214 results about "Feature engineering" patented technology

The application discloses a circuit pre-training method and device based on a graph neural network and electronic equipment, comprising the following steps: for a large-scale circuit, a circuit network graph is constructed, a random walk road is generated, and the road is used as a sample for model training; a node2vec model is trained using the road sample; a model with the minimum loss function value is selected as an optimal model; and the optimal model is loaded, and representation vectors of all nodes are output and saved. In the application, the node2vec pre-training model of the graph neural network is used, the information of a long-distance standard unit is fully absorbed to absorb local circuit structure features, deep representation of nodes is automatically learned from the circuit structure, and tedious manual feature engineering is avoided, thereby providing high-quality representation vectors for a downstream model. The node representation learned by the node2vec model is a task-independent general feature, can be flexibly applied to various downstream tasks, and provides support for subsequent circuit analysis and optimization design work.

The application provides a kind of data-driven thermal power unit equipment health early warning method and device, comprising: constructing historical sample data set;The feature engineering extraction is carried out to historical sample data set, determine a plurality of sensitive characteristic parameters of the equipment health state, and construct the equipment health benchmark baseline based on sensitive characteristic parameters;From real-time operation data, the real-time characteristic parameters corresponding to sensitive characteristic parameters are extracted;The multidimensional health deviation of real-time characteristic parameters relative to health benchmark baseline is calculated;Multi-dimensional health deviation is input into the health early warning model pre-trained, and the current equipment health state score and predicted degradation trend are output by health early warning model;Combined with the preset grading early warning rule, determine whether to trigger early warning and the early warning level triggered.The application can effectively solve the problem that the traditional fixed threshold method is poor in adaptability under variable working conditions, the false alarm rate is high, improve the equipment fault prediction ability and the intelligent level of operation and maintenance decision.

The application provides an intelligent visual detection method and device for component defects, which comprises the following steps: loading a light standard three-dimensional model corresponding to a target component from a backend server; synchronously collecting real-time image data and depth data of the component on a conveying mechanism or a clamping mechanism; generating a real-time point cloud model for describing surface geometric features of the component based on the real-time image data and the depth data; matching the real-time point cloud model with the light standard three-dimensional model to locate a point area of interest in a high-defect area of the component; spatially mapping and fusing texture features in the real-time image data and the depth data in the point area of interest; and identifying defects in the point area of interest of the component through a joint probability model. The application can realize high-speed and accurate defect detection of components without complex feature engineering, and greatly reduces the maintenance difficulty and implementation cost of the algorithm.

The present application relates to a bridge technical condition prediction method based on the fusion of graph neural network and time series modeling, belongs to the field of traffic infrastructure maintenance and artificial intelligence application technology, and solves the problems of insufficient utilization of space-time characteristics, lack of engineering logic constraints and rough prediction granularity of existing methods. The present application first collects historical technical condition data and pre-processes to obtain structured space-time sequence data; a double-branch feature extraction network extracts spatial correlation feature vectors and time series evolution feature vectors in parallel; a degradation trend prior feature vector is fused through a gating mechanism to generate a comprehensive feature vector; a double-branch continuous ordinal prediction head is used to output the initial continuous prediction score of the target bridge in the prediction year; finally, a time series consistency post-processing algorithm is used for logical constraint correction to generate the bridge technical condition grade prediction result. The present application effectively utilizes space-time characteristics and has strong engineering logic interpretability, and can realize continuous and accurate bridge technical condition prediction.

The application relates to the technical field of big data, in particular to a battery fault prediction method in a new energy electric vehicle scene, which comprises the following steps: collecting electric vehicle battery related information data sets and preprocessing the same to obtain preprocessed data sets; based on the preprocessed data sets, an extended TrAdaBoost algorithm is used to construct a classifier; knowledge transfer is realized by using the classifier to obtain new data sets; feature engineering is performed on the new data sets, including basic features and business features; a replacement feature importance method is used to screen features; the screened features are input into an AutoGluon framework, and a battery fault prediction result is output; the extended TrAdaBoost algorithm is used to combine data in two different fields, and the knowledge of a source field is transferred to a target field through the idea of transfer learning, so that the generalization performance of a model in the target field is enhanced.

