method comprises the following steps: constructing an irregular prediction unit based on taxi historical passenger carrying data, constructing a multi-time slice data set, obtaining a taxi demand prediction data set, obtaining a space influence factor data set, and determining the influence intensity of different space influence factors; adopting a CNN-LSTM-ResNet fusion algorithm, and determining a parallel combination prediction model based on the CNN-LSTM-ResNet; and inputting the passenger carrying data and the space influence factors at the current moment into the parallel combination prediction model to determine the taxi demand of the irregular prediction unit. According to the method, the convolutional neural network is innovatively used for extracting the space influence factors except time, the multi-source space-time big data in the fields of population, nature and the like is selected for space influence factor quantification, and compared with single time sequence prediction, the prediction reasonability and precision are remarkably improved.

The invention discloses a network flow prediction method for optimizing an extreme learning machine through an improved cuckoo search algorithm and relates to the technical field of intelligent computing. According to the network flow prediction method for optimizing an extreme learning machine through an improved cuckoo search algorithm smaller prediction error and shorter prediction time can berealized by adopting real number coding to represent each parasitic nest and adopting a snap-drive cuckoo search algorithm to perform parameter optimization on the extreme learning machine. Accordingto the method, the operations of allocating solutions to bird eggs, rejecting non-search, selecting worst parasitic nests through probability search, globally searching, updating Pm, updating Pa and the like by executing an snap-drive cuckoo search algorithm are performed; the algorithm is high in optimization capacity, low in calculation complexity, high in calculation speed and convergence speed, capable of conducting global search and capable of jumping out of a local optimal solution can be known.

The invention discloses a multi-step wind power forecasting method based on singular spectrum analysis and locality sensitive hashing. The method is charactierzed by decomposing historical wind powerdata of a wind power plant into two independent components through singular spectrum analysis, the independent components being a low-frequency average trend component for reflecting wind energy overall change trend and a high-frequency fluctuation component for reflecting intermittency and fluctuation of wind respectively; reconstructing the two components in a phase space to obtain an average trend section and a fluctuation component section; and finding similar average trend sections of an average trend section to be forecasted through locality sensitive hashing, and carrying out locality predication. To prevent accumulated errors brought by separate forecast of each component and fixed error brought by forecast of only one component, the forecast input is combination of the similar average trend sections and corresponding fluctuation component sections, and finally, a prediction result of wind power output power is obtained. The method is clear in physical significance in wind power plant generation power prediction, short in prediction time and accurate and stable in prediction results; and the prediction results do not rely on prior knowledge of users.

The invention relates to a product sale prediction method based on a support vector machine model with parameter optimization. The method is characterized by including the following steps: S1, selecting a kernel function of the support vector machine; S2, adopting a grid search method to optimize predetermined parameters in the kernel function in S1; S3, establishing a prediction model; and S4, predicting a product sale trend, and applying historical product sales data to the prediction model in S3 to obtain a prediction result of product sale. The method provided by the invention greatly improves precision of sale prediction through a machine learning method, the SVM prediction model is short in prediction time, high in prediction precision and strong in robustness, avoids the circumstance that part of nonlinear models are easy to fall in the defects of local minimums and slow convergence rate, and thus the prediction model based on SVM optimization is effective and feasible. The prediction method provided by the invention overcomes the defects of poor precision and low calculation efficiency in traditional sale prediction, can provide a relatively accurate sale prediction reference for a decision-making level, and has good application value.

The invention discloses an industrial non-stationary process soft measurement modeling method based on a local weighting factor model. According to the method, a sliding window is introduced, a factoranalysis model is established in each sliding window, an online query sample is input into each sliding window to obtain a local similarity between the query sample and a training sample and local confidence of the query sample in the sliding window, then results of all the sliding windows are integrated to obtain a global weight of each training sample, a weighted average value of a training sample set is calculated according to the global weight, and the query sample is predicted on the basis of the weighted average value to obtain a butane content prediction value of the query sample. According to the invention, the accuracy of the prediction result is improved by establishing a weight relationship between an online measurement sample and the training sample.

