200 results about "Gradient boosting" patented technology

The present invention provides an energy load forecasting system and method based on weather big data. The system includes three modules: data acquisition, model learning and load forecasting. The main steps of the method include: firstly, according to the area where the energy load is located, the weather observation data, temperature The temperature value measured by the sensor and the temperature value of the local user's smart phone, where the weather observation data includes characteristic data such as temperature, humidity, wind, rainfall and light intensity; secondly, normalize the acquired weather data to form a weather Big data training set; then, use the XGBoost gradient boosting algorithm to extract the influence weight value of the weather data on the energy load data, and then use the LSTM neural network model to construct the energy load prediction model; finally, combined with the weather forecast data in the region, use the energy The load forecasting model predicts the energy load in the area to be forecasted. The invention effectively improves the traditional energy load single time series analysis method and improves the energy load prediction accuracy.

Computer-enabled methods, apparatus, and computer-readable media are provided for verifying that a given network name, such as a URL, is an official, e.g., registered, approved, or otherwise officially recognized, network name that refers to or identifies a principal, such as a business. These techniques involve receiving a principal name and a given network name, receiving at least one feature attribute from at least one database of feature attributes, wherein the at least one feature attribute comprises a characteristic of the principal name or a characteristic of the network name, and invoking a logistic regression method to generate a probability, based upon the at least one feature attribute, that the given network name is an official network name for the principal name. The logistic regression method may include a gradient boosting tree model that generates the probability based upon the at least one feature attribute.

The invention discloses a mosquito-borne infectious disease epidemic situation prediction method and system based on a gradient boosting tree. The mosquito-borne infectious disease epidemic situationprediction method based on a gradient boosting tree includes the steps: widely collecting various factor data influencing the mosquito-borne infectious disease; cleaning the data influencing the mosquito-borne infectious disease so as to perform importance ordering on the factors influencing the mosquito-borne infectious disease, on the basis of the gradient boosting tree; according to the selected important factors influencing the mosquito-borne infectious disease, establishing a mosquito-borne infectious disease epidemic situation prediction model based on Poisson regression; by means of theselected factor and the correlation coefficients of the mosquito-borne infectious disease, initializing the prediction model, and then determining the mosquito-borne infectious disease prediction model parameters by means of S fold cross-validation; and by means of a epidemic situation hot spot graph based on geographical information and an epidemic situation outbreak graph based on a time axis,visually displaying the model prediction result. The mosquito-borne infectious disease epidemic situation prediction method and system based on a gradient boosting tree apply the gradient boosting tree and other machine learning methods to the field of mosquito-borne infectious disease epidemic situation prediction, can improve the mosquito-borne infectious disease epidemic situation prediction accuracy, can assist disease control staff to predict the mosquito-borne infectious disease epidemic situation in advance, and can timely take the corresponding measures to control large scale outbreakof the infectious disease.

The invention discloses a public bicycle flow variation volume forecasting method based on heap model fusion. The public bicycle flow variation volume forecasting method comprises the steps of: 1, adopting a method of fusing public bicycle rental record data and meteorological data for extracting features, and constructing eigenvectors from several perspectives of time, space, meteorology, history, clustering and the like; 2, adopting a distance similarity matrix combining geological positions and a rental relation, clustering by utilizing a clustering algorithm, and configuring clustering features into the eigenvectors; 3, dividing the eigenvectors into five groups according to feature types, training five basic models by utilizing a machine learning system based on a gradient boosting tree algorithm, training features by adopting a cross validation method, and training a heap model by taking results of the five groups of basic models as features. The public bicycle flow variation volume forecasting method based on heap model fusion ensures that a certain difference exists among the basic models, constructs the heap model by adopting the cross validation method finally, improves the accuracy degree of the model, has good forecasting precision, and has small errors.

