474 results about "Sequence prediction" patented technology

A method to predict equipment life is disclosed. The method includes making available a set of input parameters, and defining a model of a health of the equipment as a function of the set of input parameters. The method continues with receiving at least one signal representative of a respective one of an actual sensor output relating to an actual operation attribute margin of the equipment, predicting a remaining useful equipment life based upon a sequence of outputs of the model of the health of the equipment, and generating a signal corresponding to the remaining useful equipment life.

A multi-sensor data fusion early warning method for monitoring an electric transmission line tower comprises the following steps that 1) current electric transmission line parameters of detection sensors are obtained in real time; 2) first-class BP neural networks are fused; 3) second-class D-S evidence theories are fused, output of the BP neural networks is evaluated between 0 and 1 and processed to be elementary probabilities to serve as evidences of the D-S evidence theories, and danger classes are obtained. According to the second step, 2.1) fault monitoring is carried out on a single detection sensor: a BP network is adopted to construct a detection sensor output sequence predication model, and assuming n time output samples of the observed sensors as x(1), x(2), x(3),..., x (n), (n+1) time sensor output values are predicted; 2.2) the BP neural network of each detection sensor carries out time data fusion; 2.3) input signals are normalized; 2.4) fusion identification is carried out on characteristic layers. The multi-sensor data fusion early warning method is good in stability, high in reliability and good in real-time performance.

A method to predict equipment life is disclosed. The method includes making available a set of input parameters, and defining a model of a health of the equipment as a function of the set of input parameters. The method continues with receiving at least one signal representative of a respective one of an actual sensor output relating to an actual operation attribute margin of the equipment, predicting a remaining useful equipment life based upon a sequence of outputs of the model of the health of the equipment, and generating a signal corresponding to the remaining useful equipment life.

The invention discloses a climate time sequence forecasting method based on an empirical mode decomposition and support vector machine, belonging to the field of short-term climate forecasting. The climate time sequence forecasting method comprises the following steps of: firstly, pretreating a time sequence through an empirical mode decomposition method; decomposing the time sequence to a plurality of intrinsic mode function components and a trend component, wherein the components can more accurately reflect changes in the original sequence and keep characters of the time sequence per se; next, carrying out phase space reconstruction on each component through a time sequence forecasting method; respectively establishing different support vector machine regression models for forecasting; and combining the forecasted result of each component to the forecasted result of the original sequence. The invention has the advantages of getting help from the empirical mode decomposition method for smooth processing of the time sequence, reducing interference or coupling information among the sequences on the basis of keeping the characters of the time sequence per se, enabling the accuracy of forecasting to be higher, and especially fitting for treating non-stationary climate time sequences with yearly precipitation or changed temperature.

Techniques are described for generating predictions of future traffic conditions at multiple future times, such as by using probabilistic techniques to assess various input data while repeatedly producing future time series predictions for each of numerous road segments (e.g., in a real-time manner based on changing current conditions for a network of roads in a given geographic area). In some situations, one or more predictive Bayesian models and corresponding decision trees are automatically created for use in generating the future traffic condition predictions for each geographic area of interest, such as based on observed historical traffic conditions for those geographic areas. Predicted future traffic condition information may then be used in a variety of ways to assist in travel and for other purposes, such as to plan optimal routes through a network of roads based on predictions about traffic conditions for the roads at multiple future times.

The invention provides a method and system of the prediction of time series data. The method comprises the steps that wavelet decomposition is conducted on the sequence formed by t-1 moment data, n subsequences are obtained; stationary detection is conducted on n subsequences respectively; for non-stationary time series, an advance learning LSTM model is built using the t-1 moment data, and the values of t moment are predicted respectively, and forecasts of the non-stationary part are obtained by summing; similarly, for stationary sequences, ARMA models are respectively built and the values of t moment are predicted, and the forecasts on stationary part are obtained by summing; finally the prediction values of the non-stationary part and the stationary part at t moment are summed, to obtain the final forecast value. By the method and system of the prediction of time series data, by wavelet decomposition, the advantages of LSTM and ARMA are combined, in comparison with traditional methods, better effects are provided in dealing with non-stationary time series. In addition, by benefiting from the unique LSTM structure in the model, the forecasting and the generalization ability of the method is better, and the method is suitable for time series prediction in various fields.

