454 results about "Clustered data" patented technology

The invention relates to the technical field of coal mine informatization and data management, in particular to a data governance method and system for a coal mine and a storage medium, and provides a layered architecture from edge node preprocessing to central platform deep fusion in order to solve the problem of coal mine multi-source heterogeneous data governance. The edge nodes realize real-time anomaly judgment and output clustering data through an auto-encoder, a generative adversarial network and deep reinforcement learning; the central platform generates global collaborative model parameters and logic semantic data by using federated learning and a self-attention mechanism, then constructs a coal mine knowledge graph through unified semantic mapping and a graph convolutional network, and synchronously updates the digital twinborn platform with real-time verification data to generate real-time digital twinborn data. And finally, outputting an early warning decision under the action of a long short-term memory network and a self-attention mechanism, and performing traceability management in combination with a block chain technology, thereby realizing efficient and safe coal mine data management.

The invention discloses a multi-modal offshore wind power ultra-short-term prediction method in the field of offshore wind power plant cluster power prediction, and aims to solve the technical problems of spatial-temporal feature splitting and insufficient dynamic dependency relationship modeling. The method comprises the steps of performing anomaly detection and restoration on fan data, and generating a corrected wind power cluster data set; extracting a mean value, a standard deviation and a latest value of core operation data of each fan through a dynamic time window, and constructing a multi-dimensional node feature; a static geographic similarity matrix is generated based on geographic coordinates, a basic wake effect matrix is generated in combination with real-time wind direction data, correction is carried out through the maximum mutual information quantization time-delay effect, and then a dynamic adjacency matrix is obtained through self-adaptive fusion; and integrating the multi-dimensional node features and the dynamic adjacency matrix into a space-time diagram sequence data architecture, inputting the space-time diagram sequence data architecture into a multi-scale wake flow perception diagram space-time prediction model, and outputting a multi-fan power prediction value. According to the invention, high-precision multi-fan power prediction can be realized.

The invention relates to the technical field of electrical variable measurement, in particular to a power equipment system real-time dynamic monitoring method and system based on the Internet of Things. The method comprises the following steps: acquiring different data types; forming vibration sequences for the vibration signals, and obtaining fitting curve functions of all the vibration sequences; determining a trend sequence based on the variation coefficient of the vibration sequence; combining the trend sequence and the vibration sequence with a fitting curve function to determine a vibration trend coefficient; clustering the period based on the vibration trend coefficient, respectively obtaining a local change trend and an overall change trend based on cluster data and all data, and obtaining an equipment disturbance coefficient based on the local change trend and the overall change trend; and determining a prediction function based on the equipment disturbance coefficient, and obtaining a corrected vibration signal in combination with the vibration trend coefficient difference of adjacent periods, thereby completing dynamic monitoring. According to the invention, the monitoring accuracy of the power equipment fault is enhanced.

Hierarchical modelling and advanced feature engineering discover abnormalities in time series data with irregular trends. Data is collected in real time to ensure temporal integrity in the invention. Extraction filters and isolates useful data. Data cleansing removes noise and extraneous data after preliminary analysis identifies patterns and abnormalities. Feature engineering organizes cleansed data for machine learning algorithms. Primary storage stores this data for fast retrieval and extensive trend analysis. Holidays and weekends provide unique patterns in trend analysis. These trends are used to cluster data and create hierarchical predictive models, starting with a first-order model for general trends and increasing in order to refine residuals. Serializing these models improves storage and retrieval. Trend clusters are created from new data points, and algorithms detect pattern deviations. Statistical tests and machine learning classifiers identify anomalies and create alerts and remedial measures. The system monitors and analyzes incoming data to detect anomalies.

The invention relates to the technical field of driving control, in particular to a new energy automobile safe driving control method and system. The method comprises the following steps that a historical urgent danger avoiding data set and a corresponding user misoperation braking state are obtained through a new energy automobile control center, sample undersampling processing is conducted firstly, and an urgent danger avoiding balance sample is obtained; thirdly, analyzing user misoperation braking abnormity by using a balance sample, and performing disordered vector intensity calculation under a time dimension based on braking abnormity data to generate disordered vector segmentation clustering data; then, braking time sequence intervention correction sensing is carried out according to the clustering data, optimized braking intervention correction data are generated, finally, the optimized data are transmitted to a new energy automobile control terminal, and safe driving control is executed. According to the invention, the driving control technology is optimized, so that the driving control technology is more perfect.

