167 results about "Data balancing" patented technology

The invention relates to an automatic driving large model training optimization method based on multi-scene data balance. Comprising the following steps: (1) constructing a real vehicle high-speed driving scene library; (2) constructing a visual language automatic driving large model, training by adopting an iterative training framework based on a real vehicle high-speed driving scene library, designing a multi-task joint loss function and a weight adaptive adjustment strategy, realizing multi-task target balance, and obtaining a trained visual language automatic driving large model; (3) dynamic simulation is carried out for the automatic driving working condition, and a high-fidelity simulation data test set is constructed based on simulation data; and (4) performing hyper-parameter optimization and lightweight processing on the trained visual language automatic driving large model according to the high-fidelity simulation data test set to obtain a scene data balanced automatic driving large model. According to the method, the large model reasoning speed is increased, and resource occupation is reduced, so that the judgment capability of the large model on dynamic working conditions and high-risk scenes is remarkably improved.

The invention relates to the field of bioinformatics and drug development, and discloses an efficient siRNA effectiveness prediction method and system based on an XGBOOST algorithm. The method comprises the following steps: acquiring a training data set containing siRNA sequence information, chemical modification information and corresponding validity data; carrying out cleaning, feature extraction and data balance processing on the data; training the prediction model by using an XGBOOST algorithm, and optimizing hyper-parameters; performing model evaluation and optimization through cross validation; a new siRNA sequence is input for effectiveness prediction; explaining a prediction result by using an SHAP value analysis technology; the actual interference efficiency of the efficient siRNA sequence is verified through an in-vitro cell experiment, and the model is further optimized based on an experiment result. According to the method, the accuracy and efficiency of siRNA effectiveness prediction are improved, the influence of chemical modification is comprehensively considered, the problem of data imbalance is solved, the model interpretation is enhanced, and the calculation efficiency is remarkably improved.

The invention discloses a wind turbine generator variable pitch system fault diagnosis method for data balance. The method specifically comprises the following steps: acquiring operation data and fault data of a wind turbine generator through a monitoring and data acquisition system equipped on a wind power plant; based on the obtained data, screening sensitive features related to the fault of the variable pitch system, reducing the dimension of a data sample and reducing the input of redundant information; according to the quantitative relation, dynamic balance factors are distributed for various types of samples, the dynamic balance factors are introduced into a multi-stage diagnosis model Dualformer based on double Transformers, and a DBFDformer data balance fault diagnosis model is constructed; and training a DBFDform model by using the screened feature data, and completing dynamic updating of balance factors and optimization of model hyper-parameters through a snow ablation optimization algorithm. The wind turbine generator variable pitch system fault diagnosis method can effectively improve the diagnosis precision and performance of the wind turbine generator variable pitch system fault under the unbalanced data, has excellent generalization ability and robustness, provides a new thought for the fault diagnosis of the wind turbine generator variable pitch system under the unbalanced data sample, and is helpful for improving the utilization rate of wind energy.

The invention provides a cognitive ability decline detection method and system based on physiological indexes of wearable equipment, and relates to the technical field of feature selection and machine learning. Comprising the following steps of multi-dimensional physiological data acquisition, data preprocessing and time alignment, cognitive ability state label definition, feature engineering and data balance, cognitive ability decline detection model training and optimization, and output of cognitive ability state prediction. Multi-dimensional physiological indexes and time information of a user are collected in real time through a wearable device, and the physiological indexes comprise heart rate fluctuation features, heart rate statistical features, body temperature features, blood oxygen saturation features, motion data, electroencephalogram state features, skin electrical features, near infrared spectrum features and the like. The wearable device is used for integrating multiple types of physiological signal sensors, and continuous and non-inductive collection of multi-dimensional physiological data such as heart rate variability, electrodermal response and oxyhemoglobin saturation is achieved.

The invention discloses an electrocardiogram arrhythmia classification method and system based on a residual shrinkage network, and relates to the technical field of arrhythmia classification. According to the method, efficient electrocardiogram arrhythmia classification is achieved through multi-link cooperation, and a fine preprocessing, data balance strategy and multi-attention mechanism fusion model is designed; preprocessing provides high-quality input through wavelet denoising, precise R-wave detection and the like; the under-sampling-over-sampling mixed strategy is used for solving class imbalance and improving minority class recognition; the ResTCL-Net is fused with CNN, GRU, RCA, TSA and CLA modules, and signal features are mined in multiple dimensions; the optimization training strategy gives consideration to efficiency and stability, and is matched with comprehensive evaluation to guarantee performance. According to the scheme, the classification accuracy and generalization ability are remarkably improved, and abnormal heart beat recognition is enhanced.