The invention discloses an HTTP request double-layer progressive anomaly detection method, anomaly detection is carried out based on a double-layer progressive anomaly detection model, and the model is composed of an XGBoost gradient boosting tree module and an LSTM long-term and short-term memory network module. The method comprises the following steps: training an HTTP request data set of a certain scale; according to the model, firstly using a first-layer XGBoost gradient boosting tree module for rapid judgment, inputting a suspected HTTP request into a second-layer LSTM module for carefuljudgment, achieving HTTP request exception detection, and achieving the purpose of intercepting a malicious HTTP request through a normal HTTP request. According to the method, the accuracy of HTTP request anomaly detection can be effectively improved, the false positive of anomaly detection is reduced, and an anomaly detection model has certain long-term dependence.

Disclosed is a system for forecasting demand for goods and / or services. In at least certain embodiments the system is configurable to select a machine learning model from among multiple different machine learning models for forecasting demand for a dataset that may be continually being updated over time. The models available to the system are each based on different machine learning algorithms (e.g., linear regression, gradient boosting, neural network, etc.) as well as several variations for each algorithm available to the system. The system can monitor changes in the datasets, changes in accuracy of the machine learning results, and external factors, and based thereon, determine whether to initiate a model reselection process or a model retraining process. Each machine learning model can be evaluated against each dataset and can select the best model for the dataset.

The embodiment of the invention provides a method and a system for forecasting the spatial distribution of heavy metals in soil, and relates to the technical field of environmental monitoring. The method comprises the following steps: measuring a content value of heavy metals in soil of a soil sample at a sampling point; constructing a training data set; constructing a GBRT (Gradient Boosting Regression Tree) model by adopting an auxiliary feature and the content of heavy metals in the soil as variables, and training the GBRT model by using the training data set to obtain a trained GBRT model; constructing a data set to be measured; inputting the data set to be measured into the trained GBRT model and outputting a content value of heavy metals in the soil corresponding to a point to be measured. As a GBRT algorithm iteratively integrates characteristics of a number of weak regression devices, and auxiliary features are rationally applied, the method has a good generalization ability and is conducive to popularization and application.

The invention discloses a method for extracting texture based on a lung nodules three orthogonality position computed tomography (CT) image, which includes: a building a lung nodules three orthogonality position CT image system; and b extracting lung nodules three orthogonality position texture characteristic parameters through a disperse Curvelet transformation method. The invention further discloses a method for forecasting lung cancers, which includes: a extracting lung CT image texture characteristics; b extracting three orthogonality position CT image and video signs; and c building a forecasting model based on multi-dimensional characteristic parameters including lung nodules three orthogonality position CT image texture characteristics and video signs through a Gradient Boosting algorithm. The method for extracting texture based on lung nodules three orthogonality position CT image and the method for forecasting lung cancer can be used for early diagnosis of pulmonary nodules.

The invention relates to the field of intelligent power grid load identification technology application, in particular to a computer readable storage medium. A computer program is stored in the computer readable storage medium, and a load identification method based on power fingerprint is realized when the computer program is executed by a controller. According to the load identification method based on the electric power fingerprint, household appliance load identification features are extracted from massive household appliance load power utilization data through power calculation, Fourier analysis and other means, so that an electric power fingerprint set is obtained, training is performed through a gradient boosting tree classifier, and a gradient boosting tree classifier model is verified through a test data set; and finally, the trained gradient boosting tree classification model is used for household appliance load classification. And data mining is performed on the household appliance load electric power fingerprint by using the gradient boosting tree classifier so that rapid and accurate identification of the household appliance load of the intelligent power grid can be realized.