The invention relates to the technical field of urban fire safety rescue, in particular to a low-altitude fire fighting system and method based on urban CIM and Beidou positioning. According to the system and the method, full space-time accurate positioning is realized through the Beidou differential base station and the UWB indoor positioning equipment; integrating the urban information model system and unmanned aerial vehicle cluster data, and constructing a multi-source heterogeneous data set; a coordinate system is unified based on a space-time reference calibration module, and fire behavior prediction, path planning and resource scheduling are realized through AI computing nodes; the unmanned aerial vehicle task distribution module is used for realizing multi-vehicle cooperative rescue; and iteratively optimizing system parameters through a closed-loop adaptive optimization module. The method is suitable for fire emergency rescue in an urban complex building environment, and the rescue response speed, the decision-making precision and the cooperative combat capability are remarkably improved.

The invention discloses a 3D printing process defect monitoring method and system based on a multi-modal large model. The method comprises the steps that S1, before related operation is carried out, corresponding preposition work needs to be completed, wherein the preposition work comprises knowledge graph creation, retrieval enhancement generation, reasoning and action framework and fine adjustment of the large model; s2, collecting and preprocessing industrial camera cluster data; s3, according to the preprocessed image, segmenting the image, constructing a region adjacency graph, finding an optimal region combination by applying a random algorithm, and combining some regions to obtain image representation; s4, calling the multi-mode large model subjected to pre-training and specific field fine adjustment to realize real-time defect monitoring in the 3D printing process; s5, a detection report is given according to a defect detection result, user interaction intelligent consultation is opened, and the model is updated and adjusted according to printing data; according to the method, the problems of standard, efficiency, trust and supervision of defect monitoring in the 3D printing process can be effectively solved.

The invention discloses a low-altitude economy-oriented decentralized unmanned aerial vehicle (UAV) task allocation method and system, and the method comprises the steps: firstly registering a UAV by a smart contract, starting an anti-Sybil attack mechanism, uploading task information and paying guarantee deposit by a publisher, verifying a task by the smart contract, and carrying out the delay decision. Secondly, the UAV obfuscates own geographic position coordinates and encrypts the geographic position coordinates, and the task publisher clusters the encrypted coordinates to obtain encrypted clustering data; finally, the intelligent contract calculates the probability that each UAV arrives at the task location according to the encrypted clustering data, the intelligent contract evaluates the matching degree of the UAVs, and a candidate UAV set is screened out; and according to a delay decision result, the smart contract performs task allocation on the candidate UAV set. According to the method, the motion states of the tasks and the UAV are considered, the reasonability and the performability of task allocation are ensured, and the accuracy and the real-time performance of task matching are effectively enhanced.

The invention provides a fast recovery method for hadoop cluster data, which relates to the technical field of erasure codes, and comprises the following steps of: calculating an importance score of a data block to be recovered, determining a target copy number based on a nonlinear copy number adjustment function, calculating a load weight in combination with network topology information, and selecting candidate nodes; and predicting a node fault probability by using a long short-term memory network, calculating a migration priority based on the importance score and the fault probability, and selecting an optimal target node to execute data recovery. According to the invention, the intelligence and differentiation of the data recovery process are realized, and the data recovery efficiency and reliability are improved.

The invention discloses an Internet of Things cluster data analysis method based on big data and artificial intelligence, and relates to the technical field of data analysis, and the method comprises the steps: carrying out the initialization and parameter synchronization processing of an edge side feature coding sub-model and a cloud space-time depth model through a space-time diagram sequence, obtaining a collaborative modeling parameter set, and obtaining a collaborative modeling parameter set; carrying out self-supervised reconstruction and prediction task training on the cloud space-time depth model based on the collaborative modeling parameter set, obtaining a depth representation model, learning abnormal mode parameters by using the depth representation model, constructing an abnormal scoring function, obtaining an abnormal detection model, and after receiving real-time multi-source heterogeneous data at an edge node, carrying out real-time reconstruction and prediction task training on the cloud space-time depth model; performing state prediction and anomaly score calculation by using the anomaly detection model to obtain a real-time anomaly score result; through the intelligent data analysis method, the accuracy, the real-time performance and the interpretability of Internet of Things cluster data analysis are improved.

