77 results about "Historical model" patented technology

The invention, which belongs to the technical field of network intrusion detection, discloses an intrusion detection method and intrusion detection system based on sustainable ensemble learning. A multi-class regression model is constructed by using a class probability output and a classification confidence product of an individual learner as training data, so that the decision-making process of the ensemble learning has high adaptability to the attack type to improve the detection accuracy. At the model updating stage, parameters and decision results of historical models are added into the training process of a new model, thereby completing incremental learning of the model. According to the invention, on the basis of the ensemble learning fusion plan of the multi-regression model, the decision-making weights of the individual learner during the detection processes for different attack types are allocated in a fine granularity manner; and the parameters and results of the historical models are used for training the new model, so that the stability of the model is improved and the sustainability of the learning process is ensured. Besides, the experiment result is compared with theexisting MV and WMV plans, the accuracy, stability and sustainability of the intrusion detection method and intrusion detection system are verified.

A system and method for measuring, analyzing, and tracking the past performance of security analysts' earnings estimates and recommendations. The present invention provides a database of historical data relating to security analyst earnings estimate predictions wherein a historical model enables users to view the historical data as a time series of earnings estimates for each analyst selected, for a selected period of time, for a predetermined earnings event. Users may define a model to automatically create enhanced composite estimates wherein an improved prediction of the quantity being estimated, such as company earnings, revenue or cash flow is obtained. Users may view performance screens and historical performance data for a particular contributor or various contributors for a given security. Other views may be available.

This invention relates to one method to predict commute bus, which comprises the following steps: data collecting; data pre-processing; processing and establishing module; predicting road situation; predicting current certain time road section status combined with historical module database; predicting bus stop ; real time receiving bus route data to judge nearest bus detail position to predict the bus to bus each section future situations; computing bus each section consumption time to predict stop tie to update module.

The embodiment of the invention discloses an online model training method. The method comprises the steps of obtaining a training sample from streaming data, determining an objective function of the model according to the training sample, historical model parameters and non-convex regular terms, determining current model parameters enabling the objective function to be minimum, and updating the model according to the current model parameters. In the online training process, since the non-convex regular term is adopted to replace the L1 regular term for feature screening, the penalty deviationcan be reduced, effective features can be screened out, the sparsity is guaranteed, and the generalization performance of the model is improved. The invention further provides an information pushing method. The method comprises: obtaining user feature data and content feature data, based on the pushing model obtained by the online training model method, determining the probability that a target user is interested in target information according to the user feature data, the content feature data and the pushing model, and determining whether pushing is conducted or not according to the probability that the target user is interested in. The invention further provides an online model training device and an information pushing device.

The invention provides a content sensing method based on network stream behaviors. The content sensing method comprises the following steps: acquiring network flow in an outer network environment andextracting an observational characteristic as a training sample; training a model by utilizing the training sample; inputting a network stream with an unknown type into the model and identifying the content; carrying out incremental learning by utilizing recognized network stream flow and historical model parameters; updating the model parameters to ensure the continuity of model classes. According to the content sensing method provided by the invention, a dynamic modeling capability of a hidden Markovmoder model and a powerful nonlinear expression capability of a deep neural network are utilized; an experiment result shows the feasibility of the method and compares performance advantages of an existing technical scheme.

A system uses machine models to estimate trip durations or distance. The system trains a historical model to estimate trip duration using characteristics of past trips. The system trains a real-time model to estimate trip duration using characteristics of recently completed trips. The historical and real-time models may use different time windows of training data to predict estimates, and may be trained to predict an adjustment to an initial trip estimate. A selector model is trained to predict whether the historical model, the real-time model, or a combination of the historical and real-time models will more accurately estimate a trip duration, given features associated with a trip duration request, and the system accordingly uses the models to estimate a trip duration. In some embodiments, the real-time model and the selector may be trained using batch machine learning techniques which allow the models to incorporate new trip data as trips complete.

