204 results about "Optimality model" patented technology

A model creation tool is provided that automatically evaluates potential econometric models and, given a set of historical data on which the model is intended to operate and other constraints, automatically identifies an optimum model. The optimum model may be an autoregressive model. The optimum model is defined by a set of optimum model parameters and optimum model coefficients. The optimum model parameters may include a model order such as an autoregressive model order, a seasonality function parameter, a trend function parameter, and a parameter that sets a tolerance that is used when performing principal components dimension reduction operations on the historical data.

The invention discloses a surface roughness prediction method based on GA-GBRT and a method for optimizing process parameters. The method comprises the steps of: collecting data to construct a data set, and dividing the data set to training set data and test set data, and employing the training set data to perform training of key parameters of a GBRT model; b, performing parameter coding and population initialization: randomly generating a chromosomal sequence for increasing the number of iterations, the maximum depth of the individual regression estimator and the learning rate; c, employing the k-folded cross-validation method to train the GBRT model, and employing the genetic algorithm to calculate the fit goodness fitting value of each individual; d, when the number of cycles does not reach the maximum number of iterations, allowing the population to be selected, crossed and mutated to produce a new generation of populations, and continuously performing training of the GBRT model; and e, repeatedly performing the steps c and d until the number of cycles reaches the maximum evolution algebra or exceeds the maximum number of iterations to obtain the optimal model parameters. The surface roughness prediction method based on GA-GBRT and the method for optimizing process parameters are high in test precision and superior in prediction performance and improves the surface processing precision of the workpiece.

The invention discloses a WEB public opinion trend forecasting method based on optimal model. The basic thought of the method is to firstly classify historical public opinion events to obtain several categories of opinion events, secondly clustering the time series plot of events in each obtained category to obtain sub-categories and finally obtaining the optimal model of each sub-category while ensuring the sum of mean-square errors is least to obtain the optimal model of each main category. When given an event to be predicted, the predicted event is classified and the optimal models obtained by early training to the category of the event are selected for matching, thus the model which is more suitable for the development trend of the event and the change proportion in matching can be selected; and inverse transformation is performed to the selected models according to the obtained change proportion so as to obtain the long-term development trend of the predicted event. Therefore, the defect that the existing network forecasting method can not forecast the inflection point can be overcome, the government and the supervision department can adopt timely and effective measures and the effects of network supervision can be better realized.

The invention provides an automated method for identifying an optimal or near optimal mathematical model to describe observed data including: a) the definition of a candidate model search space, b) methods for searching said candidate model search space to identify the optimal or near optimal model within said candidate model search space. The present invention includes algorithms for writing the computer code needed to implement and evaluate candidate models in the software package NONMEM.

The invention discloses a multi-objective-based building scheme stage energy-saving optimization design mode construction method, and belongs to the field of building design. The method sequentially comprises the steps of integrating a building body and environment information, determining an optimization target and a design variable, establishing a building multi-target prediction model, coupling the prediction model and executing operation, and constructing an energy-saving optimization design mode. According to the method, the building energy consumption, the thermal comfort index and the initial investment cost are considered at the same time in the scheme stage, the support vector machine algorithm in machine learning is adopted to construct the multi-target prediction model, compared with software simulation, target prediction can be converted into a numerical calculation process from a performance simulation process, the prediction efficiency of the performance index is improved, And the synchronism of multi-target prediction can be ensured. The algorithm is specially proposed for the small sample problem, so that a large number of redundant samples can be eliminated, the optimal model is obtained under the condition of limited samples, and the construction efficiency of the prediction model is improved.