The invention, which belongs to the technical field of train seat adjustment information processing, discloses a train seat comfort evaluation method and system based on multi-source physiological signals. After combination of physiological signal data of a subject and a subjective evaluation scale, a train seat comfort database for a group is established by combining a mainstream subjective evaluation method with a physiological-signal-based objective evaluation method; and a train seat comfort evaluation model is established by selecting a LightGBM gradient boosting tree which is a machine learning modeling method with multiple characteristics of rapidness, distribution and high performance. According to the invention, physiological signal data of a tester are inputted into the established train seat comfort evaluation model to obtain a comfort evaluation result. Therefore, the evaluation deviation caused by subjectivity of evaluation personnel is effectively reduced. The irrationality caused by a direct objective analysis method is avoided; and the comfort evaluation accuracy and feasibility of the train seat are greatly improved.

The invention discloses a hot continuous rolling strip steel convexity prediction method based on a gradient boosting tree model. The method comprises the steps: selecting technological parameters, equipment parameters, strip steel parameters and the actual strip steel convexity of related hot rolling strip steel as input and output of a hot continuous rolling strip steel convexity prediction model; collecting related original modeling data at a hot-rolled strip steel production site, preprocessing the data, and obtaining final modeling data by removing missing values and abnormal values and balancing the data; dividing the final modeling data obtained through preprocessing into a training data set and a test data set according to a certain proportion; based on the training data set, establishing a hot continuous rolling strip steel convexity prediction model based on a gradient boosting tree algorithm through cross validation; determining optimal parameters of the hot continuous rolling strip steel convexity prediction model by adopting a coordinate descent method; and evaluating the performance of the established hot continuous rolling strip steel convexity prediction model basedon the test data set. According to the method, the convexity of the hot continuous rolling strip steel can be accurately predicted, and the problem of large convexity deviation of the hot continuousrolling strip steel can be solved.

The invention discloses an image recognition method combining a convolutional neural network and a gradient lifting tree, and belongs to the technical field of mechanical learning. According to the method, a group of image patches are obtained according to the size of an input image. The image patches and an original image serve as input data together. Five branches are adopted. Each branch employs an improved VGG-19 model to extract features of an input image. Then, features are inputted to a gradient boosting tree for training to acquire a basic classifer. A weighted voting method is used for combining the basic classifier into a strong classifier to be used for classifying the input image. According to the method, medical images are identified and proved, the input medical images can beaccurately and rapidly classified, doctors are assisted to diagnose diseases, the diagnosis efficiency is improved, and therefore the misdiagnosis rate is effectively reduced.

The invention discloses a voltage sag reason recognition method based on an EM algorithm and a gradient boosting tree, and aims at solving problems that voltage sags having same kind and different kinds of sag characteristics cannot be classified accurately, data characteristics are hard to extract, and the like. On the basis of voltage sag waveform data and complemented by power grid fault data,thunder data and meteorological data, the per-unit value of power grid secondary voltage is accurately extracted through the EM algorithm, the characteristic of the waveform data in a statistical sense is extracted on the basis of the per-unit value, a gradient boosting tree model is trained in single-phase and three-phase data waveforms through certain training sample data, voltage sag reason recognition is achieved, and a voltage sag caused by a system side fault and a voltage sag caused by a user side heavy user load starting are clearly distinguished.

The invention provides a service robot movement system fault diagnosis method and device based on time sequence characteristics. The service robot movement system fault diagnosis method based on timesequence characteristics comprises the following steps that: adopting a sliding window way for original data to generate a time sequence training sample; adopting a plurality of random forest models and a plurality of gradient boosting tree models to carry out weighting fusion to form a random module, and utilizing a random model to carry out fault characteristic selection in the training sample;and inputting selected fault characteristic data into a GRU (Gated Recurrent Unit) neural network, extracting the time sequence characteristics of the fault, inputting the last moment data of the sliding window into a BP (Back Propagation) neural network, extracting the fault characteristics of a current moment, and carrying out fault classification after the time sequence characteristics and thefault characteristics of the current moment are combined.

