1362 results about "Task learning" patented technology

The invention provides a method and system for analyzing defects in wafer manufacturing based on big data, and relates to the technical field of big data, and the method comprises the steps: collecting technological parameter information, defect scanning images and manufacturing process data information, and carrying out multi-scale feature pyramid network processing and fusion to obtain fusion features; the fusion features are analyzed through a deep learning network, the deep learning network adopts a multi-task learning framework to complete a defect classification task, a defect segmentation task and a defect evolution prediction task at the same time, consistency loss based on graph matching is introduced, and a defect type recognition result and a feature map of defect space distribution are obtained; based on the characteristic spectrum, when it is detected that the process parameters are abnormal, analyzing historical defect data by using a graph neural network containing a time sequence attention mechanism; technological parameter constraint conditions are set in combination with an expert knowledge base, and technological parameters are adaptively adjusted through a reinforcement learning method. According to the invention, the accuracy of defect detection and classification is improved.

Provided in the present application is an intelligent question answering method based on a collaboration between a large language model and a knowledge graph, relating to the technical fields of artificial intelligence and natural language processing, the method comprising: decomposing a complex question into a plurality of simple questions, and analyzing the degree of association between the simple questions and a basic function so as to form a multi-hop reasoning path; automatically extracting structured information from the simple questions on the basis of a multi-task learning framework of a large model, so as to construct a knowledge graph; and constructing a cumulative reasoning learning framework on the basis of a logic reasoning large model, and performing iterative verification on a process result formed by the knowledge graph on the basis of the multi-hop reasoning path, so as to correct the reasoning path until a correct answer is inferred.

The invention provides a textile defect positioning method, which comprises the following steps: according to an extracted defect feature map, fusing feature maps of different scales, extracting multi-scale defect features, and comprehensively depicting space distribution and combination information of defects; aiming at the condition that multiple flaws exist at the same time, a multi-task learning model is constructed, flaw detection, classification and positioning tasks are carried out at the same time, and a back propagation algorithm is adopted to optimize the multi-task learning model; positioning the defect area by adopting an image segmentation algorithm, obtaining the contour of the defect area according to the positioning result of the defect area and by adopting a Canny edge detection algorithm, and calculating the area and position information of the defect; and based on an evaluation result, a dynamic threshold adjustment algorithm is adopted, a defect judgment threshold is adjusted according to a production batch and a product type, and a threshold parameter is dynamically updated by analyzing historical data and a current production condition.

The invention discloses a method for correcting dialogue intention information of old people based on a deep neural network, and particularly relates to the technical field of dialogue optimization, and the method comprises the steps: collecting and cleaning multi-topic multi-round dialogue data, marking multi-round dialogue corpora, and unifying the format; performing feature extraction on the labeled corpus by using the multi-intention recognition sub-model to generate a multi-intention representation vector capable of representing each intention; screening context content related to the target intention in combination with a hierarchical attention mechanism of user context dynamic weight adjustment; processing context content related to the target intention through gating operation based on multiple rounds of interaction history, and reserving semantic fragments with reference value; performing parallel analysis on the reference semantic fragments based on a multi-task learning structure of a deep neural network, performing classification processing on different intentions, and calculating scores; and executing multi-intention fusion correction operation according to classification and scoring results, and outputting intention information capable of distinguishing each intention.

The invention belongs to the technical field of intelligent perception and mode recognition, and particularly relates to a human body behavior recognition method based on millimeter wave radar and multi-task learning. The method comprises the steps that millimeter wave radar point cloud data and RGB video streams are synchronously collected, and a five-dimensional point cloud scene is generated through three-dimensional analysis and dynamic target extraction; a hierarchical Point Transform network is constructed to extract radar space-time features, and human body key point features are generated by using visual auxiliary attitude estimation; radar and key point features are dynamically fused through a cross-modal attention mechanism, a multi-task joint optimization strategy is combined, behavior classification serves as a main task, attitude estimation serves as an auxiliary task, and a behavior recognition result is output. According to the method, through multi-task cooperation and cross-modal feature interaction, on the premise of ensuring privacy security, the accuracy and robustness of human behavior recognition in a complex scene are remarkably improved, and efficient and reliable technical support is provided for the fields of intelligent monitoring, human-computer interaction and the like.

The invention relates to the technical field of distributed resource management and optimization, in particular to a distributed resource life cycle evaluation and dynamic updating method, which comprises the following steps: data acquisition: collecting multi-modal data of distributed resources, including resource state data, use behavior data, environment data and historical maintenance data; feature extraction: carrying out feature engineering processing on the multi-modal data, and extracting key indexes such as a resource health state, a degradation rate and environmental sensitivity; building a life cycle evaluation model, building the model by using the time sequence data of the resources based on a multi-task learning framework, and predicting the health state and the residual life of the resources; and optimizing a dynamic updating strategy, and monitoring and dynamically adjusting in real time. The method provided by the invention overcomes the problems of insufficient life cycle dynamics and rigid updating strategy in a traditional method, has the characteristics of high efficiency, intelligence and accuracy, and is suitable for intelligent management and optimization of new energy equipment, industrial assets and other distributed resources.

