74results about How to "Solve the low prediction accuracy" patented technology

The invention discloses a medium and long term hydrologic forecasting method based on empirical mode decomposition. The medium and long term hydrologic forecasting method based on empirical mode decomposition comprises the steps that firstly, a hydrologic forecasting model is built according to the following procedures of (101) empirical mode decomposition, wherein empirical mode decomposition is carried out on a hydrologic time sequence s(t) of a forecast drainage basin, (102) kernel principle component analysis, wherein kernel principle component analysis is carried out on n intrinsic mode function components Fj obtained through empirical mode decomposition and a trend item rn, and p main components F'k are extracted, (103) building of a training sample set, wherein the training sample set is built according to the extracted p main components F'k, (104) building of a support vector machine model, and (105) training of the support vector machine model; secondly, annual runoff data of the years needing forecasting are forecast through the built hydrologic forecasting model. The medium and long term hydrologic forecasting method based on empirical mode decomposition is simple in step, convenient to realize, easy and convenient to operate, good in use effect and capable of effectively resolving the problem of low forecasting accuracy of an existing hydrologic forecasting method.

The invention provides a coupling large-scale data flow width learning rapid prediction intelligent algorithm based on network community detection and GCN. The algorithm comprises the following steps:step 1, community detection; step 2, space-time feature extraction; step 3, width learning rapid prediction; and step 4, large-scale real-time prediction of the space-time coupling width learning neural network. Compared with the prior art, the algorithm has the beneficial effects that intelligent community detection and GCN feature extraction are adopted, width learning is combined, the problemof large-scale node prediction is solved, and the algorithm has the advantages of being high in calculation speed, high in prediction precision, high in adaptive capacity and the like.

The invention discloses a medicine sales forecasting method based on a BP neural network and an ARIMA combination model; the problem of low forecasting accuracy of a single forecasting method based on a traditional research method or an artificial neural network is solved. This method first uses the ARIMA model to predict the historical annual sales volume of a certain type of drug, and its linear law information is included in the prediction error of the ARIMA model, and then uses the BP neural network to predict the error of the ARIMA model so that its nonlinear law is included in the prediction error of the ARIMA model. In the prediction results of BP neural network. Finally, the prediction result of ARIMA and the prediction of BP neural network are added to obtain the prediction value of the combined prediction model; this method can overcome the defects of the time series method in predicting drug sales to a large extent, and significantly improve the prediction accuracy of drug sales; it can be compared Good forecasting of drug sales can be used as a method for predicting future drug sales; the process of predicting drug sales can be easily realized through Eviews software, which is practical and easy to promote and apply.

The invention discloses a combined wind power prediction method based on wind speed fluctuation characteristic extraction. The combined wind power prediction method includes the following steps that wind speed data acquired by training samples are normalized; time windows are established for the normalized wind speeds, and multifractal spectrum analysis is performed in the time windows; the widths omega of singular index alpha value taking intervals of the time windows and symmetry parameters S of peak value differences Deltaf (alpha) and f (alpha) of a singular spectrum function f (alpha) are analyzed and compared. The wind speeds are classified according to the parameters [omega, Delta f (alpha), S], and the sizes of the time windows are further adjusted. Divided categories are sequentially trained by using an extreme learning machine, a support vector machine and an optimization regression power curve method, average monthly precision comparison is conducted on produced prediction results, one of the methods is selected to serve as an optimum single algorithm for the categories, and trained models are obtained. Same classification and modeling are conducted on test samples, corresponding optimum single algorithms are selected for different models for respective prediction, and finally final prediction results are obtained through combination.

The invention discloses a novel load prediction method and device based on deep learning. The method comprises the following steps that: obtaining the input variable of a power grid to be predicted, wherein the input variable is used for indicating the parameter of the power grid to be predicted; inputting the input variable into a trained power grid load prediction model for a model operation soas to obtain an operation result; and according to the operation result, determining the load of the power grid to be predicted. Through the method, an effect that the prediction accuracy of a power grid load prediction method is improved is achieved.