The invention provides an internet event propagation outbreak prediction method and device, electronic equipment and a medium. The method comprises the following steps: setting a first media matrix; merging the Internet events according to minutes to obtain a plurality of event groups; monitoring each event group mode, wherein the modes comprise hot spots and non-hot spots; monitoring the minute propagation quantity of the hotspot mode event group and a second media matrix; wherein a time point when a speech center point appears and the same number of identifiers of the second media matrix andthe first media matrix reaches a second threshold value is a prediction starting point; obtaining a fast trend curve and a slow trend curve of the non-hotspot mode event group through the minute propagation quantity; monitoring minute propagation quantity and a third media matrix; wherein when the fast trend curve exceeds the set time of the slow trend curve, a speech center point appears and thesame number of identifiers of the third media matrix and the first media matrix reaches a fourth threshold value, the time point at the moment is a prediction starting point; and predicting the timeperiod of the event group from the prediction starting point to exponential-level outbreak through the prediction model. Prediction before outbreak is carried out.

The invention discloses an industrial high-order dynamic process soft measurement method based on a multi-hidden-layer weighted dynamic model. According to the method, a sliding window is introduced,and a multi-hidden-layer dynamic model is established in each sliding window for each group of online samples, so the local autocorrelation of data in a hidden space and the high-order dynamic relation of hidden variables in time sequence are fully considered, and the description of the data can be more accurate; in combination with a support vector data description method, a global weight of an online sample is calculated, and the multi-hidden-layer weighted dynamic model is established; after the parameters of the model are obtained, a locally weighted linear regression model is establishedso as to obtain a quality variable estimation value of the online sample.

The embodiment of the invention aims to provide a drug-target relation prediction method, and aims to solve the problems that in the prior art, target prediction by using machine learning is not high in accuracy, target prediction based on a graph model is poor in unknown drug-target relation prediction, and the prediction time is long. The invention provides a new similarity matrix of drugs, and between drugs and targets, and the similarity matrix is used to predict the relationship between the drugs and the targets. By utilizing the method, the target prediction accuracy can be improved, the time required for target prediction can be shortened, and the method has important significance on development of new drugs.

The invention discloses a power short-term load predicating method based on fast periodic component extraction. According to the method, training data signals are subjected to spectral analysis, periodic and aperiodic components of the signals are sequentially extracted, then, the periodic components are subjected to cyclic predication, the aperiodic components are subjected to difference autoregressive moving average model predication, and the load condition of one day (the predication day) is obtained. The invention provides a novel fast detecting method for the short-term load predication of a power system, and solves the problem of the load short-term predication of a nonlinear system with complicated network structure and unstable parameters in the existing power system.

The invention provides a real-time monitoring method for a clevidipine butyrate crude drug synthesis process. A Raman spectrum analysis technique is adopted, a model curve which can predict concentrations of a product clevidipine butyrate and a reactant chloromethyl butyrate in a synthesis reaction process is provided, and reaction degree and good and poor technological parameters can be judged. According to the technical scheme of the invention, operation is easy, detection is rapid, quantities of reactants and products in synthesized reaction solution can be evaluated in a few minutes, results are accurate, application range is wide, concentrations, which are predicted according to the model curve provided by the invention, of reactant and product components are close to concentrate data obtained by adopting high performance liquid chromatography, and the real-time monitoring method for the clevidipine butyrate crude drug synthesis process is not restricted to the clevidipine butyrate crude drug and also can be applicable to real-time monitoring of a synthesis process of other crude drugs.

The invention provides an internet event propagation outbreak prediction method and device, electronic equipment and a medium. The method comprises the following steps: setting a first media matrix; merging the Internet events according to minutes to obtain a plurality of event groups; monitoring each event group mode, wherein the modes comprise hot spots and non-hot spots; monitoring the minute propagation quantity of the hotspot mode event group and a second media matrix; wherein a time point when a speech center point appears and the same number of identifiers of the second media matrix andthe first media matrix reaches a second threshold value is a prediction starting point; obtaining a fast trend curve and a slow trend curve of the non-hotspot mode event group through the minute propagation quantity; monitoring minute propagation quantity and a third media matrix; wherein when the fast trend curve exceeds the set time of the slow trend curve, a speech center point appears and thesame number of identifiers of the third media matrix and the first media matrix reaches a fourth threshold value, the time point at the moment is a prediction starting point; and predicting the timeperiod of the event group from the prediction starting point to exponential-level outbreak through the prediction model. Prediction before outbreak is carried out.

The embodiment of the invention provides a terminal space-time movement predication method and device. The method includes: obtaining movement information of a terminal, wherein the movement information includes a cell to which a terminal is switched in a movement process, a switching time used for switching to the cell and a duration for staying in the cell; according to the obtained movement information of the terminal, generating a historical movement sequence of the terminal and updating a historical movement sequence set of the terminal according to the historical movement sequence; and at a to-be-predicted time point, based on a cell which the terminal is currently in and a current time point of the terminal, according to the corresponding cell in the historical movement sequence set and the switching time and duration of the cell, predicting subsequent movement information of the terminal. The terminal space-time movement predication method and device are used for optimizing a method for predicting a user-terminal movement path which changes with time.