The invention discloses a mine water inflow prediction method and system based on adaptive graph convolution, and relates to the field of coal mine safety. Comprising the steps of making a clustering data set based on historical data; training a water burst scene classifier by using the clustering data set; using the trained water burst scene classifier to identify historical data to obtain different water burst types corresponding to mine water burst; based on different water inflow types and the clustering data set, respectively training to obtain a water inflow prediction model; identifying the new water inflow association index data by using the trained water inflow scene classifier to obtain a target water inflow type; and predicting the new water inflow data by using the target water inflow prediction model to obtain the water inflow. According to the method, a unique adaptive graph convolution formula and a hydrogeological attention mechanism are introduced, the prediction capability and scene adaptability of the prediction model are improved, working face water inflow prediction under different hydrogeological conditions is considered, and important technical support is expected to be brought to coal mine water disaster prevention and control.

The invention discloses a data deduplication method and system based on incremental calculation and feature clustering and electronic equipment, and the method comprises the steps: receiving multi-source heterogeneous text, image, audio and video data streams in real time, adding a timestamp to each data unit, and generating an input data set with a time attribute; performing multi-modal feature extraction on the input data set, and dynamically adjusting feature weights of the extracted multi-modal features through a sliding window model and a time decay factor to obtain an increment feature vector set with the weights; executing two-stage clustering based on the feature vector set to generate a plurality of fine-grained data clusters; and comparing the intra-cluster data of each fine-grained data cluster in pairs by adopting a composite similarity model to determine intra-cluster repeated data, and performing data deduplication on the input data set according to the intra-cluster repeated data to obtain a deduplicated input data set. According to the method, the processing efficiency of the real-time data is improved, and the storage and calculation cost is remarkably reduced.

The invention provides a carrier-based aircraft intelligent re-flight decision-making method based on a Bagging algorithm, and relates to the technical field of carrier-based aircraft re-flight decision-making. According to the method, a carrier-based aircraft is simulated in different flight states to obtain carrier-based aircraft re-flight track clusters, a re-flight boundary criterion is used for carrying out flag bit definition on the re-flight tracks, and the carrier-based aircraft re-flight track clusters are obtained; therefore, a re-flight decision-making system is converted into a dichotomy problem of re-flight track cluster data, and in a carrier landing process, real-time flight state quantity of a shipboard aircraft is input into a re-flight decision-making model, and the model calculates whether current carrier landing is dangerous or not in real time. In the process, the shipboard aircraft can predict whether re-flight operation needs to be executed or not in real time and whether re-flight has risks or not. Therefore, the safety of carrier landing and re-flying of the shipboard aircraft is ensured. Finally, the method can effectively improve the safety of the shipboard aircraft during carrier landing and re-flight, ensures the carrier landing recovery efficiency, and lays a technical foundation for the carrier landing and re-flight decision of the shipboard aircraft.

The invention discloses a parameter identification method for an electrochemical-thermal-micro short circuit coupling model. The method comprises the following steps: constructing the electrochemical-thermal-micro short circuit coupling model; collecting multi-working-condition data of battery operation, and clustering the multi-working-condition data; performing feature extraction on the clustered data to obtain a feature data set; performing sensitivity analysis on parameters in the electrochemical-thermal-micro short circuit coupling model, and estimating the influence degree of the parameters on the characteristics; and parameters with high influence degrees are preferentially considered, probability influence evaluation is carried out on the parameters, and a parameter identification result is obtained. According to the method, on the premise of accurately judging the potential influence of the parameters, the accurate identification of the operation characteristic parameters of the battery system when the micro short circuit occurs can be realized.

A method and apparatus for assessing the hazard of rainfall-induced landslide clusters, and a computer device. The method comprises: acquiring historical rainfall-induced landslide cluster data from past intense rainfall events to construct landslide sample data; performing analysis on the basis of landslide attribute parameters and rainfall-induced landslide cluster data distribution information of the landslide sample data to obtain landslide development characteristics; on the basis of landslide geographical features and geometric features, using the landslide sample data and a plurality of candidate landslide-influencing indexes to analyze underlying patterns of landslide occurrence, so as to establish a landslide hazard evaluation index system; and performing correlation analysis on the plurality of candidate landslide-influencing indexes to determine a target landslide-influencing index, and combining the target landslide-influencing index with the landslide hazard evaluation index system to obtain a landslide hazard assessment model constructed on the basis of an automated machine learning framework. The efficiency of rainfall-induced landslide cluster hazard assessment can be improved and the accuracy of rainfall-induced landslide cluster hazard assessment can be effectively enhanced.