The invention relates to a nonlinear partial least square optimizing model-based forest carbon sink remote sensing evaluation method which comprises the following main steps of: (1) mapping the original variable to a high-dimension space to obtain a new variable by adopting a kernel function and carrying out standardization treatment; (2) carrying out regression analysis on the extracted component by adopting a least square method and reducing a regression coefficient; (3) evaluating a model by adopting LOO cross effectiveness; (4) repeating the step (2) to step (3) and adding 1 to the component number every repetition till the extracted component number reaches the maximal value; (5) repeating the step (1) to the step (4) and adding 1 to the subsection number M in the step (1) every repetition till M is equal to the preset number; and (6) searching the model with maximal related coefficient of an estimating value and a practical value from all the models and modeling with the M and extracted component number at the moment for being used as a final estimation model. The invention uses the optimized nonlinear partial least square regression for establishing a forest carbon storage predicting model and improves the forest carbon storage predicting precision.

The invention provides a marginal hierarchical social relation perception method. A semi-supervised algorithm is decomposed into N+1 parallel subtasks to be allocated to the centralized server and theN F-APs for training and prediction for the centralized server and N fog computing access points (F-AP) of a fog wireless access network based on the integrated semi-supervised algorithm by using multiple classification and regression tree (CART) models for parallel training and prediction so as to reduce the computing burden of the centralized server. The position information required for socialrelation perception is acquired from the F-APs and the conversation record information is acquired from the centralized server. The centralized server and the N F-APs perform parallel computing so that the time overhead can be reduced; and the centralized server and the N F-APs only require to perform information interaction of small data volume with no requirement for uploading all the originaldata to the centralized server so that the burden of the pre-transmission link can be reduced.

The invention discloses a multi-model landslide displacement prediction method and system, and aims to solve the problems of low accuracy of conventional single-model prediction and difficulty in determining a weight of a combination prediction model. The method comprises the following steps of obtaining landslide displacement monitoring data, and performing equal-time-interval processing on the monitoring data to obtain a displacement time sequence of a landslide; selecting multiple displacement prediction models to predict landslide displacement in a time range corresponding to the displacement time sequence to obtain a predicted displacement time sequence; through the predicted displacement time sequence and the displacement time sequence, obtaining a prediction model accuracy matrix, and according to the prediction model accuracy matrix, determining a prediction model weight matrix; and obtaining a final predicted displacement result in combination with the prediction model weightmatrix according to displacement data predicted by the selected displacement prediction models.

The invention discloses a wireless link quality prediction method based on an LSTM neural network. The prediction method comprises the following steps: a wireless communication device collects and stores a wireless link quality signal-to-noise ratio signal sequence as a communication link quality original signal sequence; a mean filtering method is adopted to decompose the communication link quality original signal sequence into a stable sequence and a noise sequence, the noise sequence calculates a noise standard deviation, training and prediction application are performed on the two parts ofdesigned LSTM neural network models respectively, and finally a confidence interval of a required communication link is calculated. The predicted lower bound is compared with the lowest communicationreliability requirement of the intelligent power grid to judge whether the lowest communication reliability standard is met or not. The method can be widely applied to the field of wireless sensor networks, effectively predicts the link quality, and improves the stability and reliability of link transmission.

The invention discloses a battery SOC estimation method based on a model fusion idea. According to the method for estimating the SOC of a battery by using an integrated algorithm for fusing improved support vector regression PSO-SVR, AdaBoost and random forest RF based on a Stacking model fusion thought, firstly, feature expansion and feature screening are carried out on feature engineering of thebattery SOC, in order to reduce an overfitting risk, a data set is processed by using a K-fold cross validation method, then a support vector machine algorithm is improved by using a particle swarm algorithm, and finally, the battery SOC is estimated by using a proposed model fusion method. According to the method, the estimation precision of the SOC of the battery is superior to the estimation precision of three single models of SVR, AdaBoost and RF on the SOC of the battery, the state of charge of the energy storage battery can be accurately estimated, accurate estimation of the SOC of thebattery is a guarantee for efficiently and safely charging and discharging the battery and prolonging the service life of the battery and is the premise of fault diagnosis and is an important guarantee for stable, safe and efficient operation of a power system and is one of necessary ways for accelerating promotion of an intelligent power grid.