The invention belongs to the technical field of hydraulic slide valve internal leakage detection methods, and particularly relates to a hydraulic slide valve internal leakage detection method based onan acoustic emission technology. The method comprises the steps: collecting hydraulic slide valve internal leakage detection experiment data, carrying out wavelet packet decomposition on acoustic emission signals, serving obtained sub-band energy characteristics as input, serving the corresponding slide valve state and internal leakage amount as output, and optimizing parameters of a radial basiskernel function; training an optimization detection model; verifying the accuracy of a classification model; Verifying the accuracy of a regression analysis model. According to the hydraulic slide valve internal leakage detection method based on the acoustic emission technology, wavelet packet decomposition is carried out on the acoustic emission signals of a normal slide valve and an internal leakage slide valve, energy characteristics of all sub-bands are extracted, and a hydraulic slide valve internal leakage diagnosis database is constructed; according to the hydraulic slide valve internal leakage detection method based on the acoustic emission technology, the accuracy is higher, the prediction time is shorter, and the efficiency of non-intrusive detection of hydraulic slide valve internal leakage is effectively improved.

The invention discloses a mobile application use behavior prediction method based on Entity Embedding and a TCN model. The method comprises: obtaining a mobile application use information original data set of a user; preprocessing the original data set of the mobile application use information; performing Entity Embedding on the classified data on the basis of an Embedding layer of a neural network, and constructing feature data; constructing a TCN network prediction model by taking the feature data as input; and obtaining an optimal TCN network prediction model through training and verification, and predicting a to-be-used mobile application. The influence of an App use sequence and a context environment on App use is comprehensively considered, an Entity Embedding feature extraction method and a TCN neural network are applied to mobile application use behavior prediction, the tedious feature processing process of a traditional machine learning model is avoided, feature data are extracted by using an Entity Embedding method, and the prediction capability of the TCN model can be improved by customizing the dimension of the feature data input into the TCN model.

The invention discloses an industrial non-stationary process soft sensor modeling method based on a local weighting factor model. The present invention introduces a sliding window, establishes a factor analysis model in each sliding window, inputs online query samples into each sliding window to obtain the local similarity between the query sample and the training sample and the local confidence of the query sample in the sliding window, and then The global weight of each training sample is obtained by synthesizing the results of all sliding windows, and the weighted average of the training sample set is calculated according to the global weight, and the query sample is predicted on the basis of the weighted average to obtain the predicted value of butane content of the query sample. The invention improves the accuracy of prediction results by establishing the weight relationship between online measurement samples and training samples.

The invention provides a multi-label long text classification method introducing a multi-way selection fusion mechanism, and relates to the technical field of multi-label long text classification based on sequence-to-sequence architecture. The invention improves the effect of completing multi-label long text classification based on the sequence-to-sequence architecture. Based on the data released by a machine learning challenge, the title data and description data are spliced ​​to obtain long text data. For data without description, copy a copy of the question as Describe, and then preprocess the data to remove low-frequency words to obtain more effective data. The obtained data uses a converter model that incorporates a multi-way selection fusion mechanism to generate a tag sequence for the input long text, and effectively removes redundancy during decoding. information. Under the test data, the tag sequence generated by the model has a recall rate of 0.5% compared with the model without multiple selection fusion; the precision rate and F1 value have increased by 1 percentage point.

The present invention provides a method for automatic classification of electrocardiographic signals based on a single lead, which receives limb lead electrocardiographic signals through Web socket; uses wavelet median threshold method to remove noise; then uses Pan-Tompkins method for R-wave detection, Cutting the signal based on the R wave and calculating the RR interval characteristics, the obtained ECG segments are sequentially subjected to empirical mode decomposition, Gaussian random projection matrix, polynomial fitting, and interval extreme value operation to obtain the corresponding eigenvectors; The vector is standardized so that it conforms to the standard normal distribution, and the standardized feature vector is input into the trained XGboost model, and the corresponding detection value is output. The present invention overcomes the problem of individual ECG specificity. At the same time, the method runs on the server side, reducing the pressure on the client side. This method has a high reference value for the detection results of N and V abnormal heart rhythms.