A method includes collecting individual-level data (for example, survey data) corresponding to each of one or more individuals who has been exposed to advertising for a product or service. A representative sample may be created from the individual-level data. A model may be created based on factors relating to acquisition of the product or service. A response to the advertisements is assessed based on the model. In some embodiments, intermediate measures to sales are included in the model. In some embodiments, the individual-level model is integrated with a historical model.

A system uses machine models to estimate trip durations or distance. The system trains a historical model to estimate trip duration using characteristics of past trips. The system trains a real-time model to estimate trip duration using characteristics of recently completed trips. The historical and real-time models may use different time windows of training data to predict estimates, and may be trained to predict an adjustment to an initial trip estimate. A selector model is trained to predict whether the historical model, the real-time model, or a combination of the historical and real-time models will more accurately estimate a trip duration, given features associated with a trip duration request, and the system accordingly uses the models to estimate a trip duration. In some embodiments, the real-time model and the selector may be trained using batch machine learning techniques which allow the models to incorporate new trip data as trips complete.

The invention provides an automatic dynamic bus scheduling system and an automatic dynamic bus scheduling method. The method comprises the following steps of: acquiring time for a bus to arrive at an initial station and a terminal station through global positioning system (GPS) data uploaded by a bus-mounted terminal, bus arrival and departure data and a historical model parameter; acquiring the departure interval of the next departure period through the time for the bus to arrive at the initial station and the terminal station and the operation index, bus state and departure rule data of the current line in a database; and acquiring the departure time point of the next period according to the departure interval, acquiring buses from an available bus set according to a queuing discipline, generating a departure plan, and sending the departure plan to a scheduling client. By the system and the method, the departure plan of the buses is adjusted automatically, the buses depart automatically, the work intensity of scheduling personnel is reduced, work efficiency is improved, the departure level is improved, a better travel experience is provided for travelers, and a technical foundation is laid for realizing centralized scheduling, field scheduling, unilateral scheduling and bilateral scheduling for enterprises.

The present invention provides a power grid model multi-version multi-tenant management system and method based on distributed storage. The method mainly comprises that a distributed file storage method is used to store and manage historical and future multi-service models of a power grid, so as to improve a storage capacity and availability of multiple versions of models; different versions of model data are stored by using an independent file set, so as to enable multiple users to build mutually isolated model storage and maintain work environments, and improve isolation and security of the multiple versions of model data; a data level segmentation method is used to carry out multi-file data level segmentation storage on a single version of model, so as to improve model data access performance; future model versions can be split and combined, so as to improve a multi-tenant collaborative modeling capability, and ensure online process management of future versions; and , historical versions of models are stored and managed according to differences between models, so that historical models can be managed, traced, and used.

The embodiment of the invention discloses a method and device for training a model. One specific embodiment of the method comprises the steps: receiving a model training request sent by a client, wherein the model training request comprises a target model type; matching the target model type with model types in a historical model configuration set; in response to successful matching, executing thefollowing first training steps: sending configuration data of a historical model corresponding to the successfully matched model type to the client; receiving first configuration modification data sent by the client for the configuration data; and determining a first initial target model according to the configuration data and the first configuration modification data, and training the first initial target model to obtain a target model, wherein model parameters of the first initial target model are determined according to model parameters of the historical model corresponding to the successfully matched model type. Therefore, the target model is obtained through training based on the configuration data of the historical model, and the model training efficiency is improved.

The invention discloses an enhanced message service (EMS) multi-professional-data holographic storage and panoramic accident inversion method, which comprises the following steps of: a) performing integrated version management on individual professional models, human-man graphical pictures and measurement point object data corresponding to the human-man graphical pictures of an industrial enterprise energy management and control system through a global version dictionary, and storing system historical models, graphs and data versions at all moments; and b) performing distributed storage on professional real-time data of the industrial enterprise energy management and control system as time sequence data formats into a high-speed database, wherein a triggering type or periodic sampling mode is used as a storage mode. The method is an effective holographic storage method for the multi-professional-data models, the pictures and the data of the industrial enterprise energy management and control system and an accident inversion method for a full panoramic process on the basis of the holographic storage method; and the problem that the conventional large-size industrial enterprise does not have an effective energy accident inversion analysis tool can be solved.