The application relates to the field of artificial intelligence and natural language processing technology, in particular to a document intelligent review method and system based on a multi-modal artificial intelligence and a rule engine, which solves the problems that an existing review scheme driven by a pure rule engine lacks semantic understanding and cannot understand context, a review scheme model based on a traditional natural language processing model has limited model generalization ability and still needs a large amount of feature engineering, a scheme based on a pure large language model for direct review has untraceable and unverifiable results, and the scheme has high delay and high cost, and the application comprises a document preprocessing and analysis module, a rule intelligent analysis module, a key information intelligent extraction module and a collaborative intelligent review engine. The application positions a large language model LLM as an enhanced component of a rule engine, utilizes the capability of the large language model LLM to understand natural language rules, extract complex information and perform deep semantic reasoning, and the framework of the whole review process, the final output of decision and the accurate positioning of results are dominated by the controllable and explainable rule engine.

This application relates to a method, system, and medium for diagnosing high-loss anomalies in distribution networks based on federated transfer learning. The method includes the following steps: dividing the distribution network into N regions, with each region acting as a federated learning client to collect distribution network data from each region; establishing a correlation analysis model between electrical characteristics and high-loss anomalies through data preprocessing and feature engineering to provide high-quality input data for federated learning; each client using a 1D-CNN+LSTM hybrid model for local training after data processing; a Bayesian hierarchical aggregation center receiving model parameters or update information from clients in each region, performing aggregation calculations to generate or update the global model; and rapidly deploying the global model to the target region clients for diagnosing the causes of high-loss anomalies in the target region. This application achieves efficient collaboration of model performance and lightweight, rapid deployment, significantly improving the automation level and operational efficiency of high-loss anomaly diagnosis.

This invention discloses a carbon emission prediction system and method for papermaking processes based on BO-GBDT. The system collects process parameters and energy consumption data during papermaking production, preprocesses the input data, constructs a feature set containing key influencing factors, and obtains a carbon emission dataset for model training. A Bayesian optimization algorithm is introduced to intelligently search for the optimal hyperparameters of four gradient boosting models, constructing a hyperparameter optimization space and setting an optimization objective function. Model performance is evaluated through cross-validation, and the optimal model is updated and selected. BO-GBDT is selected as the final prediction model. The trained model is then used to accurately predict carbon emissions from the papermaking process. By automatically optimizing the hyperparameters of the gradient boosting decision tree model using Bayesian optimization and combining it with multi-source data feature engineering, high-precision and high-efficiency prediction of carbon emissions from the papermaking process is achieved, providing an effective tool for carbon management in the production process.

The application discloses a kind of based on the post-investment risk early warning and exit decision auxiliary method and system of private equity investment project of stream data processing, belong to artificial intelligence and financial technology field.This method accesses multiple-source heterogeneous post-investment data in real time through distributed stream data framework, extracts static, dynamic and event trigger features after preprocessing using sliding window mechanism;Risk scoring fusion is carried out based on time series anomaly detection and gradient boosting tree model, risk classification and real-time early warning are realized;Quantitative exit decision scheme is generated by combining risk level and constraint condition through multi-objective optimization algorithm.The system includes multi-source data processing, real-time feature engineering, dynamic risk early warning and exit decision auxiliary module, can realize post-investment risk whole cycle monitoring and intelligent auxiliary decision of exit scheme, improve post-investment management efficiency, reduce investment risk.