The invention relates to the technical field of load prediction, and provides a power load prediction method and system based on time sequence decomposition and an attention mechanism, and the method comprises the steps: carrying out the adaptive time sequence decomposition of an obtained original load sequence, calculating the sample entropy of each decomposed component, and carrying out the clustering; constructing a group of encoder and decoder networks for each piece of clustered data, performing parallel encoding to extract features, performing serial decoding reconstruction on the features from low frequency to high frequency, and outputting prediction data from low frequency to high frequency step by step; the weight is initialized based on the sample entropy, the trained weight is obtained through optimization in the encoder and decoder network training process, and the predicted value of the power load is obtained through weighted fusion. According to the method, adaptive time sequence decomposition, a weight mechanism guided by sample entropy and an attention-enhanced encoder-decoder structure are introduced, so that multi-component collaborative modeling and cross-scale dynamic prediction are realized, and the prediction accuracy and stability in complex load data and small sample scenes are effectively improved.

The invention discloses an aerial target velocity vector remote sensing method. The method comprises the following steps: transmitting an annular light beam with a linear divergence characteristic through a laser radar system; the receiving system is divided into two paths, one path obtains the distance of a target through a distance measuring detector, and the other path obtains the azimuth angle of the target through an area array sensor; establishing a three-dimensional coordinate system, resolving to obtain a spatial three-dimensional coordinate of the target relative to the laser radar system based on the divergence characteristic, the distance and the azimuth angle of the annular beam, and recording a timestamp; when the target passes through the annular light beam for multiple times, clustering data points based on spatial-temporal characteristics: data continuously collected at the same position belong to the same cluster; and averaging data belonging to the same cluster, and when two clusters of effective data exist and azimuth angles are different, calculating a velocity vector of the target relative to the laser radar system. By utilizing the method, the spatial position and the velocity vector of the target can be quickly and accurately obtained under a simple background.

The invention provides a distributed cluster security backup method and system oriented to a cloud-side environment, and relates to the technical field of distributed storage systems, and the method comprises the steps: obtaining cloud-side cluster data, and adaptively selecting a backup mode based on an erasure code and a multi-copy hybrid redundancy strategy; establishing a workload-aware incremental replication mechanism, and optimizing fault recovery by using idle resources; the problem of lock competition is solved by adopting a multi-copy consistency guarantee mechanism based on two-stage submission, and lock-free reading is realized based on MVCC. According to the invention, the data backup flexibility, the fault recovery efficiency and the system consistency guarantee capability are improved.

Systems and methods are disclosed for identifying relationships between complex datasets and / or high-dimensional datasets for predictive modeling. The method includes clustering data associated with one or more entities in a first dataset based on distance data; clustering the data associated with the one or more entities in a second dataset based on longitudinal data; consolidating the first dataset and the second dataset into a third dataset based on weights assigned to one or more edges between one or more nodes in the first dataset and the second dataset; generating a diagnosis space indicating a condition of the one or more entities based on the third dataset; and determining, via input of the diagnosis space into a machine learning model, an optimization of a weighting scheme for assigning weights to one or more features within the longitudinal data.

The invention relates to the technical field of point cloud registration, in particular to a multi-view point cloud data registration method and system. The method comprises the following steps: acquiring single-view point cloud data to perform adjacent field configuration construction to obtain local structure chart data; performing resonance mode analysis on the local structure diagram data to obtain diagram embedded data; performing geometric potential energy mapping according to the graph embedded data to obtain morphological stability distribution graph data; performing spectral domain supporting point family extraction according to the morphological stability distribution diagram data to obtain a skeleton topology node set; constructing cross-view structure chart data according to the skeleton topology node set corresponding to the multiple pieces of single-view point cloud data; performing semantic perception fingerprint generation on the cross-vision-field structure chart data to obtain structure recognition cluster data; obtaining different-source structure coupling graph data according to the structure identification cluster data; and performing low-degree-of-freedom rigid body calculation according to the heterogeneous structure coupling graph data to obtain point cloud registration data. According to the invention, the stability and precision of multi-view point cloud registration are improved.