The embodiment of the invention discloses a method and a device for storing urban renewal and renovation data based on BIM, the method comprises the following steps: acquiring geometric data information and non-geometric data information of each building in a selected area; according to the geometric data information and the non-geometric data information of each building, correspondingly creatinga building model and storing the model; if the new model instruction of the building is detected, obtaining the new model and storing the new model in the storage area of the corresponding building;If the instruction of comparing the new model with the old model is detected, obtaining the building model and the new model of the selected building, and displaying the comparison results after the two models are compared. By modeling the historical data in BIM for the historical model and the current model, When comparing the new model with the old model, the new model can be intuitively compared with the new model to get the results of the comparison, which improves the information level of the building model data, and through the three-dimensional model, the model can be more intuitive andefficient to view.

The invention relates to a target tracking method based on a space coupling relation and a historical model. The main technical features are that a MIL model is established and is continuously updated in the tracking process; a target model pool is formed using the historical state of a tracking target; an optimal target model is selected according to a current frame and a target position is estimated; key points in a surrounding area of a target are detected, and a target position is predicted using space related information; and the target position estimated by the optimal model and the target position predicted by the space related information are fused to obtain a final target position. The design is reasonable, a multiple-instance learning tracking algorithm is taken as a basic tracking model, a group of target historical models are stored, and space context information is combined to assist tracking, and thus the invention achieves the excellent tracking effect, can deal with various environmental change conditions, and has good robustness and strong extendibility.

The invention discloses a BIM-based building model automatic matching method and system. The method comprises the steps of obtaining a BIM historical model database; obtaining a first block set of the first building; obtaining attribute information of each block in the first block set; obtaining a first block classification list according to the attribute information of each block; obtaining first category information of the first block classification list; inputting the first category information and each basic model information in the BIM historical model database into a model classification model to obtain category attributes corresponding to each basic model in the BIM historical model database; performing model matching on the first block set to obtain the basic model corresponding to each block in the first block set; and constructing the first building model according to the basic model. The technical problems that in the prior art, design of a building model is not intelligent enough, and the modeling efficiency is low due to the fact that the matching precision of an existing building model is low are solved.

The invention relates to a cross-boundary service demand analysis method and system and a readable medium. The method supports the modeling of the demands of a cross-boundary service from a pluralityof perspectives such as value, target, flow and service, achieves the cross-boundary service design under the guidance of value, and achieves the alignment of business and value. According to the modeling method, the mapping from a user target to a service is achieved, and a developer can be guided to rapidly develop a cross-boundary service meeting the user demand. The method has the beneficial effects that 1) historical model reuse is realized, a new model can be quickly constructed based on an existing model, and the demand analysis efficiency is improved; 2) modeling analysis is comprehensively carried out on the cross-boundary service demands from multiple perspectives such as value, target and service, and the method has excellent practicability, and 3) the subsequent cross-boundaryservice design and development work can be effectively guided.

The embodiment of the invention discloses a game interaction behavior model generation method and device, a server and a storage medium. According to the embodiment of the invention, N role groups anda historical model set corresponding to each role group can be obtained, and each role group comprises M different game roles; candidate models corresponding to the role groups at the current momentare determined in the historical model set; an interactive behavior of each game role is obtained by adopting the corresponding candidate model at the current moment; based on the interaction behaviorof each game role in the N role groups, candidate models corresponding to the role groups at the current moment are updated and trained to obtain updated candidate models of the role groups; the updated candidate models of the role groups are added into a historical model set; and the method returns to execute the step of determining candidate models corresponding to the role groups at the current moment in the historical model set until the candidate models converge, thereby obtaining interaction behavior models corresponding to the role groups. According to the scheme, the quality of the game interaction behavior models can be improved.

