114 results about "Concept drift" patented technology

The invention discloses a self-adaptive multi-dimensional adjustment metering box based on the Internet of Things, and relates to the technical field of intelligent monitoring of power equipment. The problem of measurement distortion caused by high fixed threshold false triggering rate, weak anti-aliasing capability and multi-source data asynchronization in a high-dynamic industrial scene in the prior art is solved. Multi-parameter acquisition is realized through the sensing acquisition module and the annular buffer area; a reconfigurable FIR filtering module is adopted to dynamically switch a low-pass / band elimination mode to suppress aliasing interference; noise features are extracted by combining spectral kurtosis analysis and concept drift detection, and a dynamic threshold value is generated by using exponentially weighted moving average and a sliding window standard deviation; a multi-channel time sequence is aligned through an IEEE 1588 protocol, and cloud parameter closed-loop optimization is realized based on extended Kalman filtering and a particle swarm algorithm; according to the method, the threshold fault tolerance, the high-frequency transient signal capturing precision and the multi-source heterogeneous data fusion reliability in a high-noise environment are remarkably improved.

The invention discloses an interdisciplinary scientific research potential assessment method based on a dynamic multi-modal knowledge graph, relates to the technical field of scientific research management and intelligence analysis, and aims at the dynamic requirements of college scientific research ecology by relying on an incremental assessment framework. Comprising the steps of multi-source data acquisition and incremental knowledge graph construction, time sequence trend identification, online and transfer learning optimization, scene simulation and multi-scene strategy verification and multi-modal data fusion. The map structure is updated in real time through concept drift detection and ontology extension, and emerging hotspots are identified by combining time sequence difference indexes such as citation acceleration and a cooperative network; and the accuracy and continuous iteration capability of the evaluation model are kept in multiple environments by means of online learning and a cross-domain migration strategy. Finally, potential evolution is quantified through scene simulation, strategy suggestions are output, the sensitivity to the cross domain is further enhanced with the assistance of multi-mode embedding, a closed-loop evaluation system from data to decision is formed, and college subject cross layout and resource allocation are assisted.

The invention relates to the technical field of the Internet of Things and artificial intelligence, and discloses an equipment maintenance fault intelligent analysis method and system based on the Internet of Things, and the method comprises the steps: representing an entity and a relation as a time function, and constructing a time sequence knowledge graph; monitoring vector change in the knowledge embedding space, detecting a concept drift phenomenon and mining an equipment performance evolution mode; constructing an equipment state transition model by using a recurrent neural network, predicting the future state of the equipment and evaluating the fault risk; collecting feedback processing knowledge conflicts and updating the model; generating multi-time-scale fault risk early warning and maintenance suggestions; according to the method, the knowledge acquisition efficiency is improved, dynamic knowledge support of the full life cycle of the equipment is provided, early recognition of gradient faults is realized, the method has adaptive knowledge updating capability, multi-dimensional fault analysis is provided, maintenance resource configuration is optimized, and the equipment management efficiency is improved.

The invention discloses an integrated learning classification method and system with a concept drift detection function, and relates to the technical field of machine learning and data stream processing, and the method comprises the steps: obtaining high-dimensional data of an extensible sliding window, carrying out the dimension reduction of the high-dimensional data of the extensible sliding window through PCA, and carrying out the extraction to obtain key feature data; performing real-time monitoring on the key feature data based on an adaptive window early warning detection method, automatically expanding an adaptive window during drifting early warning, and obtaining original data of the adaptive window when concept drifting occurs; and inputting the original data of the adaptive window when the concept drift occurs into a pre-established learning model based on an integrated Hough tree to obtain a trained learning model based on the integrated Hough tree, and performing real-time classification of data streams based on the trained learning model based on the integrated Hough tree. The accuracy of stream data classification can be improved, and the adaptive ability of the model to concept drift is enhanced.