The invention discloses a data processing method and device, equipment, a storage medium and a product, and relates to the technical field of intelligent driving, and the method comprises the steps: obtaining intelligent driving data, carrying out the parameter configuration of the intelligent driving data, obtaining the configured intelligent driving data, reducing the data analysis abnormality caused by the configuration error, and improving the data analysis efficiency. The data analysis quality and efficiency are improved; the configured intelligent driving data is processed to obtain intelligent driving model training data, so that a high-quality intelligent driving model training data set is finally obtained, an intelligent driving model with good performance can be effectively trained, and the safety and reliability of an automatic driving system are improved; the intelligent driving model training data is labeled to obtain the corresponding intelligent driving model training data under each label, the attributes and categories of the data can be visually described through the labels, and effective support is provided for data balance while the diversity of the intelligent driving data is ensured.

The invention relates to the technical field of nuclear power plant safety, in particular to a nuclear power plant fault diagnosis method based on data balancing, which establishes a data set balancing technology for deep learning through SA-TimeGAN, can effectively solve the problem of data imbalance caused by large normal data volume and small fault data volume in a nuclear power plant, and improves the fault diagnosis accuracy of the nuclear power plant. According to the method, firstly, small sample fault data are preprocessed, then, SA-TimeGAN is used for conducting adversarial training on the multiple types of small sample fault data, generated fault samples are added into an original training data set to achieve balance of sample data, and finally, the balanced data set is input into a CNN-BiLSTM diagnosis model for training, so that fault diagnosis is achieved. According to the method, the recognition capability of the model on minority classes of fault samples is effectively enhanced, and real data distribution is approached through adversarial training, so that high-quality synthetic data is generated.

The invention discloses an advertisement design effect evaluation method and system based on artificial intelligence, and belongs to the technical field of effect evaluation, and the method comprises the steps of data collection, data preprocessing, advertisement data balance processing, advertisement design effect evaluation model construction and advertisement design effect evaluation. According to the scheme, the number of data needing to be synthesized by each sub-cluster is obtained based on the average density, the importance of the advertisement data in the sub-cluster to which the advertisement data belong and the similarity of the advertisement data with other sub-clusters are comprehensively considered, the selection probability is obtained, new advertisement data is synthesized based on the cluster center and the random adjustment factor, and the balanced data set is constructed according to the spatial distribution entropy; the method comprises the following steps: extracting potential information by using one-dimensional convolution, adjusting an adjacent matrix, adding a self-connecting edge, calculating spatial information, carrying out feature updating based on a convolution layer, obtaining an output label, designing a self-adaptive division proportion to divide individual positions, and respectively updating the individual positions by using a sine and cosine algorithm. And the accuracy and practicability of advertisement design effect evaluation are improved.

The invention provides an industrial internet intrusion detection method based on a diffusion model. The method comprises the following steps: cleaning and standardizing original industrial internet flow data; taking the standardized industrial internet traffic data as input, generating a synthetic sample for the minority class of samples by adopting a multi-layer diffusion model, and sequentially carrying out anti-standardization and format adjustment processing and data balance sampling on the synthetic sample to obtain balance sample data; performing feature extraction and dimension reduction on the balance sample; the local time sequence features are processed through a multi-head attention mechanism, and global dependency features of the weighted key features are obtained; taking the global dependency feature of the weighted key feature as input, and extracting a multi-dimensional time sequence feature containing a long-term time dependency feature through a multi-layer bidirectional SRU module; and performing intrusion detection classification according to the multi-dimensional time sequence characteristics. According to the method, attack recognition can be efficiently and accurately carried out when high-dimensional and high-time-sequence network flow data and data category imbalance are carried out.

The invention relates to the technical field of data signal processing, and discloses a data equalization system and method, and the method comprises the steps: inputting an input signal into an improved equalization network in response to a signal compensation request, carrying out the affine transformation of the input signal through a neuron, and obtaining a transformed signal, according to a trainable coefficient of a B-spline primary function learned during affine transformation, a nonlinear activation function is determined, then weighted summation is carried out on transformed signals, nonlinearity is introduced through local interpolation and recursive definition of the B-spline primary function in the weighted summation process, nonlinear mapping signals are obtained, and then through judgment and error code detection, the non-linear mapping signals are obtained. And outputting the balanced data signal meeting the error rate threshold. According to the method, the input signal is primarily processed through affine transformation, nonlinear mapping is carried out by using the nonlinear activation function based on the B-spline primary function, and the complex high-dimensional signal processing problem is decomposed into the combination of a plurality of low-dimensional nonlinear functions in the process, so that the calculation complexity is reduced.