This invention discloses a multi-feature fusion time series prediction method, system, device, and medium for power load peak prediction, belonging to the field of power system load prediction technology. It includes: acquiring and preprocessing raw power load time series data; extracting and constructing a load peak time series for feature construction to obtain a prediction feature set; constructing supervised learning samples to transform the time series prediction problem into a regression problem; and training the supervised learning samples using a machine learning model to obtain a load peak prediction model for predicting and evaluating future load peaks. The beneficial effects of this invention are as follows: Through a feature engineering method of multi-feature fusion, this invention can effectively capture the historical dependence, short-term volatility, and long-term periodicity in power load time series, thereby significantly improving the accuracy of load peak prediction. It has a predictive advantage, especially for power systems with large load fluctuations, in environments with large-scale integration of new energy sources.

The application discloses a type 2 diabetes heart failure risk prediction system and method, and belongs to the technical field of medical diagnosis and risk assessment. The prediction system comprises the following modules: a data and feature engineering module, which is responsible for standardization processing of data and screening of key prediction factors, and obtains a core feature set for machine learning; a model construction and selection module, which uses the core feature set and trains multiple machine learning algorithms in parallel; through cross-validation and comprehensive performance evaluation, the best model is selected as a prediction model; and a model deployment and application module, which converts output results of the prediction model into a clinically usable static nomogram or online tool, and performs visual output. The application predicts by integrating clinical variables and adopting a machine learning algorithm, and provides a static nomogram and a dynamic Web application, realizes heart failure risk assessment without relying on NT-proBNP detection, and can improve the prevention and management efficiency of cardiovascular diseases.

The application discloses a concrete pouring body element grid simulation method based on real-time vibrating data, relates to the field of digital technology for hydraulic engineering and building engineering construction, and specifically relates to a concrete pouring body element grid simulation method based on real-time vibrating data. The method is based on the self-adaptive division of bin surface body elements based on Boolean operation and spatial index; then, multi-source heterogeneous vibrating data is subjected to space-time registration and feature engineering extraction; and based on the vibrating quality body element mapping and evaluation of improved Euclidean distance and fuzzy comprehensive evaluation, finally, clustering analysis is carried out on the vibrating body element center points, the discrete points are aggregated into a vibrating area, and an energy distribution thermodynamic diagram is superimposed to realize visual display. The method makes up for the defects of the prior art, realizes efficient rendering and simulation display of real-time vibrating data on a bin surface three-dimensional model, and provides an efficient and intuitive management mode for a modern concrete pouring process.

The application provides a method for predicting process adjustment parameters based on machine learning and a readable storage medium, and the method comprises the following steps: obtaining a historical data set corresponding to a current process flow, wherein the historical data set comprises online data, offline data and corresponding process adjustment parameters; performing feature engineering processing on the online data and the offline data to extract derived features; performing regression training on the derived features and the process adjustment parameters by using a machine learning algorithm to obtain a prediction model; automatically screening out current online data and current offline data according to a production identifier, extracting corresponding derived features and inputting the derived features into the prediction model to obtain a predicted value of the process adjustment parameters. The technical scheme provided by the application realizes automatic, accurate and efficient prediction of process adjustment parameters by extracting derived features and performing regression training by using a machine learning algorithm to obtain a prediction model, and greatly improves the accuracy and efficiency of process adjustment.

This invention discloses a method for constructing a risk warning model for medical device testing based on big data. This invention relates to the field of data processing and risk warning technology, and solves the technical problem of failing to effectively correlate the deep nonlinear relationships between personnel behavior, equipment status, environmental fluctuations, and final test results, resulting in a large number of risk signals latent in process data going undetected. This invention introduces deep process features such as equipment stability indicators, personnel pass rate deviation, report modification rate, and environmental parameter deviation, enabling it to capture weak risk signals that traditional methods cannot detect. Five-dimensional feature engineering covers all key aspects of laboratory operation, and the unsupervised learning model can discover hidden anomalies beyond preset rules, compensating for the shortcomings of supervised learning in recognizing unknown patterns. The warning threshold automatically floats with recent data distribution, adapting to seasonal changes, new project launches, and other scenarios, avoiding frequent false alarms caused by a one-size-fits-all approach.