The invention discloses a multi-head time sequence intelligent risk control method and system based on weighted trend and fluctuation, and relates to the technical field of financial risk management. Comprising the following steps: S1, collecting multi-channel time sequence data in real time, and carrying out data preprocessing; s2, calculating the global time weight, quantifying the multi-scale fluctuation stability of the channel pair, and judging the asynchronous alignment degree of the channel pair; s3, extracting an effective frequency band interval, calculating frequency domain characteristic parameters of the signal, evaluating frequency domain energy phase characteristics of a channel signal, and quantifying fluctuation states of a channel under different scales; s4, constructing a sparse coupling relation graph, quantifying an edge weight in the sparse coupling relation graph, and obtaining a network average coupling weight; and S5, evaluating the dynamic evolution characteristics of the channel risk state, and generating risk trend prediction and control suggestions. The problem that risk control accuracy is affected due to the fact that trend and fluctuation feature extraction of multi-source heterogeneous high-noise multi-head time series data is unstable under the condition of concept drift and multi-scale coexistence is solved.

The invention provides a cue word optimization method and system based on an approximate sub-module function and continuous learning, and relates to the technical field of artificial intelligence multi-mode perception.The method comprises the steps that a candidate cue word set is constructed, a combined objective function based on the property of the approximate sub-module function is designed, and a candidate cue word set is constructed; solving the combined objective function by adopting a greedy selection algorithm combining random disturbance, multi-round iteration and a task self-adaptive mechanism, realizing optimal selection of a candidate cue word set, gradually selecting a cue word with the maximum gain from the candidate cue word set, adding the cue word into an optimal subset, and when a new cue word is selected, selecting the cue word with the maximum gain into the optimal subset. And if the target cue words are selected, carrying out local optimization once, after all the target cue words are selected, carrying out joint optimization on all the cue words in a continuous space by adopting an alternate optimization strategy, and carrying out continuous iteration until the optimized cue words are obtained. According to the method and the device, efficient self-adaptive updating of the cue words of the language model is realized, so that the generalization performance and robustness of the model in zero-sample, few-sample and concept drift scenes are improved.

The invention provides a risk estimation method based on a knowledge graph, and relates to the technical field of risk identification, and the method comprises the following steps: S1, data collection; s2, performing entity identification; s3, constructing an atlas; s4, feature extraction; s5, risk model training; s6, risk prediction; according to the method, dynamic interaction key feature information is extracted from time sequence data contained in an existing knowledge graph based on time-frequency analysis, new features serve as additional information to be added into corresponding entities or relations, conversion from static data analysis to dynamic behavior pattern mining is achieved, and the dynamic behavior pattern mining efficiency is improved. And the situation that the meaning of some risk indexes may change and the concept drifts as time goes on is avoided, and a new view angle and technical support are provided for risk management of the financial industry.

A computer implemented method for on-line monitoring a structure, the method comprising: receiving (151) at least dynamics data (204) of dynamic variables representing a dynamic response of the structure from a plurality of sensors (10), the dynamic response of the structure being generated when the structure has been subjected to an excitation; detecting an impulse (152) created by said excitation by identifying a free load response (208) of the structure from said at least dynamic data (204); obtaining (153) a set of poles (210) comprising pairs of a natural frequency and a damping factor of the structure from the free load response (208); generating (154), using a non-supervised machine learning algorithm, a probability-based baseline model for the structure using a plurality of said poles (210) obtained along a certain period of time, the probability-based baseline model associating in a probabilistic way the dynamic variables affecting the structure and providing a plurality of baseline parameters (214); every time a new impulse is detected and a new set of poles (211) are obtained from the free load response of said impulse, feeding (155) the new set of poles (211) to a concept drift algorithm to obtain a KPI (217) of the structure.

Disclosed in the present invention is a lightweight active online modeling method fusing mechanism and dynamic memory, which comprises the following steps: S1, data preprocessing and model initialization; S2, online data representativeness evaluation based on dynamic memory of process data; S3, online data informativeness evaluation based on dynamic memory of label data; S4, online monitoring of a mechanism state; S5, parallel hybrid active sampling policy; and S6, online modeling of a soft sensor model. The present invention separately designs memory policies for process data and label data and only reserves high-value data, thereby satisfying a lightweight hardware requirement; in addition, the present invention designs a time delay-based mechanism state online monitoring method to detect a real concept drift of the process, and further provides the parallel hybrid active sampling policy for detecting both a real concept drift and a false concept drift; in combination with dynamic memory and a kernel incremental regression method, the computation amount for model updating and the hardware overhead for edge deployment are reduced.