The invention provides a numerical control machine tool energy consumption modeling and machining process optimization method which comprises the following steps: a data acquisition step: acquiring energy consumption data according to an energy consumption modeling experiment; establishing a no-load power model: fitting the machine tool no-load power model according to the energy consumption data,and measuring the machine tool no-load energy consumption; establishing a milling power model, namely training the milling power model according to a gradient boosting regression tree algorithm and the energy consumption data; a real-time power prediction step: superposing the no-load power and the milling power; a machining parameter optimization step: establishing a machining parameter optimization model by taking the machining cutting specific energy and the machining time as target functions, and solving the machining parameter optimization model; and a machining sequence optimization step: establishing a machining sequence optimization model by taking the sum of adjacent idle feed energy consumption as a target, and carrying out constraint. Energy-saving and efficient manufacturing is achieved. A numerical control machine tool energy consumption model is constructed through combination of formula fitting and a machine learning method, high prediction precision is achieved, and better generalization performance is achieved.

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 tobacco dampening machine discharge moisture content prediction method based on a gradient boosting tree and belongs to the tobacco field. According to the tobacco dampeningmachine discharge moisture content prediction method based on a gradient boosting tree. complete process parameters and environmental parameters of the tobacco dampening machine as independent variables, a dampening prediction model is established by using a gradient boosting tree algorithm, the influence of different influence factors on the discharge moisture content of the tobacco dampening machine is fully considered, and meanwhile, the prediction and interpretation capabilities of the prediction model on nonlinear characteristics of a dampening process are enhanced. The model can predictthe moisture content of the discharged material without human experience intervention in the working environment of the actual moisture regaining process, the precision of the prediction result is high, and the calculation speed is high. According to the prediction method, the discharging moisture content and the required water adding amount of the dampening machine can be automatically calculated, and then the temperature and humidity of tobacco leaves are more effectively controlled, so that the quality of tobacco shreds is improved.

The invention discloses a method and system for predicting power grid defect materials through multi-model fusion. The method comprises the steps: sequentially building a regression model, a negativefeedback neural network model, a gradient boosting tree GBDT model and an XgBoost model based on the fault probability strategies of different regions, time and equipment; uniformly inputting the acquired defective material data and the corresponding meteorological data into the regression model, the negative feedback neural network model, the gradient boosting tree GBDT model and the XgBoost model for training; respectively outputting prediction results corresponding to the models; and performing fusion processing on the trained multiple models by using a multi-model fusion strategy to form aprediction model, and solving an average value of the prediction result to obtain a final fusion prediction result. According to the invention, early storage of emergency repair material types, scales and places is guided, the prospective planning capability of material management of a power grid company is improved, and the cost of operation materials of a power grid enterprise is reduced whilethe operation reliability of a power grid is enhanced.

The invention particularly relates to a variational mode decomposition-based short-term power load prediction method. The method comprises the following steps of S1, obtaining load data and multivariate related data of a prediction day and three months before the prediction day; S2, preprocessing and associating data; S3, carrying out power load sequence modal decomposition; S4, judging the temperature correlation; S5, generating a feature vector of each component; S6, establishing an adaptive step length load prediction model; S7, establishing a power load prediction model by using an LGBM gradient boosting algorithm; S8, integrating prediction results; and S9, correcting a prediction result. The invention also comprises a power load prediction system which comprises a data acquisition module, a data preprocessing and correlation module, a load sequence modal decomposition module and a temperature correlation discrimination module to generate characteristic vectors of each component,a load fluctuation discrimination and model adjustment module, a prediction module of each component, and a prediction result integration module of each component and a prediction result correction module. The invention is suitable for the complex composition condition of each component of the power load, and is high in prediction precision, more flexible to use and good in universality.

The invention relates to the technical field of machine learning and to a method for predicting a heparin dosage in a continuous renal replacement therapy (CRPT) process, comprising the following steps: (1) installing relevant software and libraries; (2) analyzing and understanding the relevant information of a data set collected by a hospital; (3) preprocessing data; (4) extracting features, (5)determining a training set and a test set and processing unbalanced data; (6) predicting and evaluating a model; and (7) processing the abnormal values of predicted values. The method predicts the heparin dosage in the CRRT process by using a machine learning algorithm such as a gradient boosting regression model, and can give a reasonable and reliable heparin dosage reference value in the CRRT process by further processing the predicted value according to the trained model, thereby assisting the doctors, improving surgery quality, and achieving good economical efficiency and reliability.