The invention discloses a method for identifying user interest. The method includes acquiring training samples and test samples; extracting characteristics of the training samples and the test samples by the aid of first preset algorithms and computing the optimal model parameters of logic regression models according to the characteristics of the training samples by the aid of iterative algorithms; evaluating the logic regression models with the optimal model parameters according to the characteristics of the test samples and the areas under ROC (receiver operating characteristic) curves (AUC) and acquiring first theme classifiers by means of training; determining themes of text data by the aid of the first theme classifiers, computing scores of the themes of the text data according to the logic regression models with the optimal model parameters and computing confidence scores of the themes according to second preset algorithms. Training is carried out according to the text data to obtain corresponding topic models, and the topic models are manually annotated to obtain the training samples. Users are interested in the themes with the computed confidence scores. The invention further discloses a device for identifying the user interest and a computer readable storage medium. The method, the device and the computer readable storage medium have the advantages that the user interest can be identified, and accordingly enterprises can be assisted in accurately positioning potential clients.

The invention discloses a modeling method of a support vector machine (SVM)-based software measurement instrument in a biological fermentation process. The modeling method uses an optimized SVM modelto estimate fermentation products in the biological fermentation process and comprises two steps: the SVM modeling of the biological fermentation process and the confirmation of an optimal model basedon a genetic simulated annealing algorithm and an akaike information criterion. The invention provides an optimal SVM software measurement instrument model for variables difficult in online measurement in the biological fermentation process, eliminates a phenomenon of untimely control caused by time lag of a traditional offline measuring method, lays a reliable foundation for establishing the SVMmodel in the biological fermentation software measurement and has the advantages of online measurement, automatic parameter optimization, automatic input variable selection, small sample demand and high precision.

The invention discloses a method and a device for identifying tumor purity and absolute copy number based on sequencing data. The method comprises the following steps: comparing offline data after quality control to a reference genome, and carrying out variation detection and population database annotation; testing the preprocessed data of the tumor and the normal sample by using purity predictionsoftware to obtain a purity and copy number information model; and for the model conforming to normal distribution, further screening out the model with the maximum probe support number in the high tumor cell fraction subcloning region, and defining an optimal model in combination with the matching rate of BAF and alle1 and alle2 copy numbers. According to the method, the model of purity detection software is rapidly and efficiently corrected, and the purity and absolute copy number information of tumors can be obtained more accurately; the accuracy is guaranteed, the tedious process of manual verification is avoided, the labor cost is saved, and a foundation is laid for follow-up tumor genome evolution and tumor heterogeneity research.

The invention discloses a geological disaster risk comprehensive evaluation method considering spatial distribution characteristics. Comprising the following steps: aiming at spatial aggregation and dispersion characteristics of historical geological disaster points, respectively proposing a data preprocessing method which uses a clustering algorithm to extract regional clustering attributes as evaluation indexes and is based on a fishing net grid; constructing a model based on a multi-machine learning algorithm of logistic regression (LR), a support vector machine (SVM), a gradient boosting tree (GBDT) and a random forest (RF); determining an optimal model by comparing the model prediction precision with the map evaluation effect, and drawing a dangerous map; meanwhile, providing an experimental scheme for testing the technical reliability. According to the method, the model evaluation precision can be remarkably improved, the model evaluation performance can be enhanced, and the geological disaster risk map with more accurate prediction and better quality can be generated, so that a decision basis conforming to the actual situation is provided for disaster risk prevention and control planning work.

The invention discloses a causal relationship mining method based on deep learning. The method comprises the following steps: firstly, preprocessing data by using technologies such as missing data supplement, data normalization and independent hot coding; then, based on a Keras deep learning framework, carrying out univariate time sequence prediction on the target characteristics by using an LSTM;adjusting the model structure and a series of hyper-parameters to obtain an optimal model; recording the R2 score of the model on the test set; and then, predicting all the alternative features by using the model to obtain R2 scores of the alternative features on the test set, subtracting the two scores to obtain a Granger causal relationship score of the alternative features and the target feature, and obtaining a quantitative number for describing the Granger causal relationship between the alternative features and the target feature. The method is suitable for the problem of influence factor analysis of other time sequences. In conclusion, the Granger causal relationship mining method based on deep learning has the advantages of mining much data in various fields.