The invention discloses a wind farm risk evaluation method based on a Monte Carlo method. The wind farm risk evaluation method includes the steps: (1) building a wind speed sequence prediction model on the basis of a time sequence method by means of original data standardization, model recognition and parameter determination; and (2) evaluating the reliability of a large power grid system concerning a wind farm on the basis of the sequential Monte Carlo method by means of random sampling and reliability evaluation. Randomness of the wind speed is adequately considered on the basis of a large quality of wind power databases, and the method is high in speed by the aid of the Monte Carlo high-speed algorithm.

Methods, systems and computer readable code for forecasting commodity consumption and for forecasting time series are provided. According to some embodiments, the forecasting includes deriving at least one population commodity consumption forecasting model from population historical consumption data, deriving an individual commodity consumption forecasting model for at least one individual of the population from at least one population commodity consumption forecasting model and from individual historical consumption data, and forecasting future individual commodity consumption for the individual using individual commodity consumption forecasting model. According to some embodiments, the presently disclosed forecasting includes forecasting future values of an individual time series within a population of time series, where each time series on the same domain. Thus, according to some embodiments, the forecasting includes deriving at least one population forecasting model from past values of the population of time series, deriving an individual time series for at least one individual time series forecasting model from past individual time series values and from at least one population forecasting model and forecasting future values of said individual time series using the individual time series forecasting model.

In various embodiments, a natural language (NL) application receives a partial NL request associated with a first context, and determining that the partial NL request corresponds to at least a portion of a first next NL request prediction included in one or more next NL request predictions generated based on a first natural language (NL) request, the first context associated with the first NL request, and a first sequence prediction model, where the first sequence prediction model is generated via a machine learning algorithm applied to a first data dependency model and a first request prediction model. In response to determining that the partial NL request corresponds to at least the portion of the first next NL request prediction, the NL application generates a complete NL request based on the first NL request and the partial NL request, and causes the complete NL request to be applied to a data storage system.

The present disclosure is directed towards a prefetch controller configured to communicate with a prefetch cache in order to increase system performance. In some embodiments, the prefetch controller may include an instruction lookup table (ILT) configured to receive a first tuple including a first instruction ID and a first missed data address. The prefetch controller may further include a tuple history queue (THQ) configured to receive an instruction / stride tuple, the instruction / stride tuple generated by subtracting a last data access address from the first missed data address. The prefetch controller may further include a sequence prediction table (SPT) in communication with the tuple history queue (THQ) and the instruction lookup table. The prefetch controller may also include an adder in communication with the instruction lookup table (ILT) and the sequence prediction table (SPT) configured to generate a predicted prefetch address and to provide the predicted prefetch address to a prefetch cache. Numerous other embodiments are also within the scope of the present disclosure.

The invention discloses an aquaculture water quality short-time combination forecast method on the basis of multi-scale analysis. The method includes the steps that water quality time sequence data are acquired online and repaired; through empirical mode decomposition, the selected water quality time sequence sample set data are decomposed into IMF components and residual rn components, wherein the IMF components and the residual rn components are different in frequency scale; the IMF components and the rn components are classified, a manual bee colony optimization least square support vector regression machine, a BP neural network and an autoregressive sliding average model are respectively selected for forecast according to classifying features, and finally, all results are weighed and summed to obtain a water quality time sequence forecast result. According to the method, the original water quality time sequence data are decomposed into the components different in time frequency through the empirical mode decomposition, and change conditions in original water quality sequences can be mastered more accurately; advantages of the manual bee colony optimization least square support vector regression machine, advantages of the BP neural network and advantages of the autoregressive sliding average model are complemented and combined, and thus performance of a combined forecast model is effectively improved.

The invention discloses a vehicle-mounted network intrusion detection method based on message sequence prediction. The vehicle-mounted network intrusion detection method comprises the following steps:step 1, collecting reverse control over CAN bus data and private protocols related to an automobile bus, an automobile body bus and a power transmission system bus through a T-BOX vehicle-mounted terminal; step 2, forming an operation scene according to three attack ways of automobile information security to analyze possible security threats of the vehicle; step 3, performing learning training according to the data set obtained in the step 1 and the abnormal message feature library obtained in the step 2 to form an evaluation detector; and step 4, carrying out detection verification on the input message through an evaluation detector. According to the vehicle-mounted network intrusion detection method based on message sequence prediction provided by the invention, through setting from thestep 1 to the step 4, an attack type message can be learned by a system and then identified, so that an anti-attack effect is realized.

