159results about How to "Shorten forecast time" patented technology

The invention relates to a network public sentiment hotspot prediction and analysis method, which comprises the following steps: step (1) inputting public sentiment information collected in time into a hotspot public sentiment prediction model based on fast content identification; dividing the public sentiment information into hotspot public sentiment and ordinary public sentiment according to the processing results, and sending the pre-warning for the hotspot public sentiment; step (2), inputting the ordinary public sentiment into a hotspot prediction model based on numerical value expression, carrying out numerical value pattern matching on the input ordinary public sentiment information from the participating people number distribution and the time state distribution; and detecting the hotspot public sentiment information which is omitted in the detection in the first step; step 3, analyzing the hotspot public sentiment; and step 4, predicting the hotspot public sentiment. The invention combines the contents and the numerical values, and belongs to the integrated public sentiment hotspot monitoring method with the advantages of short prediction time and accurate prediction effect.

The invention discloses a ground daily rainfall predicting method based on satellite remote sensing and regression Kriging. The method comprises the steps that firstly, data are fast obtained through satellite remote sensing, and a regression relation among ground-based observation values, TRMM, DEMs and geographic positions of rainfall capacities of all levels is established according to the classification of the rainfall to obtain regression estimated values and regression residual errors of all levels; secondly, the spatial agglomeration degrees of the regression residual errors of all levels are analyzed, the trend removing is carried out on the regression residual errors, and the Kriging interpolation of the regression residual errors is carried out to obtain the regression residual error spatial distribution characteristics of all levels per 1 km; thirdly, the regression estimated values of all levels and the regression residual errors of all levels are added to obtain the ground-based predicting values of rainfall of all levels per 1 km; lastly, the ground-based predicting values of the rainfall of all levels are merged to obtain a daily rainfall predicting value per 1 km. According to the ground daily rainfall predicting method, the spatial and temporal distribution characteristics of the ground-based rainfall can be accurately predicted, the predicting precision of the ground daily rainfall is improved, the predicted space resolution is improved, and the key problem that the water conservancy department predicts the ground rainfall is solved.

The invention provides a water quality parameter time series prediction method based on relevance vector machine regression. The water quality parameter time series prediction method comprises the following steps of 1 acquiring water quality parameter historical data from an automatic water quality monitoring station and performing data pre-processing; 2 using front 2 / 3 data in the pro-processed water quality parameter historical data as a training sample set and using rear 1 / 3 data as a testing sample set; 3 using the training sample set to train an RVM, using the testing sample set to test the trained RVM so as to obtain a water quality parameter time series prediction model based on the RVM regression; 4 using the water quality parameter time series prediction model based on the RVM regression to predict new water quality parameters. The water quality parameter time series prediction method can perform time series prediction, is large in prediction range, high in accuracy and good in prediction stability, and can provide probabilistic output, give a predicted confidence interval while performing prediction, reduce the prediction time and timely observe water quality parameter change.

The invention provides a lithium ion battery health state prediction method. The lithium ion battery health state prediction method comprises the steps of: placing a lithium ion battery in a constant-temperature environment, and performing constant-current charging and discharging circulation on the lithium ion battery after standing for a period of time; after each charge-discharge cycle is performed for preset times, placing the lithium ion battery in a room-temperature environment for standing for preset time, and performing primary capacity calibration on the lithium ion battery; carryingout constant-current discharging on the battery at a certain multiplying power, measuring the alternating-current impedance of the lithium ion battery once after the state-of-charge value of the lithium ion battery is reduced to a set value, and establishing a dynamic impedance spectrum; establishing the equivalent circuit of the lithium ion battery according to the dynamic impedance spectrum, andfitting the dynamic impedance spectrum of the lithium ion battery according to the equivalent circuit to obtain fitting data; extracting the fitting data as an input parameter, and substituting the fitting data into a BP neural network model to obtain the health state of the lithium ion battery; by adopting the scheme, the health state detection reliability is improved, the prediction error is reduced, the prediction time is shortened, the data is simple and easy to obtain, and online detection can be achieved.