The invention discloses a kidney stone, hydronephrosis and pyosis detection system based on multi-scale feature fusion, and belongs to the technical field of medical detection, and the working process of the detection system comprises the following steps: S1, image data acquisition and preprocessing; s2, feature extraction and multi-scale feature fusion of the model; s3, classification decision and output layer design; and S4, model training and optimization. According to the system, from original DICOM image input to final diagnosis report generation, full-process automatic processing is achieved, manual intervention on feature engineering or intermediate steps is not needed, the system can generate a structured diagnosis report which comprises specific disease classification, confidence score and visual evidence heat map, the output format is normative, and the diagnosis report can be used for diagnosis. The method is easy to integrate with an existing image archiving and communication system of a hospital, and the model is clearly guided to pay attention to specific features related to ponding and infection in the learning process by introducing independent ponding and infection auxiliary discrimination branches and performing joint optimization with a main classification task.

The present application belongs to the field of port management and intelligent technology, and specifically relates to a port multi-category ship berth demand prediction and recommendation method and system based on an XGBoost model. The method comprises data collection and preprocessing, feature engineering and data fusion, model training and optimization, real-time data input and prediction updating, and feedback and model updating. The system comprises a data source module, a preprocessing module, a feature engineering and fusion module, a model training and optimization module, a prediction and recommendation berth module, and a result visualization and evaluation module connected in sequence. The method and system provided by the present application comprehensively analyze ship historical dynamic data, combine real-time data input and intelligent prediction models, optimize the configuration and operation scheduling of port berth resources, and thus provide more accurate decision support for port berth resource scheduling.

The present application relates to a transformer fault diagnosis method based on improved sparrow search algorithm optimization SVM, through the pretreatment and feature engineering of transformer oil chromatographic data; according to the data of pretreatment and feature engineering, the model based on oil chromatographic data is constructed, and the model parameters are optimized and trained; the input transformer oil chromatographic data is carried out transformer fault diagnosis based on the optimized and trained model.The present application improves the sparrow search algorithm by introducing the best point set strategy, the golden regular update rule, the differential mutation disturbance and the reverse learning mechanism, so as to comprehensively improve the parameter optimization process of support vector machine, the global exploration ability of the algorithm is enhanced, the risk of falling into local optimum is avoided, and the problem of premature convergence is effectively avoided.In addition, the improved algorithm pays attention to multi-link optimization, and the parameter setting is simple, the algorithm is low in use difficulty, so that the optimized support vector machine (SVM) can effectively improve the transformer fault prediction precision.

The invention discloses a method for predicting the delivery efficiency of an LNP delivery carrier based on data enhancement and an integrated learning model, and the method comprises the steps: collecting test data, carrying out the enhancement of the test data, and forming a data set of the delivery efficiency of the LNP delivery carrier; dividing a data set; building a model; model training and evaluation; predicting delivery efficiency; according to the method, through data enhancement, the diversity of a data set and the generalization ability of a model are remarkably improved; due to the application of the integrated learning model, the prediction accuracy and robustness are improved; the optimization of the feature engineering enables the model to more accurately capture the rule of the delivery efficiency of the LNP delivery carrier.

The application relates to the technical field of multi-modal information processing and intelligent health management, and discloses a multi-modal AI data tracing and early warning system and a portable sampling device thereof.The system comprises a multi-modal data preprocessing and feature engineering module, which is used for fusing spectrum, geographical space, image and text data into a unified high-dimensional feature vector; a dynamic space-time pollution atlas construction module, which is used for establishing and dynamically maintaining a nodal space-time graph structure; and a tracing analysis module based on a graph neural network, which is used for calculating a contribution score on the graph structure to determine the most important space-time source.The portable multi-modal sampling device is provided with a shell with a man-machine interaction interface, and a sample pretreatment module integrated with an integrated micro-flow channel and a replaceable unit is integrated inside.The application solves the problems of multi-source heterogeneous data fusion difficulty, poor real-time tracing and lack of a prediction mechanism, improves the real-time tracing analysis and increases the prediction ability of future risks.