The invention provides an intelligent enterprise pollutant emission monitoring method. The method comprises the following steps of: collecting an electricity consumption data of a monitored enterprise, pollutant emissions of the monitored enterprise, geographic environment information and meteorological condition information of the monitored enterprise site, and generating a weighting value of pollutant emissions of the monitored enterprise, geographic environment information and meteorological condition information of the monitored enterprise site for the electricity consumption; and judgingwhether each factor plays a critical influence or not by the comparison between a newly generated model and a historical model. According to the intelligent enterprise pollutant emission monitoring method, for example, in the case of excellent diffusion conditions, a conclusion that the electricity consumption of the enterprise is not related to the pollutant emission can be obtained, and production-limited and electricity-limited are not needed for enterprises. At the same time, the intelligent enterprise pollutant emission monitoring method does not need to set too many assumptions beforehand and can directly analyze. Thus, the model can be ensured to be consistent with the real situation as much as possible.

The invention discloses a state-based VPN server intelligent distribution method, and the method comprises the steps: carrying out the sampling of the state of a VPN server cluster, obtaining the performance influence factors of a server, and building a historical model; sampling the performance of the server in real time; and obtaining a predicted value of a to-be-added user according to the constructed historical model, associating the predicted value with the real-time sampling performance to obtain the predicted performance, and allocating the server with the optimal predicted performance.And by adopting a machine learning method, the resource utilization rate of the VPN server in the target server cluster is comprehensively improved by taking the downloading speed as an index according to each real-time and historical performance index of the server. The optimal line is dynamically distributed for each newly used user, the server resource utilization rate is improved, server resource limiting waste is reduced, and the downloading speed of the user in the using process is increased.

The invention relates to a monitoring platform and a monitoring method for tunneling, and the monitoring platform for tunneling comprises a data processing unit which is characterized in that at least one construction section of a shield tunneling machine dynamic model is corrected in real time based on laser point cloud data, and the real-time correction degree is determined; determining a second sensitivity parameter, a second controllability parameter and a settlement monitoring range threshold value corresponding to the construction section in the historical model data set in a manner of substituting the construction section information into the historical model data set for matching calculation; therefore, settlement monitoring in the construction process is achieved through the stability judgment and the settlement monitoring range threshold value. Based on the monitoring platform, the invention provides a monitoring method for tunneling.

The invention discloses an automatic interpretation method for a telemetering slow variation parameter based on a historical data statistical property. The method comprises the steps of 1, constructing a historical model database; 2, interpreting the parameter; 3, unifying sampling points of an effective data segment of parameter historical data to the sampling points of the effective data segmentof target data; 4, estimating the target data of the parameter; 5, estimating a standard deviation of the parameter; 6, dividing the target data into four intervals: 0-1 sigma, 1 sigma-2 sigma, 2 sigma-3 sigma and 3 sigma-4 sigma according to an estimated value and the standard deviation of the parameter target data; 7, counting a probability distribution of the parameter target data in each interval; 8, obtaining a potential abnormality parameter table; and 9, checking potential abnormality parameters one by one by an artificial expert based on the potential abnormality parameter table obtained by the abovementioned step and according to an auxiliary decision graph obtained by analysis to finally determine telemetering abnormality parameters in the flight process. With the automatic interpretation method for the telemetering slow variation parameter based on the historical data statistical property, the defects of low artificial interpretation efficiency and data utilization rate areovercome and a lot of manpower is saved for a test process of a carrier rocket and an aircraft.