The invention relates to a traffic flow prediction method based on a multivariable grey model time sequence. The method comprises the following steps: inputting collected observation station traffic flow, related external variable data and observation station information data; performing data preprocessing on the input data; inputting the data subjected to data preprocessing into a multivariable time sequence fusion prediction model based on data decomposition and a multivariable time sequence fusion prediction model based on result weighting for prediction; and comparing the predicted value with the actual value, and outputting a final result. The traffic flow of the expressway is predicted through fusion of multiple multivariable time sequence prediction models, the prediction precisionis improved, and through application to the expressway in the traffic field, the traffic management department can be helped to improve the intelligent management level, and the operation cost is reduced; through the display of the application demonstration system, data support can be visually provided for managers so as to make corresponding decisions in time and implement the decisions.

A recording position error measurement apparatus includes a read image signal acquisition device and a signal processing device having: a dividing device which divides pixel series of the read image signal into sequences having different remainder values so as to generate image signals of the respective sequences; a prediction signal generation device which calculates regular prediction signals which are predicted with respect to the respective sequences, according to the read image signal; a threshold value determination device which determines tone value differences corresponding to respective distances representing recording position errors from the prediction signals, and which determines threshold values corresponding respectively to the recording position errors, from the tone value differences; a change signal calculation device which calculates a change signal indicating a difference between the prediction signal and the image signal of each of the sequences; and an error distance calculation device which specifies the recording position errors of the plurality of recording elements in the recording head according to comparing the change signal with each of the threshold values.

The invention relates to a lithium ion battery remaining useful life (RUL) prediction method based on a fusion algorithm and belongs to the technical field of battery management. The method comprisesthe following steps: S1, acquiring battery capacity attenuation data, and determining parameters of a RUL prediction model based on an optimal control algorithm; S2, fitting data of a training set, iteratively outputting an optimal control algorithm mode parameter filtering estimation value and a battery capacity attenuation data filtering estimation value, and obtaining an initial RUL predicted value by virtue of the model parameter filtering estimated value; S3, based on a difference value between the filtering estimation value of the optimal control algorithm and experimental data, buildingan original error sequence, taking the original error sequence as an input of a neutral network algorithm, performing continuous iteration training on the error sequence, and outputting a predictionresult of the error sequence; and S4, after the data of the training set is used, obtaining a final lithium ion battery RUL prediction result by synthesizing an initial predicted value of the optimalcontrol algorithm and the error sequence prediction result obtained by virtue of the neutral network algorithm.

The invention discloses a hydrological time series prediction method based on a multiple-factor wavelet neural network model. The invention provides a multiple-factor wavelet neural network model used for predicting the hydrological time sequence. The model takes a multiple-time sequence message as input, and the multiple time sequence message not only comprises the current wavelet coefficient of a prediction target time sequence but also comprises the current wavelet coefficient of other time sequences relevant to the time sequence; mutual information between the multiple-time sequence message and the prediction target time sequence serves as a measurement for judging the relevance of the multiple-time sequence message and the prediction target time sequence; other time sequences of strong relevance are selected; and a wavelet function selection criteria based on the a coefficient of weighted correlation is further utilized to select the optimal wavelet function for the model. Compared with the prior art, the method disclosed by the invention has the advantages of higher prediction accuracy and better expandability and practical value.

The invention belongs to the technical field of data processing, and discloses an LSTM neural network cycle hydrological forecasting method based on mutual information, which comprises the following steps: screening and classifying original data through mutual information analysis, and taking rainfall, reservoir water level and flow hydrological characteristics as input characteristics of a long-term and short-term memory cycle forecasting model; the long-term change of flood is reflected by simulating rainfall process training and determining the structure of the LSTMC model; and verifying the output of the model by using the actual flood data. According to the method, the data set is analyzed by adopting a mutual information-based method, the flow at the current moment and each hydrological characteristic of the previous longer time period are fully captured, and the input characteristics of the model are dynamically selected. According to the method, the deep learning algorithm is utilized, the cyclic prediction model based on the LSTM neural network is adopted, when the method is used for flood flow time series prediction, the problem that the hydrological change process is greatly influenced by factors in the earlier stage is solved, and effective features can be automatically captured well.