The invention provides a deep learning based video crowd emotion analysis method, and mainly relates to classification for crowd emotions in the video by using a multi-stream neural network. The method comprises the steps of building a multi-stream neural network (pixel, optical flow and saliency), concurrently extracting features in pixel information, superposition optical flow information and saliency information of a video sequence by using the network, and finally integrating the three types of features to obtain classification for the crowd emotions. The self-learning ability of deep learning is given into full play, the limitation of manual feature extraction is avoided, and the method is enabled to be higher in adaptability. Training and prediction are performed by using structural features of the multi-stream deep learning network, classification results of multi-stream sub-networks are integrated finally, and the accuracy and the work efficiency are improved.

The invention discloses a lightweight target detection method. The lightweight target detection method comprises the steps of 1, performing data enhancement processing on a sample image; obtaining a prior bounding box size of the network model; step 2, constructing a target detection network model: the target detection network model is based on YOLOv4, a MobileNetv3 network is introduced to reconstruct a feature extraction network, standard convolution is replaced by deep separable convolution in PANet, and model parameter quantity and operand are reduced; after convolution operation is carried out on the feature layer with the same channel number, an improved CBAM attention mechanism is integrated, and the network detection performance is further improved; 3, training a target detection network model; and 4, detecting by using the target detection network model to obtain a detection result. The method has the characteristics of improving the target detection efficiency and reducing the network prediction time.

The embodiment of the invention provides a training method and device for a semantic relation recognition model and a terminal, and the method comprises the steps: inputting a sample data set into aninitial pre-training model, outputting the representation information of sample sentences, and enabling the sample data set to comprise a plurality of sample semantic units; obtaining a plurality of feature words, and splicing the plurality of feature words to obtain representation information of the spliced feature words; inputting the representation information of the sample sentences and the representation information of the splicing feature words into an initial classifier, and outputting semantic relationship categories among the sample semantic units; adjusting the initial pre-training model and the initial classifier to obtain a new pre-training model and a new classifier; and establishing a semantic relation recognition model according to the new pre-training model and the new classifier. And the feature words are used as strong features in the chapter relationship, so that the classification effect on the specific semantic relationship can be improved. When the semantic relation recognition model is used for predicting the semantic relation category, the prediction time is shortened, and the prediction efficiency is improved.

The invention discloses a method for predicating the corrosion rate of soil of a transformer substation grounding grid. A current parameter analyzing method cannot well solve the grounding grid corrosion rate predicating problem. The method includes the steps that sampling points are reasonably distributed in a selected area; soil parameters are measured at the sampling points, the soil parameters comprise the resistivity, the moisture content, the redox potential, the pH value and the volume weight, a soil sample is collected back to a laboratory to conduct the measurement of salinity and the chloride ion physical and chemical property, and meanwhile an electrochemical method is used for measuring the corrosion rate of metal materials in the soil; the seven indexes of the resistivity, the moisture content, the redox potential, the pH value, the volume weight, the salinity and the chloride ion are used for setting up a model, and a BP neural network is used for training measured data. The method is fast, effective, accurate in predication and capable of providing a reliable basis for predicating the corrosion rate of the soil of the transformer substation grounding grid.

The invention proposes a feature selection method and system for network security data. The method comprises the steps of performing data normalization processing on a KDDCUP99 data set; performing Re-ReliefF data dimension reduction on a vector set; removing unrelated data or data with relatively low relevance to form a candidate feature set; and obtaining a feature with minimum relevance with the candidate feature set by utilizing an improved Re-ReliefF algorithm. According to the feature selection method and system for the network security data, provided by the invention, for redundant features existent in the data, redundant data in the data is removed by virtue of an MRMR (Maximum Relevance Minimum Redundancy) thought, so that the efficiency of a classifier is improved.