The invention discloses a PHM-oriented equipment state monitoring and predicting method. The method comprises the following steps: acquiring a sound signal in equipment through a sensor; transmitting the sound signal data to a storage and analysis module, and storing the sound signal data in a database after the storage and analysis module receives the sound signal data; the storage analysis module inputs the sound data into the preprocessing module for preprocessing operation; inputting the preprocessed sound signal into a diffusion probability model for data balance processing; performing feature extraction and classification on the balanced data by adopting a parallel dynamic convolution-sequential network model; and judging the health state of the equipment according to the classification result. A wavelet denoising module, a diffusion probability model and a deep learning model are integrated in an equipment health state monitoring system, and acquired equipment sound data are processed and analyzed step by step. According to the method, the accuracy and reliability of fault prediction can be effectively improved, and more accurate decision support is provided for health management of equipment.

The invention relates to the technical field of visual identification, and discloses an operation stage identification method based on a dynamic data balance strategy, and the method comprises the following steps: S1, collecting operation original data, and carrying out the annotation of the collected original data; s2, preprocessing the annotated original data by utilizing a dynamic data balance strategy so as to form a data set; s3, constructing an Xception-double flow LSTM (Long Short Term Memory) recognition model; and S4, training and optimizing the Xception-double-flow LSTM recognition model by using the data set. According to the method, the problem of bias learning is effectively relieved, the influence of data volume difference on a model learning process is effectively reduced, and deviation is reduced, so that higher classification performance is shown in an operation stage identification task; the method effectively improves the overall recognition precision of the operation stage, can provide more structured support for operation training, quality control and clinical decision making, and further lays a foundation for research and development of an intelligent operation auxiliary system.

The embodiment of the invention relates to the field of distributed technologies, and discloses a distributed vector data balancing method, device and equipment based on Redis and a storage medium, the method comprises the steps that at least one to-be-migrated data subset is obtained, and the data subset is obtained by dividing a vector data set according to data dimensions and the value domain range of each data dimension; obtaining an available resource total score corresponding to the at least one candidate Redis node in real time; sequentially storing the at least one to-be-migrated data subset to the candidate Redis node with the highest available resource total score according to a preset sequence; and generating an index directory according to the value domain range corresponding to the at least one to-be-migrated data subset and the correspondingly stored candidate Redis nodes. By acquiring the total score of available resources corresponding to each candidate Redis node in real time, the load of the node is fully considered, the data subset is stored to the node with lower load, the load balance of the candidate Redis nodes is realized, the utilization rate of each candidate Redis node is improved, and the stability of a system where the candidate Redis nodes are located is improved.

The invention relates to a data balancing method and device and computer equipment. The method comprises the following steps: determining a hard disk to be migrated out and a hard disk to be migrated in based on the access heat of data blocks contained in each hard disk; obtaining a to-be-immigrated hard disk sequence based on each to-be-immigrated hard disk; respectively determining to-be-migrated data blocks corresponding to the to-be-migrated hard disks to obtain a to-be-migrated data block sequence; traversing the to-be-immigrated hard disk sequence, determining a target to-be-immigrated data block matched with the to-be-immigrated hard disk in the to-be-immigrated data block sequence for each to-be-immigrated hard disk, determining a target data block in the to-be-immigrated hard disk, and performing data migration based on the target to-be-immigrated data block and the target data block, according to the technical scheme, data balancing and capacity balancing among the hard disks can be carried out on the basis of considering the access heat among the hard disks, and the reading flow can be uniformly guided to the hard disks, so that the access heat of the data is used as a basis for balancing, and the read-write performance of the hard disks can be improved to a greater extent.