The application relates to a kind of social platform robot detection methods based on multi-agent, comprising: creating social platform API connection instance to collect target user comment history data to obtain original data set;Build the orchestrator-multi-agent collaborative detection model driven by large language model;Data processing agent constructs comment level relationship to input data, and carries out cleaning, standardization and quality detection, forms the structured data set after preprocessing and writes into shared memory;Feature engineering agent reads structured data from shared memory, calculates user-level multi-dimensional behavior feature vector and writes into shared memory;Rule discrimination agent reads feature data, executes rule discrimination, judges whether feature hits piece by piece, thereby outputs rule risk score and writes into shared memory;Modeling agent reads feature data, trains anomaly detection and supervised classification model, executes multi-model fusion to obtain robot risk determination result and writes into shared memory;Explanation generation agent reads model detection result and rule data, and fusion outputs the final interpretable social robot detection result.The application reduces the dependence on single rule or single model by adopting multi-agent cooperation to complete data processing, feature construction, rule discrimination and multi-model fusion detection, greatly improves the accuracy and robustness of social platform robot detection.

This application discloses a method for predicting OTA upgrade completion rate and cycle time based on multimodal data fusion. The method first collects multimodal data on vehicles, users, environmental networks, and upgrade packages, and generates feature vectors reflecting multidimensional influencing factors through preprocessing and feature engineering. Subsequently, the feature vectors are input into a pre-trained multi-task learning model, which outputs in parallel the upgrade completion probability and predicted upgrade cycle time for the target vehicle. Finally, based on these prediction results, a global completion rate curve and optimization strategy are generated. This application overcomes the limitations of traditional methods that rely on single historical data by fusing multi-source heterogeneous data, achieving a shift from "post-event statistics" to "pre-event prediction." It can accurately predict key indicators of OTA upgrade activities and provide quantitative basis for proactive intervention and resource planning, significantly improving upgrade completion rate and operational efficiency.

The application discloses a surface water quality prediction method based on a self-attention mechanism coupled KAN network, and steps include: S1, integrating multi-source heterogeneous water environment data of a target basin, pre-processing and multi-granularity feature engineering are performed, and a three-dimensional space-time feature tensor is constructed; S2, a composite base function library is constructed, and an optimized base function set is obtained through double sparse screening; S3, the feature tensor is divided according to three feature channels, a space-time attention network is constructed, and a global attention weight matrix is calculated; S4, the base function set and the weight matrix are fused, a KAN network is constructed and is sparsified, and an Att-KAN coupled model is formed; S5, a multi-objective loss function is established, and the coupled model is optimized and trained in stages; S6, water quality space-time prediction is completed by inputting pre-processed target data, and attribution analysis and uncertainty evaluation are carried out. The application adopts the above method, fuses the advantages of the self-attention network and the KAN network, and realizes the unity of water quality prediction precision and model transparency.

The application relates to the technical field of energy storage, and discloses a distributed photovoltaic energy storage matching and operation method based on artificial intelligence, which comprises the following steps: acquiring initial multi-source data sets of a distributed photovoltaic system on the user side; based on the initial multi-source data sets, performing feature engineering and gradient boosting tree classification model processing to obtain a matching necessity probability value and to judge whether the user side needs matching; when the user side needs matching, performing optimal energy storage configuration on the distributed photovoltaic system by using a bidirectional long short-term memory network model and an improved genetic algorithm, and obtaining a next-day prediction data set and a target configuration parameter set; based on the target configuration parameter set and the next-day prediction data set, performing dynamic self-adaptive regulation and control method processing to obtain a target charging and discharging strategy of the distributed photovoltaic system; and using the target charging and discharging strategy to control the operation of the distributed photovoltaic system, so that the matching operation result is obtained, and the renewable energy consumption efficiency and comprehensive income of the distributed photovoltaic matching project are improved.