The invention relates to the technical field of semiconductor packaging detection and repair, in particular to an FCWB copper column bump in-situ optical detection and laser repair integrated method and system, comprising: constructing a repair parameter database which comprises a plurality of repair parameter nodes, and the repair parameter nodes comprise fault repair parameters and reference clustering data; fault diagnosis is carried out on the initial chip to obtain a to-be-repaired chip or a first fault chip, if the initial chip is the to-be-repaired chip, fault diagnosis is carried out on the to-be-repaired chip to obtain diagnosis detection data, and after target detection parameters are retrieved in a parameter database by utilizing the diagnosis detection data, the to-be-repaired chip is subjected to fault diagnosis; and if yes, repairing the to-be-repaired chip by using a fault repairing parameter corresponding to the target detection parameter and a laser repairing unit to obtain a first repaired chip, and performing secondary diagnosis on the first repaired chip. According to the invention, the accuracy and the intelligent degree of parameter setting required by laser repair can be improved.

The invention provides a bacterial colony heterogeneous data separation method and device, equipment and a storage medium. Relates to the technical field of microbial colony heterogeneous data separation. The method comprises the following steps: performing hierarchical clustering on microbe single-cell Raman spectrum data to identify outliers; based on the outliers, performing de-noising processing on the single-cell Raman spectrum data of the microorganisms to obtain de-noised data; performing spectral clustering based on the de-noised data to obtain a spectral clustering result; wherein the spectral clustering result is used for identifying or marking bacterial colony cells in different growth periods; and evaluating the spectral clustering result by using a contour coefficient and a CH index to determine an optimal clustering number, and performing spectral clustering on the de-noised data based on optimal clustering data to obtain a final clustering result. According to the method, outlier detection, pruning optimization, spectral clustering and quantitative evaluation are combined, so that the precision and stability of microbe heterogeneity analysis are remarkably improved.

The invention provides a face image clustering method and device, equipment and a storage medium. Relates to the technical field of image processing. The method comprises the following steps: inputting required clustering data, wherein the clustering data is a high-order graph regular non-negative matrix; calculating and constructing a high-order similarity matrix; constructing a high-order graph constraint; defining and initializing each layer of basis matrix and representation matrix of regular non-negative matrix factorization of the high-order map; obtaining an updating formula of each layer of basis matrix and an expression matrix of the regular non-negative matrix factorization of the high-order map; updating the basis matrix and the representation matrix of each layer of the regular non-negative matrix factorization of the high-order map; obtaining a low-dimensional representation matrix; and outputting a clustering result. According to the invention, for the face image data with complex similarity, similarity information of more levels can be captured, so that the clustering performance is improved.

A power failure anomaly recognition method and apparatus. The method comprises: acquiring cluster data from a target cluster (S1); on the basis of the cluster data, acquiring distribution transformer data of each distribution transformer at each moment, the distribution transformer data comprising three-phase voltage imbalance, a three-phase voltage, the fluctuation amplitude of a phase current corresponding to voltage sag, a three-phase current, the sag amplitude of a three-phase instantaneous total active power, and the sag amplitude of each phase current (S2); on the basis of the distribution transformer data, screening out a corresponding distribution transformer that meets a first preset anomaly condition, so as to achieve preliminary anomaly recognition (S3); and, on the basis of the distribution transformer data, performing in-depth anomaly recognition on the distribution transformer that meets the first preset anomaly condition, so as to screen out a corresponding distribution transformer that meets a second preset anomaly condition and obtain a distribution transformer having power failure anomaly (S4). On the basis of the three-phase voltage imbalance, the three-phase voltage, the three-phase current and other distribution transformer data, preliminary anomaly recognition and in-depth anomaly recognition are performed; compared with manual inspection, the present method reduces manual workload and eliminates factors that affects power failure anomaly recognition such as limited human energy, thereby improving the timeliness and accuracy in power failure anomaly recognition.

The invention discloses a DNA sequence reconstruction method and system based on multi-scale attention and contrast learning, and relates to the technical field of DNA storage data reconstruction. Comprising the following steps: collecting a plurality of DNA sequence copies, screening out abnormal length sequences, and constructing a standardized clustering data set; performing one-hot coding and filling processing on the DNA sequence; extracting context dependent features and cross-sequence variation features; an Inter-Sequence multi-head attention mechanism is constructed to calculate the similarity between the sequences, and a weighted sequence tensor is generated; a global dependency relationship in the sequence is extracted through an Intra-Sequence multi-head attention mechanism; local offset features caused by insertion and deletion errors are extracted through a multi-size convolutional network; inputting a double-layer long-short-term memory network for sequence-level modeling, and outputting base reconstruction probability distribution; and constructing positive and negative sample pairs, calculating comparison loss, combining cross entropy loss to form a joint loss function, and outputting a high-precision DNA sequence reconstruction result. The method has high accuracy and robustness under the conditions of complex noise and multiple types of errors.