The invention relates to the field of intelligent water affairs, and discloses a residential water consumption prediction method based on an MIC-XGBoost algorithm, and the method comprises the steps: obtaining influence factor values and water consumption of each month, constructing a water consumption-influence factor value corresponding table of water consumption influence factors, and employing a maximum information coefficient algorithm to obtain a water consumption-influence factor value corresponding table of the water consumption influence factors; calculating a target influence factor which has the maximum influence on the water consumption, training the original XGBoost algorithm model according to the water consumption and the influence factor numerical value of each month to obtain a target XGBoost algorithm model, and predicting the water consumption by utilizing the target XGBoost algorithm model according to the influence factor numerical value of the current month to obtain the predicted water consumption of the current month. The invention further provides a resident water consumption prediction device based on the MIC-XGBoost algorithm, electronic equipment and a computer readable storage medium. According to the method, the problem that the prediction accuracy is low when the water consumption of residents is predicted by only depending on the XGBoost algorithm can be solved.

The invention discloses a traffic flow prediction method and device, computer equipment and a readable storage medium. The method comprises the following steps: acquiring original traffic data of a target road section within a preset duration; processing the original traffic data to obtain influence parameters of external factors on the traffic flow of the target road section; processing the original traffic data according to the influence parameters to obtain corrected traffic data; dividing the corrected traffic data to obtain training set data and test set data; training an initial causal convolutional recurrent neural network model by using the training set data to obtain an intermediate causal convolutional recurrent neural network model; testing the intermediate causal convolutionalrecurrent neural network model by using the test set data to obtain an evaluation result, and determining the intermediate causal convolutional recurrent neural network model as a target causal convolutional recurrent neural network model when the evaluation result meets a predetermined condition; and predicting the traffic flow of the target road section at the to-be-predicted time point by usingthe target causal convolutional recurrent neural network model.

The invention discloses a data processing method. The method comprises the steps of obtaining a target user identifier and candidate object information corresponding to the target user identifier; obtaining specific time information for accessing the candidate object information by a computing device corresponding to the target user identifier; according to the target user identifier, the candidate object information and the specific time information, determining an attention sequence of the target user to the candidate object information by adopting a pre-trained neural network model, whereinan attention mechanism based on time characteristics is added into the neural network model. By adopting the method, the problem that the prediction accuracy of the attention sequence of the user tothe candidate objects is relatively low is solved.

The invention relates to a wind turbine generator failure prediction method, in particular to a wind turbine generator multivariate failure prediction method based on data driving. The problem that an existing failure prediction method based on data driving is low in accuracy is solved. The wind turbine generator multivariate failure prediction method based on data driving comprises the following steps that 1, state data collecting is conducted on monitored wind turbine generator parts; 2, denoising processing is conducted on the characteristic quantity by adopting a five-point sliding average method; 3, a correlation degree R of the characteristic quantity and remaining life prediction is calculated; 4, a multivariate least square support vector machine prediction model is built; 5, optimization is conducted on a regularization parameter gamma and a nuclear parameter sigma<2> of the multivariate least square support vector machine prediction model; 6, the effectiveness of the multivariate least square support vector machine prediction model is verified; 7, the remaining effective life of the wind turbine generator parts is predicted. The wind turbine generator multivariate failure prediction method based on data driving is suitable for wind turbine generator failure prediction.

According to a network security risk event prediction method and device provided by the embodiment of the invention, the risk event category of the to-be-predicted network data is obtained through the twin neural network classification model, so that the problem that the network security risk event cannot be predicted when the network data sample size is too small or the network data sample distribution is unbalanced, and the accuracy of predicting the network security risk event is not high can be solved. In view of the problem of less marked information or unbalanced network data sample distribution in an actual application scene, the network data sample distribution imbalance degree is reduced by the simplest means, and for a small data set, the number of samples is greatly increased, and the possibility is provided for subsequently performing risk prediction by using a deep learning algorithm with higher fitting capability. When the network data sample size is sufficient and the network data sample distribution is balanced, the twin neural network classification model in the embodiment of the invention can achieve the best performance, and has the best AUC, GM and F1 performance.