The invention discloses an electric power and electric quantity anomaly detection method based on big data, relates to the technical field of electric power and electric quantity anomaly detection, and provides online concept drift detection, model adaptive updating, group anomaly judgment and multi-scale fusion recognition for the problems of high false alarm and missing alarm caused by dynamic change of power consumption behaviors of power consumers. Potential mode changes are found through a sliding window and a threshold value, and a suspected exception list is generated; correcting the short-term model by using incremental training and periodically updating the long-term model; then, multi-scale error measurement is fused into a multi-scale anomaly score, and key anomalies and conventional anomalies are judged in combination with a group deviation rate; and finally, self-adaptive adjustment is carried out through a threshold correction index and the like, and a manual annotation result is returned to the model library, so that accurate distinguishing and continuous learning of normal behavior transformation and real abnormity are realized. According to the scheme, the false alarm rate and the missing report rate can be remarkably reduced, and the reliability and the refinement degree of power dispatching and monitoring are improved.

The invention discloses an encrypted traffic adaptive update classification method and system oriented to an open network environment. The method comprises the following steps: firstly, extracting endogenous semantic features and exogenous environment features based on a causal decoupling mechanism, and stripping an environment confusion factor through anti-fact disturbance and invariance constraint to obtain invariant semantic representation; constructing a macroscopic drift state vector representing a network situation, and inputting a trained meta-learning super-network intelligent decision adaptive control hyper-parameter; performing cross-modal element calibration by utilizing large language model thinking chain reasoning, and calculating the subspace direction consistency of an instantaneous gradient vector of a candidate sample and a category optimization trajectory prototype so as to screen credible samples; and in combination with the capacity-limited playback queue, gradient orthogonal projection constraints are introduced to update low-rank adaptation layer parameters. According to the method, the concept drift problem is solved through causal decoupling and meta-learning decision, forgetting prevention is achieved through orthogonal projection updating, and the online adaptability and robustness of the model in the open environment can be improved without additional manual annotation.

A predictive analytics system and method in the setting of multi-class classification are disclosed, for identifying systematic changes in an evaluation dataset processed by a fraud-detection model by examining the time series histories of an ensemble of entities such as accounts. The ensemble of entities is examined and processed both individually and in aggregate, via a set of features determined previously using a distinct training dataset. The specific set of features in question may be calculated from the entity's time series history, and may or may not be used by the model to perform the classification. Certain properties of the detected changes are measured and used to improve the efficacy of the predictive model.

Embodiments of the present disclosure provide a method, a device, and a computer program product for updating a model. The method includes: determining a performance metric of a trained machine learning model at runtime; determining a homogeneity degree between a verification data set processed by the machine learning model at runtime and a training data set used to train the machine learning model; determining a type of a conceptual drift of the machine learning model based on the performance metric and the homogeneity degree; and performing an update of the machine learning model based on the type of the conceptual drift, where the update includes a partial update or a global update. In this way, a desired performance of the machine learning model can be maintained, while avoiding excessive time costs and computational resource costs caused by frequent global updates.

The invention discloses an interference prediction method and device fusing an incremental KD tree and a graph convolutional neural network. The method comprises the following steps: S1, extracting time sequence features of sparse historical interference samples; s2, constructing a dynamic graph structure according to an incremental KD tree model to capture a spatial relationship among sparse historical interference samples; s3, extracting a local sub-graph based on the spatial relationship between sparse historical interference samples, and inputting the local sub-graph into the graph convolutional neural network to obtain a prediction result; s4, obtaining a new sparse historical interference sample, and carrying out incremental parameter updating on the local subgraph based on a prediction result to obtain an incremental parameter updating result; and S5, judging whether a concept drift problem exists in the confrontation process, if the concept drift problem exists, adopting an online updating mechanism to adapt to environmental changes, then returning to S1, and if confrontation is finished, exiting. According to the method, the problem that interference behavior time change and space correlation are difficult to describe at the same time in a traditional method is solved, and the problem of calculation efficiency of a traditional graph model in a dynamic scene is solved.