The invention relates to a remote sensing satellite task constraint verification-oriented digital satellite system modeling method, which comprises the following steps of: establishing a basic support layer of a digital satellite system, and providing modularized simulation support service; establishing a model layer of the digital satellite system, realizing simulation operation calculation of each simulation subsystem, and generating simulation test data; a service layer of the digital satellite system is established, and the service layer forms a command and control interface interacting with a command and dispatch system, a measurement and control data transmission plan interface interacting with an operation and control system and a measurement and control instruction interface interacting with a measurement and control system on the basis of the basic support layer and the model layer; and completing system integration of the basic support layer, the model layer and the service layer based on a remote sensing satellite task constraint verification process. The digital satellite system established by the invention faces remote sensing satellite task constraint verification, and simultaneously considers satellite-ground resource window constraints such as measurement and control, data transmission segmental arc, energy balance, data balance, sunlight avoidance and the like and satellite in-orbit constraints.

The invention discloses a lung cancer patient dynamic survival prediction model based on multi-dimensional clinical data, and relates to the technical field of cancer survival prediction, and the key point of the technical scheme is that the dynamic survival prediction model comprises a core model module and a data support module, the core model module comprises a graph convolutional neural network model, a long-short-term memory network model and a generative large language model, and the data support module comprises three data processing systems, namely a time-dependent variable, a time independent variable and a data balancing strategy. According to the dynamic survival prediction model based on combination of traditional Chinese medicine and western medicine, abundant multivariate data in EHR can be fully utilized, individualized and precise diagnosis and treatment of the patient can be realized, artificial intelligence is assisted, and dynamic prediction of the survival rate of the lung cancer patient in one year, two years and three years can be effectively realized.

The present invention relates to a method for detecting network abnormal behaviors based on dynamic data balancing and generation, including: acquiring original network traffic data, inputting the original network traffic data into a dynamic abnormal data balancing model to obtain an enhanced data set; marking the original network traffic data based on the abnormal data balancing model, performing selective undersampling on the marked data, adjusting the ratio of normal data to abnormal data to obtain a balanced data set, generating new abnormal data according to the balanced data set, and merging it with the balanced data set to obtain an enhanced data set; inputting the enhanced data set into a TimeDiT denoising model for denoising to obtain a denoised target data set, and inputting it into an LSTM-Transformer weighted training model for detecting network abnormal behaviors, weighting the abnormal data in the target data set in combination with a weighted loss function to obtain a detection result. The present invention can improve the detection efficiency and accuracy of network traffic abnormal behaviors.

The invention discloses a federal learning data balance training method and device based on CVAE and a multi-head attention mechanism, and aims to solve the problem of unbalanced data distribution of local equipment and the problems of low precision and slow convergence of a federal learning model caused by Non-IID data problems in the fields of data security technology and mobile edge computing. The invention provides a federal learning data balance training method based on CVAE and a multi-head attention mechanism. Compared with an existing federal learning algorithm, the method has the advantages that local missing feature sample data is generated according to global feature distribution, so that the unbalance degree of data distribution between local devices is reduced. Considering that the local model deviates to feature extraction under Non-IID data, so that model training is trapped in a local optimal solution and even becomes worse, the teacher model is used to help the local model learn the features of the data, so that the student model is ensured to learn richer feature representation, and the generalization ability of the model can be improved.

The invention discloses a reservoir lithology identification method and device based on data balance. The method comprises the following steps: acquiring original logging data of a to-be-identified reservoir; inputting the original logging data of the to-be-identified reservoir into a pre-trained reservoir lithology identification model to obtain a reservoir lithology identification result; the reservoir lithology identification model is pre-trained and generated according to original logging data of historical oil wells and a training set of category labels of corresponding reservoir lithology, and the training set is a balanced sample training set obtained by analyzing features of minority class samples by using an SMOTE algorithm, synthesizing new sample data and adding the new sample data into the minority class. According to the method, the problems of model overfitting and low sample utilization rate in a traditional data balance algorithm can be solved, so that the identification effect of the model on minority class lithology is effectively improved, and the reservoir lithology identification precision is improved.

The invention provides a data balance processing method, system and device, equipment and a storage medium, relates to the technical field of computers, and aims to solve the problem of how to realize efficient data migration in a multi-engine multi-channel DMA (Direct Memory Access) and how to accelerate performance convergence during performance balance scene testing. The method comprises the steps that based on the number of engines of a direct memory access module, first equalization processing is conducted on data to be processed when the data are input into all the engines of the direct memory access module, and the first equalization processing is used for evenly distributing the data to be processed to all the engines; performing second equalization processing on the to-be-processed data allocated to the single engine when the to-be-processed data is input into each channel of the single engine, wherein the second equalization processing is used for uniformly allocating the data allocated to the single engine into each channel of the single engine; and transmitting the to-be-processed data to a data processing module connected with the direct memory access module through each channel of each engine of the direct memory access module.