The invention discloses a building energy consumption prediction method based on a sub-metering time sequence, a system and a building. The prediction method comprises steps of acquiring and storing data of energy consumption and temperature of the building; using the acquired and stored data of energy consumption and temperature as input parameters for a time sequence analysis method; according to sub-metering and correlation analysis, using the trends of the energy consumption and the temperature and time factors predicted by the time sequence analysis method as main influence factors of the energy consumption of the building; and using the determined main influence factors and acquired energy consumption as parameters in a built BP neural network model to predict the energy consumption of the building in the future. According to the invention, the BP neural network is low in learning efficiency, slow in the convergence speed and quite sensitive to parameter selection, and the building energy consumption prediction algorithm based on the sub-metering and the time sequence is added on the basis of the BP neural network, so precision of energy consumption prediction is improved, prediction time is shortened, and predicted data is quite precise.

The invention discloses a power supply and distribution intelligent detection system for a ship. The power supply and distribution intelligent detection system comprises an on-off detection module, an electric quantity acquisition module, an industrial controller, a power supply module and a CAN (Controller Area Network) bus, wherein the electric quantity acquisition module comprises a three-phase intelligent ammeter and a current acquisition module; the on-off detection module and an electric parameter detection module are used for detecting the electric parameter of each transmission line of a main power distribution board, the on / off states of a circuit breaker and a fuse, the closure situations of the relay and the contactor, and the contact position of a control switch. Acquired data are uploaded to the industrial controller through the CAN bus, and the industrial controller is used for making analysis and judgment through circuit fault diagnosis software based on a continuous hidden Markov model and performing short-term state prediction on the power supply and distribution system of the ship through a prediction mechanism based on a grey model. The power supply and distribution intelligent detection system has the beneficial effects that by adopting modular design, high portability and high generality are realized; fault points are accurately located to fault devices, thereby increasing the detection accuracy; the grey model prediction mechanism is self-adaptive.

The invention discloses a prediction method for predicting the service life of a proton exchange membrane fuel cell. The prediction method comprises the steps of determining input network parameters:according to parameter data of a stack of the proton exchange membrane fuel cell, determining the number of input network types and the number of output network types; performing discrete wavelet transformation: carrying out discrete wavelet transformation on the parameter data, selecting basic wavelet types according to parameter data, and obtaining a denoised training sample through discrete wavelet transformation; and training an extreme learning machine neural network: determining the training sample and a test sample, and by utilizing an extreme learning machine, through randomly selecting an input weight and a deviation of a hidden layer, calculating an output matrix and an output weight of the hidden layer, and performing prediction through an inverse process of calculating the output weight by an evolutionary algorithm to output a prediction result. According to the method, the service life prediction precision of the proton exchange membrane fuel cell is high, and the prediction time is short.

The invention provides a short-term traffic flow prediction method based on a Spark platform. A parallel KNN algorithm is applied to the field of short-term traffic flow prediction. Compared with the conventional KNN algorithm based on single-computer computing, the problems of small system storage capacity and slow computing speed in performing data computing on the single physical computer can be solved by the method, and the problem of low neighbor matching efficiency in the neighbor search process of the KNN algorithm can also be solved by the method. According to the method, the computing efficiency of the algorithm can be enhanced under the premise of guaranteeing prediction precision, the practicality of the KNN prediction algorithm can be effectively improved and the system has great extensibility and speed-up ratio. The method also has reference meaning for other applications requiring large-scale data processing.

The invention discloses a wind power prediction method based on a deep convolutional neural network, and the method comprises the following steps: selecting and collecting the data of a wind power plant, and enabling the real coordinates of a wind driven generator to be mapped to a plane grid through employing a grid space embedding method; Filling the output of all wind turbines in the wind powerplant at a certain moment into a grid according to a mapping result to obtain scene characteristics corresponding to the moment, and arranging a plurality of continuous scene characteristics according to a time sequence to form a multi-channel image, namely, a space-time characteristic; Constructing three deep convolutional network models on the basis of the space-time characteristics to predictthe wind power; And analyzing and comparing the wind power prediction effect of each model. According to the method, STF in a multi-channel image form is constructed by embedding a grid space of a wind turbine in a wind power plant area, and the space-time transformation process of air flow is fully expressed; Three deep convolutional network models are provided, and each model can predict a largenumber of wind power of the wind turbine at the same time.