The invention relates to the field of federated machine learning, and discloses a prediction-based federated learning communication optimization method and system. The method comprises the steps thatfirstly, a global model and global variables needed in the method are initialized, each terminal user carries out local model training according to local data of the terminal user, and local model updating is obtained; then, a cloud center predicts the local model update of each terminal user according to the historical model update trend of each terminal user; then, a prediction error threshold value of each terminal user is set by calculating changes of prediction updating and global model loss functions adopted by the terminal user, and the prediction error threshold value comprises two steps of setting an initial threshold value and setting a dynamic threshold value; finally, a global model updating strategy is designed according to the set prediction error threshold value, and the cloud center adopts accurate prediction updating to replace local model updating to calculate global model updating. The problem of high communication cost caused by frequent transmission of update parameters between the terminal users and the cloud center in the federated learning technology is solved.

The invention relates to an optimized energy storage configuration method for a wind power plant. The method comprises the steps: S1, obtaining a wind power historical model value by using parameter historical data and a traditional wind power prediction model; S2, obtaining a wind power historical error according to the wind power historical model value and the wind power historical measured value; S3, minimizing the average absolute value of the wind power historical errors of different prediction points in the same wind field to obtain the optimal correction amount of the wind power historical errors; S4, combining the wind power historical error with the optimal correction amount to obtain a wind power correction historical error; S5, obtaining the energy storage configuration capacityand the energy storage configuration power according to the wind power correction historical error. Compared with the prior art, the method improves the accuracy of the grid-connected planned power of the wind power plant, and reduces the required energy storage configuration capacity and power.

The invention belongs to the technical field of machine learning, and discloses a model dynamic training, checking, updating maintenance and utilization method under a cloud platform. The resource manager obtains a workflow table according to different service requests and historical model training results; The model is verified by the verification data, and the result is notified to the resourcemanager; The service manager releases resources; And the resource manager re-issues the service to the scheduler of the service pool, and starts a new computing module for the service module. According to the invention, a lot of manual labeling cost is reduced; A large amount of model monitoring statistical data is obtained through the resource management module and used for solving the problem ofexploring and utilizing balance of the model monitoring statistical data and the original data, the model trained in the process and the original data are multiplexed to a certain extent, and after alarge amount of data is accumulated, a set of efficient workflow can be completed through excellent intelligent arrangement of the model monitoring statistical data. According to the method, hardwareresources are virtualized by utilizing the characteristics of a cloud platform, the characteristics of all functional modules are fully utilized, and the resources are utilized to the maximum extent.

The invention discloses a power system static equipment model time sequence construction method and device, and the method comprises the steps: adding an effective time domain and a failure time domain into an original model table to form a historical model table; copying the content in the original model table to the historical model table, and assigning the effective time domain and the failuretime domain of each equipment model in the historical model table to complete initialization of the historical model table; and in response to the external operation information, setting an effectivetime domain and a failure time domain of the equipment model corresponding to the external operation information in the historical model table, and completing modification or deletion of the equipmentmodel in the historical model table. According to the invention, the maintenance operation of the model table can be completely recorded, so that the model at any historical time can be quickly queried.

A system for interest rate risk management, comprising: an input device, configured to receive as input a first group of data indicative of a first group of financial instruments to be protected; a second group of data indicative of a second group of financial instruments aimed at protecting said first group of financial instruments; and an interest rate risk minimization device, connected to said input device, configured to receive as input said first and second group of data, a data feed of current market prices of said first and second group of financial instruments, a set of parameters of a term structure model, and historical zero-coupon term structures of interest rates, and to generate historical model errors and, considering these errors, the optimal amount to be invested in each financial instrument which shall be used to protect the balance sheet or portfolio. The interest rate risk minimization device is further configured to generate a residual risk estimation.

The invention provides a method and device for generating an investment portfolio, and relates to the financial field. The method for generating the investment portfolio mode provided by the present invention uses the historical model and the real-time model to jointly calculate the final investment portfolio mode scheme, which obtains the target set to be calculated first, and uses the historical model and the real-time model respectively in the target set to be calculated Calculate each element of each element, and then generate the first estimated result and the second estimated result of each element, and finally determine the most optimal portfolio mode based on these two estimated results. Since the real-time model takes into account the current market conditions, the calculated results are more suitable for the current market conditions, while also taking into account the stability of historical data.