The invention discloses a network intrusion detection method and system based on reliability sample selection, and the method comprises the following steps: carrying out the data enhancement of network intrusion data, and generating a more representative sample; then selecting a high-quality sample for initial model training through reliability evaluation; online training is carried out on the model, and when concept drift is detected, a reliability sample selection strategy guided by an attention mechanism is adopted, and samples valuable for model training are preferentially selected to be updated; when the number of the reliability samples is insufficient, selecting a sample with a high model loss value from the non-reliability samples as a supplementary sample; and inputting a to-be-detected sample into the trained model, and outputting a classification prediction result of the sample by the model. Evaluating the performance of the model by comparing a prediction result of the model with a real label; through the above steps, the abnormal behavior in the network intrusion can be effectively detected, and the adaptability and robustness of a network intrusion detection system are improved.

The present invention relates to a software defect prediction method, device and storage medium based on multi-model selection. First, the present invention uses the incremental learning method to train a random forest model M0 one by one. Then, the ADWIN concept drift detection mechanism is used to detect the dynamics of the sample mean, and the data collection mechanism is used to obtain data blocks D1 and D2. Again, the SMOTE algorithm is used to balance the class distributions in D1 and D2, and data blocks D1' and D2' are obtained respectively. For the obtained data blocks D1, D2, D1', and D2', four classification models M1, M2, M3, and M4 are respectively established based on the random forest model, and the best-performing model among the five software defect flow data classification models M0, M1, M2, M3, and M4 is selected as the final software defect prediction model M. Finally, the class of software defect data is predicted based on the M classification model of the random forest, so as to realize software defect prediction.

The invention discloses a flight landing time prediction method based on machine learning, and relates to the technical field of flight pre-judgment, and the method comprises the steps: carrying out the preprocessing of data based on historical flight data and historical meteorological data, and outputting feature sample data; performing data cleaning and standardization processing on the feature sample data, and outputting a training data set and a test data set in combination with actual landing and planned landing moment information of historical flights; establishing a landing prediction model, and performing optimization by combining reinforcement learning and a generative adversarial network; using the average absolute percentage error to test the landing prediction model and output a trained landing prediction model; and outputting a dynamically updated landing prediction model by adopting a concept drift detection algorithm, online gradient lifting and adaptive weighted fusion. Through combination of the deep Q network and multi-agent reinforcement learning, the prediction result is optimized in real time, and dynamic adjustment can be performed according to changes of actual flights.

The invention relates to the technical field of building automation and intelligent operation and maintenance, in particular to a building equipment fault positioning method based on a sensor network and a layer model, and the method comprises the steps: data collection and preprocessing, dual-time scale concept drift detection, dynamic model management and adaptive correction, and fault positioning and output. According to the method, system changes with sudden change and slow change can be captured at the same time through the double-time-scale design, the dynamic model pool and the historical concept reproduction recognition mechanism enable the system to memorize and reuse the historical operation mode, false alarms caused by periodic changes such as season switching are avoided, and the reliability of the system is improved. The problems that a traditional static or simple online updating model is poor in adaptability in a dynamic environment and the accuracy rate is reduced are fundamentally solved.

The invention relates to the cross technical field of artificial intelligence and network data analysis, in particular to a network data concept drift self-supervision detection classification method and system.The network data concept drift self-supervision detection classification method comprises the steps that label-free network data is obtained from a data stream, and a positive sample pair is constructed through a mixed data enhancement strategy; performing feature extraction on the positive sample pair and the negative sample pair by using a Vision Transform feature extraction network of a double-branch fusion hybrid expert model, and optimizing the model through an InfoNCE loss function; in the downstream task stage, a classification branch and a confidence coefficient branch are added in parallel on the basis of the pre-trained Vision Transform feature extraction network fused with the hybrid expert model, and the category prediction probability and the confidence coefficient are output respectively; calculating a confidence coefficient threshold value by utilizing samples which are correctly classified and wrongly classified in the verification set and combining a grid search method; and judging whether the sample data has concept drift or not according to a confidence coefficient threshold value. According to the method, the detection and classification performance of network data concept drift is remarkably improved in a few-label scene.