The invention discloses a track irregularity prediction method based on a hybrid intelligent optimization LSTM, and the method comprises the steps: firstly carrying out the preprocessing of time series data, then optimizing the hyper-parameters of an LSTM model through a PSO algorithm, and determining a network structure of the LSTM model; and optimizing the initial weight threshold of the LSTM model by using a GA algorithm, and determining the weight threshold of the LSTM model. And finally training and predicting track irregularity data by using the determined hyper-parameter and weight threshold. According to the method for predicting the track irregularity data based on the LSTM-PSO-GA model, the problem that the precision is not high in the prediction process of a traditional prediction method is solved, the LSTM parameters are optimized through the PSO and GA algorithms, the problem that the model falls into a local optimal solution is avoided, and the prediction convergence speed is increased. Finally, the track irregularity data is predicted, and the track irregularity phenomenon is predicted more accurately.

The invention provides a terminal equipment replacement prediction method and device, a storage medium and electronic equipment. The method comprises the steps that target operation data of a target object is obtained, and the target operation data is operation data generated by multiple selection operations executed by the target object on resource information on a target page within target time; according to the target operation data, a target time sequence and target object features are obtained, the target time sequence is used for recording multiple selection operations according to the sequence of execution time, and the target object features are used for describing a target object; and according to the target device data, the target time sequence and the target object features, machine change prediction is performed to obtain a first prediction result, the target device data being device data of a target terminal device held by the target object, and the first prediction result being used for indicating whether the target object changes the target terminal device or not. Whether the target terminal equipment is replaced by the target object can be objectively reflected, and the accuracy of machine replacement prediction is improved.

The invention discloses a large-range landslide deformation prediction method based on InSAR inversion and multiple impact factors, and the method comprises the steps: carrying out the InSAR inversion processing of SAR image data, and obtaining time sequence deformation data; clustering the time sequence deformation data to obtain a plurality of categories of time sequence deformation data; decomposing the time sequence deformation data of each category into a periodic term deformation sequence and a trend term deformation sequence; determining influence factors significantly related to periodic term deformation; respectively establishing an LSTM (Long Short Term Memory) model of each type of time sequence deformation data to predict each type of deformation; and adding the periodic term deformation prediction value and the trend term deformation prediction value of each category to obtain a deformation quantity prediction result of each category, and merging the deformation quantity prediction results of each category to obtain a large-range landslide deformation prediction result. According to the method, effective prediction of large-range landslide deformation can be realized, and the defects of small prediction range, high cost and the like in the prior art are overcome.

The invention discloses a trust prediction method based on an exponential smoothing method and a grey model. The method comprises the steps of obtaining a historical trust value sequence of a target vehicle in a vehicle-mounted ad hoc network; carrying out smoothing processing on a historical trust value sequence according to an optimal value of a smoothing coefficient in a predetermined exponential smoothing method to obtain a smooth trust value sequence, and then inputting the smooth trust value sequence into a grey model constructed based on the optimal value of the smoothing coefficient toobtain a trust prediction result. Visibly, according to the method, the historical trust value sequence is processed by utilizing an exponential smoothing method, the interference of sequence randomfluctuation on a prediction result is reduced, and in addition, parameter optimization is performed on a smoothing coefficient of the exponential smoothing method and an objective function of a grey model, so that the accuracy of the prediction result is further improved. The invention also provides a trust prediction device and equipment based on the exponential smoothing method and the grey model, and a computer readable storage medium, and the effects of the trust prediction device and equipment are corresponding to those of the method.

The invention discloses a hybrid modeling method for predicting key operating parameters of a drying part of a toilet paper machine. The method comprises the following steps: S1, acquiring historicaloperating parameters and abnormal cleaning data of the drying part of the toilet paper machine; S2, establishing a paper sheet drying mechanism model, predicting key operation parameters of the historical operation parameters in the S1 by using the mechanism model, and calculating mechanism prediction errors of the key operation parameters; S3, preprocessing the mechanism simulation error data inS1 and S2 by using a normalization method, and dividing the data into a training set and a test set according to a certain proportion; S4, establishing a mechanism prediction error compensation modelof the key operation parameters according to a BPNN algorithm principle, and using the training set data to train adjustment and optimization of the model and algorithm hyper-parameters; S5, integrating the paper sheet drying mechanism model and the mechanism prediction error compensation model, and predicting key operation parameters of the test set data. The invention solves the problem that a mechanism modeling method is low in paper sheet drying key operation parameter prediction precision.