A railway wagon bottom floor damage fault detection method belongs to the technical field of railway wagon safety detection. The invention aims to solve the problems of poor reliability and low efficiency due to the adoption of manual image inspection for safety detection of the bottom floor of the existing rail wagon. The method comprises: collecting and processing whole-wagon linear array imagesof the bottom of a running railway wagon under different conditions to obtain a sample data set; constructing a semantic segmentation neural network model based on an encoder and a decoder, performing training by adopting a sample data set, finding an optimal weight coefficient, and updating the semantic segmentation neural network model; and then collecting a current whole-train linear array image at the bottom of the running rail wagon, obtaining a current to-be-detected target area by adopting the updated semantic segmentation neural network model, carrying out pre-judgment, and carrying out fault detection on the current to-be-detected target area which is pre-judged to have a fault by adopting a single-stage ssd model. According to the invention, automatic identification of bottom floor damage is realized.

The invention relates to the field of image recognition, and discloses a bone age prediction method and device, a medium and electronic equipment. The method comprises the steps: acquiring to-be-predicted hand bone data; determining the gender of the to-be-predicted hand bone data; according to the gender, inputting the to-be-predicted hand bone data into a convolutional neural network model corresponding to the gender to obtain a predicted bone age output by the convolutional neural network model, wherein the convolutional neural network model is obtained by training in advance based on a triple loss function and a regression loss function. According to the method, the bone age prediction is carried out by utilizing the convolutional neural network model obtained by training based on thetriple loss function and the regression loss function, so that the bone age prediction efficiency is improved, the bone age prediction cost is reduced, and the bone age prediction accuracy is improved. In addition, the hand bone data of the corresponding gender is predicted specially by using the convolutional neural network model of the specific gender, so that the bone age prediction accuracy isfurther improved.

The invention discloses a surrounding vehicle behavior adaptive correction prediction method based on a driving prediction field, which comprises the steps of: S1: carrying out surrounding vehicle behavior discretization and data set preprocessing, i.e., partitioning surrounding vehicle behaviors into N typical behaviors according to a transverse direction and a longitudinal direction; S2: acquiring traffic environment participation vehicle time series data, i.e., enabling each traffic environment participation vehicle to acquire a position, a speed and an acceleration of the vehicle at each moment in real time by using a positioning system; S3: establishing the driving prediction field, i.e., establishing the driving prediction field EP based on three elements of safety, efficiency and driving comfort, wherein EP=ES+EE+EC; S4: establishing a surrounding vehicle behavior prediction model on the basis of a maximum likelihood estimation method; and S5: carrying out surrounding vehicle behavior real-time prediction and model adaptive correction. According to the invention, safety, efficiency and driving comfort which influence driver behaviors are comprehensively considered; the driving prediction field is established in a driving region of a target vehicle and qualitative and quantitative analysis is carried out; and a new idea is proposed for surrounding vehicle behavior prediction.

The invention relates to a medical rehabilitation instrument. In order to achieve the purpose of accurately and fast forecasting angles of joints of lower limbs and controlling a myoelectric prosthesis, the technical scheme includes that a myoelectric prosthesis control source lead optimization method based on correlation coefficients comprises the following steps of extracting myoelectric signals of all muscles of a human body in processes of deep squatting, standing, extension of knee joints and walking; recording movement three-dimensional coordinates of the human body by utilizing a three-dimensional movement capturing system, and accordingly solving information of angles of the knee joints of the lower limbs of the human body; extracting a root mean square value of myoelectricity as a feature parameter and calculating the correlation coefficients of the feature parameter and the angels of the knee joints of the lower limbs; and sequentially removing uncorrelated muscle leads according to the correlation coefficients, establishing a lower limb muscle and bone kinetic model by utilizing an artificial neural network (ANN), predicting the angels of the joints, and comparing errors of different results so as to obtain the best lead optimization mode under different actions. The myoelectric prosthesis control source lead optimization method based on the correlation coefficients is mainly used for design and manufacture of medical rehabilitation instruments.