The invention relates to a dynamic semantic evolution tracking and concept drift adaptive updating method and system, and the method comprises the steps: obtaining scene demand information, and determining an information collection frequency and a historical data time window; obtaining target multi-source data and historical multi-source data; constructing an initial map; semantic feature vectors are extracted, and similarity matching is carried out; when the semantic similarity is lower than a similarity threshold value, updating the initial graph to generate an incremental graph; when it is monitored that the context distribution difference value of the nodes in the incremental atlas is larger than a drift threshold value, it is judged that a concept drift event exists, and a drift response strategy is triggered; according to the method, continuous tracking of semantic evolution and self-adaptive updating of concept drift are realized by constructing the initial map, matching semantic features, generating the incremental map and triggering the response strategy; the method has the advantages of tracking a semantic evolution process, adaptively updating the knowledge graph, responding to a concept drift event, and improving semantic analysis accuracy and real-time performance in a dynamic scene.

Approaches herein relate to model decay of an anomaly detector due to concept drift. Herein are machine learning techniques for dynamically self-tuning an anomaly score threshold. In an embodiment in a production environment, a computer receives an item in a stream of items. A machine learning (ML) model hosted by the computer infers by calculation an anomaly score for the item. Whether the item is anomalous or not is decided based on the anomaly score and an adaptive anomaly threshold that dynamically fluctuates. A moving standard deviation of anomaly scores is adjusted based on a moving average of anomaly scores. The moving average of anomaly scores is then adjusted based on the anomaly score. The adaptive anomaly threshold is then adjusted based on the moving average of anomaly scores and the moving standard deviation of anomaly scores.

Probabilistic models may be used in a deterministic workflow for robotic process automation (RPA). Machine learning (ML) introduces a probabilistic framework where the outcome is not deterministic, and therefore, the steps are not deterministic. Deterministic workflows may be mixed with probabilistic workflows, or probabilistic activities may be inserted into deterministic workflows, in order to create more dynamic workflows. A supervisor system may be used to monitor an ML model and raise an alarm, disable an RPA robot, bypass an RPA robot, or roll back to a previous version of the ML model when an error is detected by a data drift detector, a concept drift detector, or both.

The invention provides a self-adaptive network security situation awareness method and device combined with online learning, and the method comprises the steps: collecting real-time network state data and system load index data of a preset data source, carrying out the preprocessing of the data to form a multi-modal time sequence segment, inputting a pre-training time sequence data prediction model activated by employing Monte Carlo Dropout through a sliding window, and carrying out the prediction of the real-time network state data and system load index data. The method comprises the following steps of: calculating a prediction value of a next time period, outputting a prediction value and an uncertainty quantity of the next time period, calculating a threat probability through historical residual probability distribution fitting, realizing double-index risk assessment based on a preset threshold interval system, dividing into three types of states, and finally, respectively triggering online learning, configuration maintenance or intervention disposal flow for different states. According to the method, by introducing uncertainty quantized double-index evaluation and state-driven online learning closed loop, crossing of network security situation awareness from static detection to dynamic self-adaption is achieved, and the two core problems of insufficient real-time performance and concept drift in an edge computing scene are effectively solved.

This invention discloses a charger life prediction system based on big data analysis, belonging to the field of computer-aided design technology. It includes a data fusion unit, a benchmark modeling unit, a drift monitoring unit, a risk assessment unit, and a strategy generation unit. The data fusion unit generates structured feature vectors and sends them to the benchmark modeling unit. The benchmark modeling unit establishes a benchmark life prediction model based on the structured feature vectors and sends the benchmark life prediction model to the drift monitoring unit. The drift monitoring unit generates a conceptual drift index and sends it to the risk assessment unit. This invention profoundly reflects the inherent attributes and operating conditions of the equipment, and its accuracy and reliability far exceed those of traditional models relying on a single data source.