The invention relates to a flood probability forecasting method based on multi-source uncertainty. The method comprises the following steps: estimating surface rainfall probability distribution basedon an incomplete rainfall monitoring group; constructing a suboptimal solution set of each model parameter, and obtaining probability distribution of the model parameters; randomly generating a combination of surface rainfall and parameters, and driving the hydrological model to generate a plurality of groups of initial forecast flow processes; estimating the probability that each model is relatively optimal by using a BMA method; and randomly extracting the optimal model and the corresponding initial forecast flow, and estimating the posterior probability distribution of the forecast flow torealize probability forecast. According to the invention, uncertainty in three aspects of rainfall input, parameters and a model structure is comprehensively considered. A calculation method for realizing flood probability prediction can be widely applied to the situations that rainfall input errors are obvious. A model or a model structure suitable for flood prediction cannot be accurately predicted, and model parameter uncertainty is high, the accuracy and reliability of a flood prediction result can be improved, and technical support is provided for flood control and disaster reduction work.

The invention provides a model optimization method and device based on scene adaptation, and the method comprises the steps: determining N algorithm models corresponding to a set scene based on the specific scene information provided by a user; Wherein N is a positive integer; Dividing the data set provided by the user to obtain a training set and a test set; Training each algorithm model throughthe training set to obtain one or more sub-models corresponding to each algorithm model; And testing one or more sub-models corresponding to the algorithm models through the test set, and determiningan optimal model corresponding to the specific scene information. The technical problems that in the prior art, an appropriate algorithm is difficult to select in an actual scene, and assistance is provided for industrial development are solved. The technical effects that all indexes and the generalization capability of the model are guaranteed, the capability of finally confirming the optimal model is achieved, convenience and rapidness are achieved, the accuracy is high, and the optimal model with the high generalization accuracy, performance and adaptability level is obtained are achieved.

The invention discloses a recurrent neural network architecture search method and system based on an improved evolutionary algorithm, and a medium. The method comprises the following steps: training aplurality of recurrent neural network sub-models to update a shared weight; initializing a generation population and a historical record table for recording the performance of all the recurrent neural network models; randomly sampling from the population to generate a sample, selecting a sample optimal model to perform mutation operation, removing the oldest or worst model in the population witha specified probability, and adding mutated child nodes into the population and the historical record table; and judging whether a preset ending condition is met or not, if not, continuing to carry out sample variation, and otherwise, outputting an optimal model in the historical record table. According to the method, the search process of a recurrent neural network architecture can be accelerated, the performance and the search time are considered simultaneously when the population is updated in each step, and the search efficiency of the recurrent neural network architecture can be greatly improved.

The invention is applicable to the technical field of transformers, and provides a method and terminal device for predicting the winding hot-spot temperature of a transformer. The method comprises thesteps of collecting temperature data of a winding hot-spot of the transformer, and enabling the collected temperature data to serve as sample data; training an artificial neural network model according to a particle swarm algorithm and the sample data to obtain optimal model parameters of the artificial neural network model, and enabling the artificial neural model based on the optimal model parameters to serve as a prediction model; collecting operating parameters and structural parameters of the transformer; and performing winding hot-spot temperature prediction on the transformer accordingto the operating parameters, the structural parameters and the prediction model. According to the invention, the particle swarm algorithm is introduced into parameter optimization for the artificialneural network model, and prediction is performed on the winding hot-spot temperature of the transformer by using the optimized prediction model, so that the accuracy of prediction obtained by the prediction model for the winding hot-spot temperature of the transformer can be improved.

The invention provides a steam turbine health state prediction method based on E-LSTM. The method comprises steps of collecting turbine operation data from a sensor, and preprocessing the turbine operation data; feeding the preprocessed data into an LSTM network, and performing iterative training for multiple times; inputting a plurality of trained model parameters into a genetic algorithm to serve as an initial population, operating the genetic algorithm, and selecting a model parameter with an optimal effect; performing generalization performance verification on the optimal model by using more steam turbine operation data; and predicting the test data set according to the optimal model parameters, and evaluating model errors. According to the method, the accuracy of model prediction canbe improved, over-fitting can be avoided, and multivariate linear regression prediction can be realized, so that the prediction model has a better fitting effect on real data, the error of manual monitoring can be greatly reduced, the fault diagnosis efficiency can be improved, and the occurrence of faults can be informed in advance. The method can be widely applied to state management of variousfirepower and nuclear power plants and even steam turbines of ships.