The invention discloses an intelligent logistics service system. The system comprises a data acquisition layer for acquiring data of a logistics service scene, a cloud computing processing center forcomputing, processing and analyzing the data of a specific logistics service scene, an external access interface provided and supported by the cloud computing processing center, a data storage centerfor storing various data required by a user, and an edge computing center. The edge computing center is provided with a unified access gateway used for receiving the data acquired by the data acquisition layer, and the edge computing center has a function of preprocessing the data acquired by the data acquisition layer and a function of predicting a specific scene or a special situation through amodel at the same time. Through the introduction of the edge computing center, the intelligent logistics service system can achieve the quick analysis and early warning of special conditions in the logistics service, also can improve the safety of logistics service management, and also reduces the analysis and prediction time of numerous and numerous logistics service data.

The present invention discloses an advertisement flow prediction method and device. The method includes the following steps that: a keyword to be tested of an advertisement to be tested is obtained; the keyword to be tested is compared with keywords in a reverse index table, so that a reference keyword matched with the keyword to be tested can be obtained, wherein the reverse index table is built in advance according to historical flow data corresponding to advertisements delivered in history; an advertisement corresponding to the reference keyword is determined as a reference advertisement, and reference advertisement flow corresponding to the reference advertisement is obtained from the reverse index table; and calculation is carried out by using the reference advertisement flow, so that predicted advertisement flow corresponding to the advertisement to be tested can be obtained. With the method and device adopted, the reference advertisement corresponding to the advertisement to be tested can be determined fast and conveniently, so that the predicted advertisement flow can be obtained by using the reference advertisement flow of the reference advertisement, and therefore, prediction time can be greatly shortened, and the real-time performance of the prediction of advertisement flow prediction can be enhanced, and guide for the delivery strategy of the advertisement to be tested can be benefitted.

The invention belongs to the artificial intelligence field and discloses an intrusion detection system and a method based on machine learning. Referring to the existing network security model and theintrusion detection model, the invention constructs an intrusion detection system frame based on machine learning according to the requirements in practical application. Firstly, a feature selection method is used to reduce the high dimension of security data in intrusion detection. Secondly, the particle swarm optimization artificial neural network algorithm is used to improve the detection accuracy. Thirdly, two typical clustering algorithms are used to eliminate false positives in intrusion detection. The intrusion detection system frame based on machine learning constructed by the invention adopts modular design, The invention presents a new intrusion detection method, which combines protocol analysis technology with clustering support vector machine to improve the detection efficiencyof clustering support vector machine. The algorithm is improved by computer software to further improve the detection rate and reduce the false positives rate.

The invention relates to a prediction method of pharmacokinetic property and toxicity of a drug molecule based on a support vector machine, which belongs to the molecule design field assisted by computers. The method fully takes advantage of the statistical learning modeling of the support vector machine, adopts an integrated method and simultaneously carries out the selection of a drug molecule descriptor and the optimization of SVM parameter. The method thereof comprises the following implementation steps: the descriptor is calculated and pre-treated, a descriptor data set is re-scaled, and the integrated method is adopted to carry out the selection of the descriptor and the optimization of the SVM parameter simultaneously. The optimization of the SVM parameter uses a conjugate gradient method to optimize penalty function C and kernel function Gamma. Genetic algorithm is used for selecting the descriptor and the individual fitness degree function adopts the fitness function which can comprehensively reflect prediction accuracy and the number of descriptors. In the integration of the selection of the descriptor and the optimization of SVM parameter, fitness degree function of each individual is calculated by SVM optimized parameter to complete the data integration of roulette, hybridization and mutation. The method fully takes two processing advantages of SVM and computer and significantly improves prediction result and efficiency.

The invention discloses a deep learning-based intelligent X-ray film diagnosis method. The method comprises the following steps: image data of chest X-ray films stored in a preset memory are acquired;the image data of the chest X-ray films are augmented in a preset mode; the augmented image data are subjected to preprocessing operation to enable data features to meet a standard; the preprocessedimage data are classified to form a plurality of classification models, and model training is designed for each classification model to generate a model training prediction result; a final CNN featuremap is acquired, and according to the CNN feature map, a thermodynamic map is generated; and according to the model training prediction result of each classification model, an integration result is calculated and obtained. Deep learning on pulmonary disease characteristics in a large number of chest X-ray films can be carried out, multiple pulmonary diseases can be accurately predicted, a diagnosis basis is provided, the prediction time is greatly less than manual diagnosis time, and the time of a doctor is saved.