A clustering foundational AI model analyzes for, collects data about, and then outputs the role and / or function of an entity in a network and / or in an organization. The clustering foundational AI model clusters data together so that similar roles and / or functions can be readily identified to supply additional contextual information about the entity involved in the alert and / or event, and then outputs the role and / or function for the entity associated with the alert and / or event to assist in an investigation. The clustering foundational AI model adds the additional contextual information about the role and / or function of the entity upon receiving the alert and / or event. A UI receives the additional contextual information about the role and / or function of the entity in the network and / or organization and then presents both the alert and / or event and the additional contextual information that allows a user to gain contextual information about the alert and / or event.

Systems and methods are disclosed for predicting unnecessary resource utilization. A processor receives a first data object and generates for each member of a plurality of members a usage indicator for a pre-determined time period and a usage rate for the pre-determined time period. The processor generates each member of the plurality of members, based at least on the first classification data set, the second classification data set, the usage indicator, and the usage rate, a member optimization parameter. The processor generates based at least on the usage indicator, the usage rate, and the member optimization parameter for each member of the plurality of members, a plurality of cluster data objects, where members of each cluster data object are unique from members of any other cluster data object. The processor causes at least one of the plurality of cluster data objects to be displayed on a Graphical User Interface (GUI).

The invention discloses a web application firewall attack real-time traceability system based on real-time monitoring, and relates to the technical field of monitoring analysis, and the system comprises the steps: carrying out the attack behavior fingerprint clustering analysis according to the attack technique difference data, obtaining the attack behavior fingerprint clustering data, carrying out the cross-session attack association evaluation based on the attack behavior fingerprint clustering data, and carrying out the cross-session attack association evaluation based on the cross-session attack association evaluation. Cross-session attack association data is obtained, and attacker group portrait estimation is carried out according to the cross-session attack association data to obtain attacker group portrait data; performing context influence factor identification on the attacker group portrait data to obtain a context influence factor, and performing portrait correction on the attacker group portrait data based on the context influence factor to obtain corrected attacker group portrait data; and carrying out attacker traceability model construction on the corrected attacker group portrait data to obtain an attacker traceability model so as to execute attack real-time traceability and alarm response. The method has the effect of improving the attack tracing efficiency.

The invention relates to the technical field of foundation engineering filling, in particular to a method and system for automatically detecting the compaction settlement amount of a full-automatic heavy compactor, and the method comprises the steps: determining a historical change relation and a current change relation between soil settlement amount data and hammering pressure data detected by the heavy compactor during each hammering; according to the historical change relation, historical change relation curves corresponding to the different soil positions after being beaten by the heavy tamping machine for multiple times are determined, and the historical change relation curves are matched and clustered to obtain multiple sets of historical cluster data; and performing similarity calculation by utilizing the multiple groups of historical cluster data and the current change relationship to obtain target hammering pressure data, and regulating and controlling the heavy tamping machine by utilizing the target hammering pressure data. According to the method for automatically detecting the compaction settlement of the full-automatic heavy compaction machine, the operation progress can be judged and automatically adjusted, manual intervention is reduced, and it is ensured that the construction progress can be efficiently carried out with high quality.

The invention belongs to the technical field of medical apparatus and instruments, and discloses a mechanical intelligence-based incubator closed blue light irradiation data management system, which comprises a state cluster data module for collecting multi-source heterogeneous data, correcting clock skew between an incubator and a preset biosensor through a space-time alignment protocol, and sending the corrected clock skew to a server; dynamically adjusting the sampling rate of the multi-source heterogeneous data based on the infant physiological data in the multi-source heterogeneous data after time alignment, and outputting a state cluster data packet with a timestamp; the layered extraction module is used for constructing a three-level feature processing assembly line based on the state cluster data packet, and comprises a base layer for calculating a base statistical magnitude through a sliding window; the logic layer generates a temperature and light effect index and a sealing safety coefficient according to the basic statistics and in combination with the multi-source heterogeneous data; the decision-making layer establishes a treatment efficiency prediction matrix based on a temperature-light effect index and a sealing safety coefficient, and predicts a correlation trend of blue light irradiation and bilirubin metabolism; the safety and the curative effect stability of infant treatment are improved.