The invention discloses a load prediction method based on STL decomposition and concept drift detection, and relates to the technical field of power system prediction, and the method comprises the following steps: collecting historical power load data of a power system, checking whether there is a missing value in the collected data, and carrying out the interpolation processing of the missing value, decomposing the processed load data into three components, namely a trend component, a period component and a residual error component through STL (Standard Template Library) decomposition; secondly, carrying out concept drift detection on the trend component through an ADWIN algorithm, capturing load characteristics changing along with time, and dynamically adjusting the size of a time window trained by a prediction model; and under the dynamically adjusted time window, training a Transform-based prediction model for the three components, and adding prediction results of the three components to realize load prediction. According to the method, STL decomposition is carried out on the load data, the load data is decomposed into trend, period and residual components, refined model training is carried out for different components, and finally high-precision load prediction is realized.

The invention relates to the technical field of network security and information security, and particularly discloses an internal threat detection method based on entity behavior analysis, which comprises the following steps of: firstly, converting a multi-source heterogeneous log into a unified event semantic vector, and constructing a double-flow neural network architecture: extracting a long-term portrait vector from a long-period history by utilizing LSTM (Long Short Term Memory), and in the other path, a short-term intention vector is extracted from the neighbor session by utilizing Transform. And the short-term intention is projected to the semantic space of the long-term portrait to carry out consistency residual calculation, so that the hidden slow poison attack is effectively identified. In order to solve the problem of misinformation caused by feature dimension stability difference, a variance estimation and anisotropy measurement mechanism is introduced, portraits are modeled as probability distribution, variance vectors are utilized to carry out reliability weighting on residual errors, and global entropy regularization is combined to dynamically adjust penalty sensitivity, so that the probability distribution of the portraits is determined. Therefore, legal concept drift and malicious behaviors can be accurately distinguished in a complex dynamic scene.

This invention discloses a method and system for identifying encrypted traffic based on spatiotemporal features and semantic alignment. The method first extracts the spatial and temporal feature sequences of the network flow, and uses a byte-pair encoding algorithm to convert the spatial packet length into discrete symbols. Then, the discrete symbols and temporal features are mapped to a high-dimensional space and fused together. A global spatiotemporal feature vector is extracted through a network using a concatenated one-dimensional convolution and multi-head self-attention mechanism. Next, the text semantic bullseye matrix of fine-grained behaviors of various known applications is obtained offline from a large language model. Finally, the similarity between the spatiotemporal features and the text bullseye is calculated, and a multi-instance learning max-pooling mechanism is introduced for dynamic routing. Based on this, a contrastive learning loss function optimization model is constructed or cross-modal inference is performed. This invention completely overcomes the conceptual drift problem caused by changes in encrypted features, achieving extremely high generalization accuracy and feature interpretability across generations.

The invention relates to the technical field of computer security, in particular to a dynamically updated Android malicious software continuous learning detection method, which comprises the following steps of: acquiring and sequencing historical Android application samples according to time to construct a training set, and training a hierarchical comparison classifier; new applications are collected regularly to form a to-be-tested batch, and the current classifier is used for prediction; for each sample in the to-be-detected batch, executing pseudo loss uncertainty calculation to obtain an uncertainty score of the sample; according to the score, selecting a predetermined number of most uncertain samples for labeling, and obtaining a real label; adding the new labeled sample into the training set, and performing incremental training by adopting a hot start mode based on the weight of the current classifier to obtain an updated classifier; and circularly executing, and detecting the new application in the next period by using the updated classifier. According to the method, the labeling cost can be remarkably reduced, the concept drift can be effectively coped, and efficient, stable and continuous malicious software detection is realized.

The application discloses a dynamic cross-modal hash retrieval method and system based on concept vector learning, and the method comprises the following steps: constructing a concept vector matrix based on current label information and a Hadamard matrix; the concept vectors in the concept vector matrix are used for representing that samples with the same label have the same and invariable vectors; constructing a data similarity matrix based on current label information and historical label information; obtaining target hash codes based on the concept vector matrix and the data similarity matrix; obtaining a target hash function based on the target hash codes; analyzing the to-be-queried data according to the target hash function to obtain to-be-queried hash codes corresponding to the to-be-queried data; and performing similarity calculation on the to-be-queried hash codes and the target hash codes in a database to obtain target retrieval results. The application can overcome the defects of the concept drift problem in a dynamic data environment, improve the adaptability of a cross-modal hash model to the concept drift, improve the dynamic cross-modal hash retrieval performance, and can be widely applied to the technical field of information processing.