The invention relates to a main contribution structure prediction method for vibration noise of an electric drive axle of a new energy automobile, which comprises the following steps: S1, establishinga multi-body dynamics calculation model of the electric drive axle based on a multi-body dynamics method, and calculating an excitation force caused by a gear transmission error and bearing contact;S2, establishing a dynamic finite element model of the electric drive axle based on a finite element dynamic method, and calculating the frequency response of the drive axle under the action of the excitation force calculated in the step S1; And S3, establishing a boundary element model of the electric drive axle noise radiation based on a boundary element theory, and performing numerical calculation on the electric drive axle noise radiation based on the data obtained in the step S2. According to the method, the prediction time of the main noise contribution structure is saved, the experimentcost is reduced, and the accuracy of the important noise contribution structure under the predicted peak frequency is improved.

The invention discloses a method for predicting a frequency spectrum of a CRN (Cognitive Radio Network) on the basis of a GCV-RBF neural network. The method comprises the following steps of: S1: acquiring channel historical data information; S2: using the channel historical data information as a preset input sample of an RBF neural network, training the RBF neural network by an OLS algorithm, and acquiring an optimal RBF neural network structure by a GCV evaluation method; and S3: according to the channel historical data information, by the optimal RBF neural network structure, predicting a current frequency spectrum state. Compared to the prior art, according to the invention, the optimal RBF neural network structure is obtained by the GCV evaluation method, so that the problem of overfitting in the training process is solved, and prediction accuracy is improved. Further, the RBF neural network structure, as a local approaching network, has the advantages of simple structure, high convergence rate, high real-time performance and the like, and can be sufficiently adaptive to changes of the network and improve self-adaptation of the network.

The invention discloses a regional house rent prediction method based on big data, which adopts FFM algorithm to carry out data cleaning, feature extraction, data conversion, feature modeling on a large amount of house rent information, predicts rent rental and finds out abnormal rental data by using the constructed model. The invention not only realizes the prediction of the house rent by cleaning the data, extracting the features and modeling, but also can detect the abnormal house rent information well. The rent forecasting method based on the FFM algorithm provided by the invention can well cope with the situation of sparse house data, can automatically learn hidden relations between features, and is a very effective method for rent forecasting.

The invention provides an IES incomplete data load prediction method and system based on C-GAN transfer learning. The method comprises the following steps: firstly, collecting original sample data andnormalizing the data; secondly, extracting sample features of the normalized sample data by adopting a depth variation self-encoding network; inputting the extracted sample features into a constructed first C-GAN generator; when a game of the generator and a discriminator reaches Nash equilibrium, expanding the incomplete sample data; inputting the expanded sample data set into a constructed generator of a second condition C-GAN; when the game of the generator and the discriminator reaches Nash equilibrium, predicting electricity, gas and heat loads in parallel; judging the prediction precision based on the C-GAN discriminator, continuously correcting and improving the prediction precision of comprehensive energy load prediction in the process that the generator and the discriminator playa game to achieve Nash equilibrium. The prediction system provided by the invention is used for load prediction, parameters required by network training are reduced, and meanwhile, the prediction time is shortened.

The invention relates to a microgrid short-term load prediction method based on empirical mode decomposition, and the method comprises the steps: S1, collecting original load data, carrying out the preprocessing, and obtaining a total load curve; S2, decomposing the total load curve into a trend load curve and a fluctuation load curve through empirical mode decomposition; S3, constructing a grey prediction model based on an amplitude compression method to obtain a trend load prediction value; S4, establishing an MIC historical matrix considering load similar day relevance and a UTCI historicallibrary including meteorological and geographical location factors, and obtaining a fluctuation load prediction value; and S5, reconstructing the trend load prediction value and the fluctuation loadprediction value, and obtaining a short-term load prediction value to control the working state of the distributed power supply in the microgrid. Compared with the prior art, the method has the advantages that the influence of meteorological and geographical location factors on load prediction is comprehensively considered on the basis of analyzing load characteristics, high prediction accuracy isachieved, and prediction speed is increased.