The invention discloses a tunnel event detection method based on integrated learning time sequence prediction, which mainly solves the problem that similar methods fail to accurately predict values of a sensor and cannot effectively control tunnel ventilation. The method comprises the detection steps of: pre-processing acquired highway tunnel data to form a training data set; training a plurality of basic predictors according to the training data set, and forming a strong predictor by the weighted combination of the basic predictors; utilizing the strong predictor to calculate a predictive value of smoke concentration of a tunnel at t+1 time according to a value of a tunnel sensor at the current t time, and dynamically adjusting the basic predictors which take part in the integration according to prediction error; comparing the predictive value of the smoke concentration of the tunnel at t+1 time with a smoke concentration threshold, and judging whether the smoke concentration is an over-standard event; and for the over-standard event, calculating control parameters of a ventilation controller, and reducing the smoke concentration. The method has the advantages of strong prediction function and high control accuracy of the ventilation controller, and is used for operation monitoring, energy conservation and emission reduction of highway tunnels.

The invention provides a core user mining method based on a deep neural network and a graph network. The core user mining method comprises the steps of constructing a user- game history information database; Performing data preprocessing; According to the game historical sequence observation data of the game user after data preprocessing, establishing a directed graph with a game name as a node and a time sequence as an edge, and inputting the directed graph into a graph network embedding method so as to predict a game which is interested in next time; And establishing the directed graph for each game user to obtain an expression of each game, carrying out feature splicing on the obtained expression of each game and the personal information of the corresponding user, and fusing and inputthe expression and the personal information into a deep neural network so as to predict whether the user is a core player of the game or not. According to the invention, the problem of sequence prediction is solved based on the fusion method of graph network embedding and the deep neural network, the time sequence information is fully learned in the form of the graph network, and higher-level interactive expression is learned by fusing the deep learning method, so that the model prediction accuracy is improved.

The invention relates to an air quality index prediction method, in particular to an air quality index prediction method based on spatial and temporal distribution characteristics. The method adopts a time series prediction method to predict time orientation, i.e. an air quality index at certain time in the future; and then, a Kriging interpolation method is adopted, and the known latitude and longitude coordinates of a monitoring site are combined with a time series prediction result to carry out interpolate estimation on the air quality index of any place of a whole area. Therefore, the air quality index prediction method has the following advantages that the complexity of a model is lowered, whole calculation time is shortened, and meanwhile, the accuracy of the model is guaranteed; and in addition, the air quality indexes in multiple time periods in one day at each place within the area can be more accurately predicted.

The invention discloses a wireless intelligent alarming system for automatically monitoring multivariate information of a tunnel. The wireless intelligent alarming system is characterized by being realized by the following steps of: firstly, identifying surrounding rock multivariate parameters based on a differential evolution algorithm and a support vector machine; secondly, predicating a surrounding rock multivariate information time sequence based on the differential evolution algorithm and the support vector machine; and thirdly, carrying out judgment and alarm on the safety of the surrounding rock according to filed monitoring information and an information predication result obtained in the second step. According to the system, the requirements on a field communication cable and the field processing capability are reduced, and thus an acquisition-transmission instrument of a filed part is integrated on a circuit board; a wireless data transmission technology is adopted on site and field monitoring data is comprehensively and smoothly transmitted to a data processing system for an ultra-long distance; and an analytic result alarms through short messages, emails and qq, so that the pertinence, the flexibility and the reliability of alarm are improved. The direct contact between relevant persons and complex and danger surrounding rock field can be avoided and status information of the surrounding rock can be safely and remotely acquired.

Disclosed are methods and apparatus for predicting an expected value associated with an end event of an event sequence. In one embodiment, the following operations are performed: (a) providing a current set of input attributes and contextual data collected during performance of previous event sequences; (b) predicting a plurality of expected values for going from a first event of a known event sequence to each of a plurality of subsequent events of the known event sequence based at least on the current set of input attributes and the collected contextual data; and (c) predicting an expected value for going from a first event of an unknown event sequence to an end event of such unknown event sequence based on at least two of the expected values predicted for the known event sequence and based at least on the current set of input attributes and the collected contextual data. The expected value for reaching the end goal of the unknown event sequence cannot be determined with a degree of certainty that is higher than a predetermined value and wherein the expected value for reaching the end goal of the known event sequence can be determined with a degree of certainty that is higher than the predetermined value.