The invention discloses an ocean buoy life prediction method based on a multi-class machine learning method, and the method comprises the specific steps: S1, building different buoy life prediction models based on the machine learning method, carrying out the feature selection of each hardware feature of a buoy, obtaining the static attribute of the buoy, and enabling the survival time of the buoyto serve as a dynamic attribute, jointly forming a data set for training a buoy life prediction model, and further evaluating the prediction accuracy of the buoy life prediction model; wherein the buoy life prediction model comprises a regression decision tree, a gradient boosting regression tree, a random forest and a support vector regression machine; and S2, respectively inputting the to-be-predicted data set into the trained buoy life prediction model to obtain four prediction results, and obtaining a final prediction result according to the four prediction results. According to the method, prediction results of various models are comprehensively considered to make an optimal decision, and the accuracy of the prediction method is effectively improved.

The invention discloses a power distribution network single-phase earth fault line selection method based on a gradient boosting tree algorithm. According to the method, data processing is carried outon a line zero-sequence current sampling value after a fault; according to the method, the zero-sequence current sampling value data of each line after normalization is obtained, then the current data of each line is used as the input of the gradient boosting tree model, the line corresponding to the maximum output value of the gradient boosting tree model is selected as the faulty line, and finally a good line selection effect is obtained. The method is not influenced by the network structure change and fault conditions of the power distribution network, and is of great significance to the fault diagnosis and safe and reliable operation of the power distribution network with a complex and changeable structure.

The invention discloses a network real-time intrusion detection method based on a fast gradient boosting tree model. The method comprises the following steps: training a fast gradient boosting tree classification model by using training data; capturing network traffic data in continuous time windows, wherein the original traffic data captured in each time window is used as a data block; performingstatistical analysis on data blocks to generate a plurality of feature vectors; classifying the feature vectors by adopting a fast gradient boosting tree classification model, and distinguishing normal behaviors from network intrusion behaviors; and if the behavior is judged to be the network intrusion behavior, outputting a network intrusion alarm signal. The category deviation problem of classification performance can be solved, the false alarm rate can be reduced, and meanwhile, the matching process of the model establishment stage and the decision stage meets the real-time requirement.

The invention relates to an image aesthetics quality evaluation method based on multi-domain knowledge driving, and the method comprises the steps: S1, designing a dense connection network as a backbone network, and extracting the aesthetics features of images; S2, designing a semi-supervised learning algorithm, learning style features from labeled and unlabeled images and extracting the style features of the images; S3, training a scene semantic classification model and a sentiment classification model by using the scene semantic classification data set and the sentiment classification data set, and extracting semantic features and sentiment features of the images; and S4, performing feature screening and fusion on the extracted features by using a gradient boosting algorithm XGBoost, andtraining an SVM classification model and an SVR regression model to predict the aesthetic quality of the image. According to the invention, the aesthetic quality prediction precision can be significantly improved.

The invention discloses a big data-based method and system for predicting the variable sliding time and relates to the field of civil aviation information. The method comprises the following steps: analyzing historical data in an airport collaborative decision-making system database, extracting flight departure operation data from the historical data, determining main influence parameters of flight departure, setting a weight value and dynamically adjusting a weight proportion; modeling by adopting a gradient boosting tree model to obtain a variable sliding time dynamic estimation model, inputting actual data of main influence parameters into the variable sliding time dynamic estimation model, estimating probability distribution of variable sliding time, and calculating an expected value;performing visual display on the variable sliding time. According to the method, the variable sliding time is intelligently predicted through manual prediction based on the constructed A-CDM system. The large-area flight delay is prevented and relieved through big data visualization auxiliary decision making.