The invention discloses an energy-based multi-objective optimization fitting prediction method for atomic structures and electron density maps. The method comprises the following steps: building a reference data set of a predicted structure and an electron density map according to three-dimensional structures and electron density maps of proteins, and generating an initial model; preliminarily moving the predicted atomic structure to the center of the density map by using the information of the electron density map to generate N initial models; selecting a Pareto set through a multi-objectiveparticle swarm optimization algorithm, selecting an optimal model from the Pareto set through a Knee algorithm, and acquiring a fitting result between the atomic structure and the electron density mapthrough calculation. According to the invention, the problem of potential deviation caused by minimizing a single energy function can be solved.

Systems and methods for generating at least one of an automated machine learning (ML) model, artificial intelligence (AI) model or quantum ML model for a user via a model generation framework are provided. The method includes receiving a user input including at least one of a data, one or more tasks and a metadata, from the user, the metadata including least one of: a selection of domain, a selection of sub-domain, or one or more keyword tags. One or more building blocks are determined in the selection of domain or said selection of sub-domain by performing a meta-learning, a transfer learning or a neural architecture search. An optimal model is iteratively determined based on the building blocks and a performance estimation of the building blocks, the optimal model including at least one of AI model, ML model or quantum ML model. The optimal model is rendered to the user.

The invention discloses a multi-gaussian kernel self-optimization relevance vector machine based wastewater quality soft-measurement method. The method comprises the following steps: (1) removing abnormal data from wastewater input and output data, performing normalization processing due to different dimensions of input variables and normalizing to be within an interval from 0 to 1; (2) forming a multi-gaussian kernel function relevance vector machine soft-measurement model module; (3) performing multi-gaussian kernel function nuclear parameter self-optimization algorithm; (4) forming an initial parameter optimization module by the genetic optimization algorithm; and (5) realizing modeling of a multi-gaussian kernel function relevance vector machine soft-measurement model. According to the method, nuclear parameters of all dimensions can be determined by a self-optimization method, the initial parameters are optimized by the genetic optimization algorithm, an optimal model is built, the output precision of BOD (Biochemical Oxygen Demand) in wastewater is effectively improved under conditions that the convergence and the sparseness of the model are ensured.

The invention belongs to the technical field of bearing fault diagnosis, and discloses a bearing fault diagnosis method, system, device and terminal, and the method comprises the steps: extracting the time-frequency features of an original vibration signal of a bearing through continuous wavelet transform, and converting the time-frequency features into a 32 * 32 pixel two-dimensional image; using an improved AlexNet model to carry out fault feature extraction on the time-frequency spectrogram; for fault diagnosis classification, selecting optimal model parameters through an LGBM classification algorithm and Bayesian optimization. The bearing fault diagnosis method provided by the invention has optimal fault diagnosis accuracy. Through experimental comparison, compared with other seven methods, the method provided by the invention has the highest accuracy rate of 99.712%, the prediction time consumption of 1.47 seconds for 1800 samples is also at the same order of magnitude as the time consumption of other models, the variance of the accuracy rates of five times of prediction is only 0.063, and the method provided by the invention is more stable than other six methods and has optimal comprehensive performance.

The invention discloses an optimized modeling method of a soft measuring instrument for a magnetic chain based on a least squares support vector machine during the stable suspended movement process of a bearingless asynchronous motor. The modeling method comprises the following two parts: modeling via the least squares support vector machine and the determination of an optimal model based on the improved particle swarm optimization. A soft measuring instrument model based on the least squares support vector machine is provided for the magnetic chain variable of the bearingless asynchronous motor difficult to measure in an online way, and the defects of low control accuracy and the like caused by time lag in the traditional offline measuring method can be well overcome. The modeling method has various advantages of automatic parameter optimization, high prediction accuracy, small sample demand size, strong anti-interference capability and the like.