The invention discloses a method for predicting the critical frequency of the ionospheric F2 layer based on the ELM: obtaining the measured data of the influence factor of the foF2 and the value of the foF2 hour, and dividing the data into three groups as the training data, the test data and the verification data of the ELM model respectively; determining the input and output variables of the ELMmodel; the training data and test data are imported into the ELM model, and the ELM model is trained. The prediction error RMSE of the ELM model was compared with the expected value of the precision of the ELM model until the prediction error RMSE was less than the expected value and the optimal ELM model was determined after the training. The validation data is imported into the trained optimal ELM model, and the accurate output prediction value is obtained. The invention can realize accurate and fast prediction of foF2.

The invention provides a method and a system for forecasting the loss of telecommunication customers, and an electronic device. The method comprises the following steps: calculating the sample dividing proportion corresponding to the on-line customers and the off-line customers respectively in a telecommunication network, and determining the improved Gini base number of a random forest algorithm based on the sample dividing proportion corresponding to the on-line customers and the off-line customers respectively; based on the improved Gini cardinality, using a stochastic forest algorithm to predict the loss of telecommunication customers. The sample division ratio represents the proportion of the sample size of the selected corresponding category customers in the network state to the totalnumber of the category customers in the network state. The invention can effectively solve the problem of low accuracy rate of customer churn prediction caused by imbalance between categories under unbalanced data, especially low accuracy rate of high-value user prediction.

The embodiment of the invention provides a water-drive reservoir yield prediction method and device and a storage medium, belongs to the technical field of oil exploitation, and solves the problem of low prediction accuracy caused by single consideration factor of production well yield prediction in the prior art. The method comprises the following steps: acquiring space information and time information of a target oil reservoir well pattern, and establishing a space-time diagram structure data set of the target oil reservoir well pattern; performing sliding time window division on the space-time diagram structure data set, and establishing a training sample set of the target oil reservoir well pattern; training by utilizing the training sample set to obtain a multi-layer space-time diagram neural network for predicting the yield of the water-drive reservoir; and obtaining a yield prediction result of a production well of the target oil reservoir well pattern by using the multilayer space-time diagram neural network. The embodiment of the invention is suitable for yield prediction of the production well in the water-drive reservoir well pattern.

The invention discloses a solid state disk life prediction method and device and a computer readable storage medium. The method comprises that the daily user system write-in amount of a to-be-tested solid state disk in a preset historical time period is acquired, and a differential historical time sequence used for representing the daily change condition of the user system write-in amount of the to-be-tested solid state disk in the preset historical time period is generated; the differential historical time sequence is input into a pre-trained exponential smoothing model to obtain a daily variation prediction value of the user system write-in quantity of the to-be-tested solid state disk in a preset future time period; the service life of the to-be-tested solid state disk is predicted according to the actual user system write-in quantity of the day before the prediction day, the daily variation prediction value and the total write-in data of the to-be-tested solid state disk; and the weights given to the write-in change values at different times are different, so that the predicted value can accurately reflect the change condition of recent data write-in, and the prediction accuracy of the SSD life is effectively improved.

The invention relates to a nonlinear partial least square optimizing model-based forest carbon sink remote sensing evaluation method which comprises the following main steps of: (1) mapping the original variable to a high-dimension space to obtain a new variable by adopting a kernel function and carrying out standardization treatment; (2) carrying out regression analysis on the extracted component by adopting a least square method and reducing a regression coefficient; (3) evaluating a model by adopting LOO cross effectiveness; (4) repeating the step (2) to step (3) and adding 1 to the component number every repetition till the extracted component number reaches the maximal value; (5) repeating the step (1) to the step (4) and adding 1 to the subsection number M in the step (1) every repetition till M is equal to the preset number; and (6) searching the model with maximal related coefficient of an estimating value and a practical value from all the models and modeling with the M and extracted component number at the moment for being used as a final estimation model. The invention uses the optimized nonlinear partial least square regression for establishing a forest carbon storage predicting model and improves the forest carbon storage predicting precision.