The invention discloses a time sequence coal reservoir permeability prediction method based on improved deep learning, and belongs to the technical field of coal bed gas exploration and development and deep learning. According to the method, standardized preprocessing is performed on logging data, a time sequence sample is constructed by using a sliding window, stratified sampling is introduced to optimize K-fold cross validation, and an AdamW optimizer is adopted to train a double-layer LSTM network, so that the problems of low precision and poor generalization ability of a traditional prediction method are effectively solved. Experimental results show that the method can accurately capture the time sequence characteristics of the logging data, data balance is ensured through stratified sampling, overfitting is inhibited in combination with weight attenuation, and the prediction result has high stability and reliability.

The invention relates to the field of industrial network security, and particularly discloses a picture network intrusion detection method based on feature selection and data balance, which comprises the following steps of: firstly, acquiring an intrusion detection data set containing non-numeric data and numeric data, performing one-hot coding on the non-numeric data, performing maximum and minimum normalization on the numeric data, and performing one-hot coding on the numeric data; dividing a training set and a test set; performing feature selection on the training set by using a mixed feature selection strategy of a ReliefF filtering method and a Boruta packaging method, and performing resampling processing on the training set by using an SMOTE-ENN mixed sampling method; then, converting the resampled training set and the original test set into grayscale images, and performing training by using an OfficientNet image classifier to obtain a trained model; and finally, the model is used for carrying out classification prediction on gray level images of a test set. The problems of data redundancy, dimension disasters and class imbalance in an IIOT environment are effectively solved, and the accuracy and stability of an intrusion detection system are remarkably improved.

The invention provides a multi-center data standardization method for evaluating the performance of an AI ultrasonic model, and relates to the technical field of ultrasonic data processing, and the method comprises the steps: collecting original ultrasonic image data from different data sources, and carrying out the standardization preprocessing, and obtaining a standardization preprocessing image; a multi-task quality evaluation model is constructed and pre-trained; applying a pre-trained multi-task quality evaluation model to carry out quantitative evaluation on the image to obtain a quality evaluation index, and based on the quality evaluation index, carrying out quality screening by adopting a batch adaptive threshold strategy and a multi-stage screening mechanism to obtain a qualified labeled image; and layered sampling is carried out on the labeled images with qualified quality, and a standardized multi-center ultrasonic evaluation data set is obtained. According to the method, the problems that quality evaluation is difficult due to heterogeneity of multi-center ultrasonic data sources, a fixed threshold strategy cannot adapt to quality baseline differences of different data sources, artifact evaluation lacks spatial position consideration, and a multi-center data balance sampling mechanism is lacked can be solved.

This disclosure describes a module and method for configuring a generative adversarial network-based continual oversampling (GAN-COS) module to augment a training dataset. The augmented training dataset may be used to train a classifier module to detect defects in a dataset. During the configuration of the GAN-COS module, datapoints associated with tasks in a training dataset are balanced using a data balancing module before the balanced dataset is then processed by a shared generator and a private generator of the GAN-COS module. The latent vectors produced by these two types of generators then undergo latent space oversampling and are used in the training of a discriminator of the GAN-COS module. The latent vectors may then be combined and used to form the augmented training dataset.

The invention discloses an API (Application Program Interface) semantic vectorization Transform malicious code detection method and system. The method comprises the steps of obtaining a malicious code data set based on a PE file, extracting an API sequence and an MD5 value of a malicious code in the malicious code data set, and determining a malicious code family label according to the MD5 value; constructing a FastText word embedding model based on negative sampling optimization, and generating a word vector from the API sequence of the malicious code based on the FastText word embedding model; the method comprises the following steps: performing oversampling processing on a malicious code API vector by using a BorederLine-1SMOTE algorithm, and performing data balance by synthesizing minority class samples of a class boundary; constructing a Transform malicious code feature extraction model, and performing feature extraction and semantic analysis on the malicious code of the malicious code API vector; an average pooling layer and a full connection layer are added to the Transform malicious code feature extraction model, a malicious code detection model is constructed, and malicious code classification detection is achieved. According to the scheme, the malicious code recognition capability and detection performance are improved.