The invention discloses a tea disease identification method and system based on lightweight optimization Yolo v4, and the method comprises the steps: collecting a tea disease picture, and carrying out the preprocessing of the tea disease picture, and taking the preprocessed tea disease picture as a data set for training a Yolo v4 model; performing lightweight optimization on a feature extraction trunk module and a feature extraction fusion module in the Yolo v4 model to obtain an optimized Yolo v4 model; the optimized Yolo v4 model is trained and verified through the data set for training the Yolo v4 model, and the optimal Yolo v4 model for recognizing the tea diseases is obtained; and identifying the tea disease image by using the obtained optimal Yolo v4 model. According to the invention, the huge parameter quantity and the model volume of the original Yolo v4 network model are effectively reduced, and the detection efficiency and the identification precision of the tea disease target are improved.

The invention discloses a TBM tunneling control parameter intelligent prediction and optimization decision-making method, and the method comprises the steps: carrying out the preprocessing of TBM tunneling parameters and the vibration acceleration of a cutter, and carrying out the training of an LSTM model through cross validation, and obtaining an optimal model hyper-parameter of the LSTM model;training the LSTM model of the optimal model hyper-parameter; training the secondary learner through cross validation and an improved loss function to obtain an optimal model hyper-parameter of the secondary learner; training the secondary learner to obtain a final stacking integration model; and based on a stacking model prediction result, generating an optimal tunneling control parameter by adopting a multi-target particle swarm algorithm. The invention provides a real-time prediction and optimization decision-making method for tunneling control parameters in the excavation process of a full-face tunnel boring machine, solves the problems of automatic selection and adjustment of TBM tunneling process parameters in complex stratums, and has important significance for safe, efficient and intelligent tunneling of the full-face tunnel boring machine.

The invention discloses a cascade hydropower station generating capacity prediction method based on a long-term and short-term memory network. The method comprises the steps of performing stability test on a hydropower station generating capacity time sequence; performing correlation test on the generating capacity time sequence; converting the generating capacity time sequence data into supervised learning data; establishing a generating capacity prediction model based on the long-term and short-term memory network; performing integrated empirical mode decomposition on the generating capacitydata to obtain a training set and a test set; training the generating capacity prediction model by using the training set, and performing model hyper-parameter optimization by using an improved discrete differential evolution algorithm to obtain optimal model parameters; and predicting the generating capacity of the cascade hydropower station by adopting the generating capacity prediction model.The method is suitable for predicting the generating capacity of large and medium-sized cascade power stations; the power generation quantity prediction model based on the LSTM neural network has moreadvantages for predicting the power generation quantity of the power station adjusted for many years, and the fitting precision of the model is improved through hyper-parameter optimization of the model.

The invention relates to a scientific and technical literature quoting number predicting method based on a time sequence. The predicting method includes the steps that statistics is carried out on scientific and technical literature quoting numbers, and then average literature quoting numbers of all the months are calculated; in combination with the average literature quoting numbers of all the months, the quoting numbers of the corresponding months are processed in a normalization mode to obtain the quoting time sequence; cluster analysis is carried out according to the time sequence, and a quoting number model with the optimal predicting performance is obtained through dividing training sets and verifying sets, building a regression model and performing error analysis; according to similarity analysis of literature to be predicted and time sequences of various kinds of literature, the class with the highest similarity is obtained, and the quoting number, in the next month, of the literature to be predicted is obtained through the model with the optimal predicting performance. The quoting conditions of each piece of published literature can be automatically analyzed, the average literature quoting numbers of all the months are obtained, different quoting modes of the literature are excavated through clustering, and then the future quoting number is predicted according to the existing time sequence of the literature to be predicted.