The invention provides a thin layer prediction method and a device, wherein the method includes that log data and seismic data of a target layer are obtained; wave impedance type of the target layer is obtained according to the log data; a corresponding wave impedance model is established according to the wave impedance type; a corresponding tuning model is ensured according to the wave impedancemodel; the thin layer prediction is conducted according to the tuning model and the seismic data. The corresponding wave impedance type of the target layer is distinguished through the seismic data, and then the corresponding wave impedance model and the tuning model is ensured according to the different wave impedance types, which can make the thin layer prediction more accurately. So the existing technical problems of low accuracy and large error in the thin layer prediction are solved. And the technical effect of tuning effect analysis and thin layer prediction can be achieved for differenttypes of the wave impedance thin layer.

A protein and nucleic acid binding site prediction method based on graph neural network characterization comprises the following steps: constructing a protein and nucleic acid interaction data set, extracting the position and feature information of each residue in the protein and the structural context thereof after sample fusion processing, and constructing graph representation of the structural context of the residues according to the graph representation. And predicting the graph representation of the to-be-predicted protein through the hierarchical graph neural network to obtain the probability that each residue is combined with DNA / RNA, thereby realizing prediction of the binding site of the protein and the nucleic acid. Key structures and feature patterns of binding sites are learned from graph representations through graph representations based on residues of structural context and a hierarchical graph neural network model.

The invention relates to a vehicle travel time prediction method and device and a computer device. The vehicle travel time prediction method e comprises the following steps: acquiring time informationand position information of target vehicle driving, determining time slice information of a target vehicle according to the time information, determining spatial slice information of the target vehicle according to the position information, when it is determined that the target vehicle is located in a target space slice according to the position information, according to the position information,the time slice information and the historical time information, calculating a time feature vector of the target vehicle; according to the spatial slice information and the reference speed in the spatial slice, calculating a spatial feature vector of the target vehicle; according to the time information, the position information, the time feature vector and the space feature vector, predicting thetravel time of the target vehicle. The problem that the prediction accuracy is low under the condition of carrying out long-term prediction on a vehicle in a public transport system based on a spacefeature vector is solved, and the prediction accuracy of the travel time of the vehicle is improved.

The invention discloses an effluent total phosphorus prediction method based on a fuzzy neural network, electronic equipment and a medium. The method comprises the following steps: determining a characteristic variable of effluent total phosphorus as an input variable; constructing an initial prediction model based on a fuzzy neural network; obtaining a training sample, inputting the training sample into the initial prediction model, and determining a final prediction model through a multi-target particle swarm optimization algorithm; and inputting the input variable into the final prediction model, and calculating the total phosphorus in the effluent. According to the method, a prediction model based on the fuzzy neural network is established to predict the effluent total phosphorus, an asymmetric membership function is adopted to describe the distribution characteristics of variable data, a multi-target particle swarm optimization algorithm is utilized to dynamically adjust the structure and parameters of the fuzzy neural network at the same time, and real-time prediction of the effluent total phosphorus concentration of sewage treatment is achieved.

The invention discloses a method for predicting the resource performance of a cloud server based on an LSTM-ACO model. The method comprises the steps of firstly carrying out the preprocessing of time sequence data, and mapping original sequence data to a [0, 1] interval; then determining an LSTM model, training and predicting existing data, and optimizing the LSTM model by using an ant colony algorithm; and finally, inputting the prediction result of the LSTM model for the data of the moment t and the data of the moments t-1, t-2,..., t-n into the LSTM-ACO model, and predicting the data of the moment t. According to the method for predicting the resource performance of the cloud server based on the LSTM-ACO model, the problem that a traditional prediction method is not high in precision in the prediction process is solved, the LSTM parameters are optimized through ACO, the problem that the model is caught in a local optimal solution is avoided, and the prediction convergence speed is increased; and finally, cloud server resource and performance prediction is realized, and the software aging phenomenon is predicted more accurately.