The invention discloses a real estate supervision method and system based on artificial intelligence. The method comprises the steps of data acquisition, data preprocessing, real estate data balancing, real estate property right risk assessment model construction and real estate supervision. The invention belongs to the technical field of real estate management, and particularly relates to a real estate supervision method and system based on artificial intelligence. According to an included angle between real estate data, favorable neighbors are screened out to construct a self-adaptive neighborhood set, similarity measurement is introduced to calculate the number of synthetic data, and the data are synthesized; detecting the quality based on the classification confidence of the synthetic data, and constructing a real estate data balance set according to the high-quality synthetic data; a dynamic frequency factor is designed to calculate a frequency adjustment loop learning rate, a dynamic cutting threshold is introduced to carry out gradient norm constraint, depth Q network parameters are updated, a real estate property right risk assessment model is constructed, the real-time requirement is met, and it is ensured that different types of real estates can be equally and effectively supervised.

The invention discloses an intelligent equipment management system integrating charging and data dump, and relates to the technical field of equipment management, and the system comprises an identity authentication module which is used for carrying out the dual authentication of a user identity through a biological feature and an authority token, extracting the three-dimensional feature of a task from a production system, and storing the extracted three-dimensional feature; calculating the adaptation degree of the task and the equipment through a feature matching algorithm to generate a candidate equipment pool; the working condition sensing module is used for generating working condition fingerprints, and comparing historical fingerprints through a dynamic time warping algorithm to judge an abnormal level; the charging transmission and storage module is used for constructing an energy data balance model according to the priority score and the task characteristics, dynamically adjusting the charging power and the dump bandwidth, and scheduling equipment through a predictive migration mechanism; and the resource management module is used for constructing a full file of the equipment, dynamically adjusting an equipment matching rule and a user permission based on an association rule algorithm, and generating a predictive maintenance work order, so that the use safety, the matching accuracy and the maintenance efficiency of the equipment are remarkably improved.

The invention relates to the technical field of enterprise data analysis, in particular to an enterprise data intelligent analysis and traceability method, which comprises the following steps: performing enterprise data acquisition on a target management enterprise to obtain initial enterprise data, obtaining historical enterprise data of the target management enterprise and classifying the historical enterprise data to obtain a plurality of enterprise data analysis areas, according to data fluctuation parameters determined by the enterprise data in the current detection period of each type of enterprise data analysis area, determining an abnormal trend type; selecting and determining an analysis mode corresponding to the enterprise data analysis area according to the abnormal trend type so as to determine abnormal enterprise data, performing enterprise data balance verification on the data source, and determining whether a current enterprise data statistical inspection result is valid or not according to a verification result so as to determine whether the abnormal data source is traced or not. According to the invention, through a layered and classified enterprise data intelligent analysis and dynamic traceability mechanism, the accuracy and efficiency of data anomaly detection are significantly improved.

The present invention discloses a mine environment monitoring method and system based on machine learning, and the method includes: data collection, data preprocessing, mine environment data balance, construction of a mine environment status assessment model and real-time monitoring of the mine environment. The present invention belongs to the field of environmental monitoring technology, and specifically refers to a mine environment monitoring method and system based on machine learning. The scheme designs weighted standard distance and exponential decay distance to obtain the total distance between data, screens out noise data, determines the number of data to be synthesized based on discreteness and nearest neighbor data, synthesizes new data based on discrete data sets and introduces random noise; designs local density and global density to obtain total density, calculates distance threshold based on minimum distance, calculates density threshold based on adjustment factor and total density, constructs candidate center data set, obtains cluster center data set according to total density and number of overlapping data, improves the quality of the entire assessment model, and is conducive to accurately assessing the state of the mine environment.

The invention discloses a femoral head necrosis prediction system and method based on Transform, and belongs to the technical field of image processing, the system comprises a data acquisition module, a data processing module, a data balancing module, a feature extraction module, a feature merging module and a necrosis classification module; the method comprises the following steps: acquiring preoperative hip joint CT data, pre-processing to unify specifications, adopting directional data enhancement to solve the problem of class imbalance, extracting features by virtue of Video Swin Transform, merging and modeling a spatial relationship through volume blocks, and finally outputting a necrosis classification result. According to the femoral head necrosis prediction system and method based on Transform provided by the invention, the prediction accuracy and robustness are effectively improved, the problem of model instability caused by strong subjectivity, hysteresis and data imbalance in the prior art is solved, an objective and efficient prediction tool is provided for clinic, and the risk of postoperative complications is reduced.