The application discloses an AI-based diode model parameter automatic extraction and optimization method, and belongs to the technical field of electronic design.The application solves the problem that the existing diode model parameter extraction method is seriously dependent on manual experience, the process is complicated and time-consuming, and it is difficult to stably obtain high-precision and strong generalization capability model parameters from nonlinear data under the influence of temperature and process deviation.The application constructs an intelligent mapping model from macro electrical characteristics to microscopic model parameters by adaptively denoising and multidimensional feature engineering on measured data, and finally generates a high-fidelity diode electronic design model by combining global optimization of physical constraints, so as to realize high-precision, high-efficiency and full-automatic extraction and optimization of key model parameters such as diode saturation current, ideal factor and series resistance, thereby improving the efficiency, automation level and cross-condition robustness of model extraction.

This invention discloses a method and system for detecting abnormal states of wind turbine generators based on Extreme Learning Machine (ELM). The method includes: real-time acquisition of multi-dimensional operational data from the wind turbine generator; data aggregation and timestamping; preprocessing of the multi-dimensional operational data including cleaning, feature engineering, and normalization to obtain a feature dataset; construction of an ELM model comprising an input layer, hidden layer, and output layer; construction of a training set and a validation set based on pre-set historical operational data of the wind turbine generator; training the ELM model using the training set and optimizing hyperparameters using the validation set; deployment of the trained ELM model at the edge of the wind turbine generator; real-time reading of the feature dataset and execution of inference to obtain anomaly judgment results, thus completing the detection of abnormal states of the wind turbine generator. This invention utilizes the ability to learn from large amounts of data to identify abnormal patterns such as excessively high temperatures and abnormal vibrations, improving detection accuracy and flexibility.

The present application introduces a kind of intelligent prediction method of coal mine water inflow quantity fusing multi-dimensional factors and deep time sequence model, it is related to wisdom mine and industrial big data analysis and prediction technical field, specific steps are: S100: the acquisition and feature engineering of multi-dimensional heterogeneous data;S200: data preprocessing and space-time alignment based on digital elevation model DEM;S300: feature dynamic weighted fusion based on gated attention mechanism;S400: water inflow quantity inference prediction based on deep time sequence model;S500: online closed-loop self-correction based on reinforcement learning.The present application compared with traditional single data source method, prediction accuracy is improved;Through digital elevation model, the space-time alignment relationship of surface water to downhole catchment point conduction is constructed, the prediction result is more in line with objective law;The unique reinforcement learning self-correction mechanism makes the model can learn and optimize independently according to real-time feedback, realizes the continuous evolution effect of more and more accurate, without manual intervention can adapt to the dynamic change of mine.

Systems and methods include reception of a dataset comprising a plurality of features, prompting of each of a plurality of text generation models to generate code to create one or more features based on the dataset, execution of the code generated by each of the plurality of text generation models on the dataset to create a first set of candidate features, discarding of non-interpretable features of the first set of candidate features to create a second set of candidate features, determination of a performance of a machine learning model trained using the second set of candidate features, and determination to add the second set of candidate features to the dataset based on the determined performance.

This invention relates to the field of forestry remote sensing monitoring technology and discloses a leaf SPAD prediction method based on UAV multispectral imaging. The method involves acquiring images of the target forest area through low-altitude multispectral aerial photography using a UAV, and extracting spectral data based on the locations of measured ground sampling points. Enhanced feature engineering is applied to the data to construct a multi-source feature set including original bands, basic vegetation indices, red-edge vegetation indices, and texture features. An early fusion strategy is used to construct a high-dimensional fusion feature vector, which is then input into an ensemble learning regression model based on ExtraTrees for training and prediction. Finally, a spatial distribution map of leaf SPAD values ​​in the forest area is generated. This invention effectively overcomes the shortcomings of conventional multispectral imaging, such as limited band size and saturation. By fusing the red-edge index and the ensemble learning algorithm, it significantly improves the accuracy and robustness of SPAD inversion, providing an effective technical means for precise management and rapid nutrient diagnosis of plantations.