The invention discloses a prediction model selection method based on the applicability quantification of a time series prediction model, and relates to the field of time series prediction model prediction. The invention aims at solving problems that a conventional time series characteristic prediction method is small in number of prediction angles of prediction results outputted by a prediction model, cannot achieve the complete and comprehensive prediction of the performances of the prediction model, and causes poor prediction effects. According to a prediction step P, a true value xk and an output result (shown in specifiction) of each prediction model, the method obtains the errors and prediction efficiencies of all prediction models. According to prediction demands, the optimal prediction model meeting the prediction demands is selected from m prediction models through combination of the errors and prediction efficiencies of all prediction models. If the number of prediction models meeting the prediction demands is one, the prediction model is the optimal prediction model; if the number of prediction models meeting the prediction demands is greater than one, the verification of the difference of prediction capability is carried out between the prediction models, thereby obtaining the optimal prediction model. The method can be used for the prediction of the prediction models.

A system and a method for time series forecasting. The method includes: providing input feature vectors corresponding to a plurality of future time steps; performing bi-directional long-short term memory network (BiLSTM) on the input feature vectors to obtain hidden outputs corresponding to the plurality of future time steps; for each future time step: performing temporal convolution on the hidden outputs using a plurality of temporal scales to obtain context features at the plurality of temporal scales, and summating the context features at the plurality of temporal scales using a plurality of weights to obtain multi-scale context features; and converting the multi-scale context features to obtain the time series forecasting corresponding to the future time steps.

The invention provides a grey support vector machine-based multi-stress accelerated life testing forecasting method, and belongs to the technical field of life forecasting. The method comprises eight steps, namely acquisition of multi-stress accelerated life testing data, determination of reliability with an empirical distribution function method, level ratio inspection of product failure time data, accumulated generating operation (AGO) of the product failure time data, construction of a support vector machine forecasting model, forecasting with the constructed support vector machine model, reduction of an AGO generating sequence forecasting value and life distribution fitting. The grey support vector machine-based multi-stress accelerated life testing forecasting method provided by the invention can be used for forecasting without knowing the information such as a specific accelerated model and product life distribution and the like, so that difficulty in establishing the accelerated model and introduction of system errors in forecasting are avoided, a complex multiplex likelihood equation set does not need to be solved, and the method has stronger engineering applicability and universality for different products or stress categories.

The invention discloses a method for predicting the residual service life of a rolling bearing based on EEMD-MCNN-GRU. The method comprises the following steps: preprocessing full-life-cycle vibrationdata of the rolling bearing; utilizing the ensemble empirical mode to construct degradation characteristics of the bearing vibration signals on different time scales so as to construct a degradationcharacteristic data set of each rolling bearing; integrating a multi-scale convolutional neural network feature extraction layer and a gate control cycle unit neural network time sequence prediction layer to build a residual life prediction model; normalizing the degradation characteristic data set input into the prediction model; carrying out segmentation processing on the normalized data; dividing the degradation feature data set into a training set and a test set, and taking the training data as the input of the whole prediction network model; and smoothing the prediction model output by using a moving average method, and outputting an optimal life prediction result. The prediction method which is more reliable and higher in robustness and generalization is provided for analysis of theresidual service life of the rolling bearing.

The invention provides a machine learning method used for electrolytic-cell fault early-warning. The method is used for establishing an anticipation model for electrolytic-cell faults. A main processof the method includes: extracting detection point sequence data; preprocessing the data; inputting a training data set to a GMM clustering model; defining abnormality discrimination rules; optimizingdiscrimination parameters; improving the GMM clustering model; and evaluating a fitting effect of the trained model. The invention also provides an application of the machine learning method used forelectrolytic-cell fault early-warning. A main process of the application includes: extracting new detection point sequence data; preprocessing the data; predicting a time sequence; and carrying out early-warning fault judgment of the trained model. According to the method and the application, benumbing of traditional conditional value alarming on operators can be effectively reduced, experiencedoperators can be replaced for judging the faults, and error judgment of human factors can be avoided.

A recording position error measurement apparatus includes a read image signal acquisition device and a signal processing device having: a dividing device which divides pixel series of the read image signal into sequences having different remainder values so as to generate image signals of the respective sequences; a prediction signal generation device which calculates regular prediction signals which are predicted with respect to the respective sequences, according to the read image signal; a threshold value determination device which determines tone value differences corresponding to respective distances representing recording position errors from the prediction signals, and which determines threshold values corresponding respectively to the recording position errors, from the tone value differences; a change signal calculation device which calculates a change signal indicating a difference between the prediction signal and the image signal of each of the sequences; and an error distance calculation device which specifies the recording position errors of the plurality of recording elements in the recording head according to comparing the change signal with each of the threshold values.