The invention discloses a multi-scale difficulty vector classification method for graded reading materials. The classification method comprises the following steps: firstly, constructing word matchingfeatures, context features, topic features and the like to enrich feature representation; a light and comprehensive sentence difficulty vector is obtained in combination with the most prominent characteristic in previous research, and then is input into a classifier such as a GBDT (Gradient Boost Tree), so that a very good effect is achieved on educational graded reading linguistic data and general linguistic data. According to the method, feature representation is simplified, sentence difficulty can be reflected only through 21 vectors, multi-scale features are introduced, difficulty featurerepresentation is enriched, and model generalization is enhanced; a difficulty vector representation system suitable for sentence levels and article levels is constructed by combining newly used context information, and good effects are obtained in two data sets of the sentence levels and the article levels; the classifier uses a gradient boosting tree, the training speed is high, and a feature importance sequence can be obtained.

The invention discloses a dioxin emission concentration prediction method based on hybrid integration of a random forest and a gradient boosting tree, and the method comprises the steps: firstly carrying out random sampling of a training sample and an input feature for DXN modeling data with a small sample high-dimensional feature, so as to generate a training subset; then, establishing J DXN sub-models based on RF on the basis of the training subsets; then, carrying out I times of iteration on each RF-based DXN sub-model, and constructing J * I GBDT-based DXN sub-models; and finally, combining the prediction outputs of the DXN sub-models based on the RF and the GBDT by adopting a simple average weighting mode to obtain a final output. By adopting the DXN prediction model construction method integrating RF and GBDT, the DXN online prediction precision can be improved, the operation optimization of MSWI process operation parameters is assisted, and the economic benefits of enterprises are improved.

The invention relates to a DDoS attack detection method based on an SDN, belonging to the field of software defined networks. Normal and attack flow tables are collected in the SDN, and then the flow tables are processed; logistic regression, a support vector machine, K nearest neighbor, a decision tree, a random forest, integrated learning bagging, stochastic gradient, integrated learning Gradient and a limit gradient boosting tree are used for training the collected flow tables, and all optimal hyper-parameter models are put together; a function is called to calculate each index value, each value is put into a data model, an optimal parameter model is selected as a base model, and prediction results of the optimal model are integrated as a model; the collected flow tables are trained and predicted; the final integrated learning prediction result is obviously higher than the previous individual prediction result of each model; and finally, the trained integrated model is put into a controller, so that DDoS attack can be effectively detected, and effective defense can be carried out when an attack arrives.

A learning and discrimination device configured to perform learning by gradient boosting, including: a histogram memory unit configured to store, for each value of a feature amount, a histogram obtained by integrating gradient information corresponding to the value in sample data; a category accumulated gradient calculator configured to calculate a histogram corresponding to a value of the feature amount corresponding to a threshold for the feature amount as a cumulative sum of the gradient information; a non-category accumulated gradient calculator configured to calculate, as a cumulative sum of the gradient information, a sum of a histogram corresponding to a value of the feature amount that is classified using a classification method with the threshold different from the category accumulated gradient calculator; and a cumulative sum selector configured to output the cumulative sum calculated by the category accumulated gradient calculator or the non-category accumulated gradient calculator.

The invention relates to a temperature information prediction method and device. The method comprises the steps: acquiring observation data and mode prediction data of a meteorological station, wherein the observation data comprise a plurality of element values; generating an observation vector of the meteorological station based on the plurality of element values in the observation data and the corresponding observation time; generating fusion data of the meteorological station based on the mode prediction data; and inputting the observation vector and the fusion data into a temperature prediction model to generate temperature prediction information of the meteorological station, wherein the temperature prediction model is generated through a gradient boosting algorithm. According to thetemperature information prediction method and device, historical observation data and mode prediction data can be fused, the site data scale is expanded, and a more accurate temperature prediction result is obtained.