The invention belongs to the technical field of medical image processing, and discloses a CT image segmentation and classification system based on segmentation feature guidance, and the specific technical scheme is as follows: the system adopts a shared encoder to extract general features, realizes collaborative optimization of segmentation and classification through a double decoding path, adopts a partial decoder in a segmentation path, and adopts a partial decoder in the segmentation path; in combination with a local feature attention module, through multi-scale feature fusion and a boundary perception mechanism, the region consistency of a global segmentation map is gradually optimized, edge detail information is supplemented, and a classification path generates a space attention weight through a segmentation feature guide module by utilizing segmentation prediction; the classification network is guided to focus on a focus area and suppress background interference, a self-adaptive loss weighting strategy based on multi-task learning is adopted, the double-task gradient flow is dynamically balanced, and the gradient competition problem in the multi-task learning is effectively relieved.

The invention belongs to the technical field of knowledge maps, and discloses a traditional Chinese medicine knowledge map fusion system based on semantic alignment. Comprising a multi-modal semantic embedding and entity alignment layer, a structure embedding and attribute fusion service presentation layer, an enhanced TransR model and knowledge graph construction layer, an intelligent rule engine and conflict resolution layer, an end-to-end automatic assembly line layer, a multi-task learning framework and incremental dynamic updating. Through innovative combination of a multi-level semantic alignment technology and an intelligent traditional Chinese medicine rule engine, end-to-end automatic fusion of traditional Chinese medicine knowledge maps is realized, and heterogeneous knowledge maps from a plurality of data sources are automatically converted into a standardized unified knowledge map through an assembly line. Compared with a traditional character string matching method, the accuracy rate of the method is about half, and the alignment accuracy rate is improved to 80-90%. More importantly, under the same data condition, the fusion time is greatly shortened, and the incremental updating efficiency is remarkably improved.

The invention provides an airfoil flow field prediction method based on multi-task learning, and the method comprises the steps: firstly building an airfoil flow field prediction model based on multi-task learning, which comprises an encoder, a backbone network and a multi-head decoder; then training the established model by using an airfoil flow field data set and adopting an optimization strategy based on multi-task learning; and finally, performing geometric parameterization and grid generation processing on an airfoil which actually needs flow field prediction to obtain standardized airfoil flow field data, and inputting the airfoil flow field data into the trained airfoil flow field prediction model for prediction to obtain full flow field physical quantity distribution and lift-drag coefficient information. According to the method, a multi-task loss optimization strategy is used in the prediction model training process, so that the problem of conflict between airfoil profile surface loss and volume loss optimization can be effectively solved, and then the construction of an airfoil profile proxy model and the accurate prediction of flow field information and aerodynamic parameters are realized.

The invention discloses a radar signal detection method and system based on multi-task learning of a time-frequency domain feature fusion network, and the method comprises the following steps: building a radar signal mathematical model, and obtaining various types of radar signal sequences through the mathematical model; processing the radar signal sequence to obtain a radar signal I sequence and a radar signal Q sequence; marking the radar signal I sequence and the radar signal Q sequence, and constructing a training data set; constructing a time-frequency domain feature fusion network, and training the time-frequency domain feature fusion network by using the training data set to obtain a signal detection and recognition model; and obtaining a to-be-detected radar signal, and completing radar signal detection by using the signal detection identification model. According to the method, combined modeling analysis is carried out on various radar signals, local features and global features of the signals are extracted by adopting a time-frequency domain feature fusion network from the angle of signal I / Q, and the influence of noise on signal identification is reduced by fusing the features of the frequency domain.

The invention relates to the technical field of phishing detection, in particular to a phishing detection method and system based on multi-modal fusion features, and the method comprises the following steps: S1, constructing a phishing sample based on multi-modal simulation and data generation; according to the method, the multi-modal simulation and data generation technology is utilized, phishing samples with high diversity and authenticity are created, a solid foundation is provided for the model, diversified interactive behavior data are generated through the user behavior simulator designed by the reinforcement learning model, the system can simulate and understand behaviors of different types of users, and the user experience is improved. The system improves the detection precision and efficiency through a non-traditionally designed multi-mode fusion deep detection model in combination with a multi-branch neural network and multi-task learning, realizes the real-time adaptive detection and defense capability through adversarial training and federated learning, greatly improves the accuracy and robustness of phishing detection, and improves the detection accuracy and efficiency. And comprehensive protection is provided for network security.

The invention discloses an air pollution risk prediction and early warning method based on multi-task learning, and the method comprises the following steps: S1, collecting multi-source air pollution monitoring data, and carrying out the preprocessing of the data, and generating preprocessing data; s2, constructing a multi-task learning model, and generating a preliminary prediction result through the multi-task learning model based on an improved multi-head attention mechanism; s3, optimizing parameters of a shared feature extraction layer and a task specific prediction layer; s4, optimizing the preliminary prediction result, and generating an optimized prediction result; s5, generating a risk level, and generating a risk assessment report by integrating the risk information of time and space dimensions; and S6, designing a trigger condition according to the risk level, defining a joint early warning rule, judging and triggering early warning through a dynamic threshold value, and generating early warning information. According to the method, multi-task learning and an improved multi-head attention mechanism are combined, multi-pollutant concentration prediction and risk early warning are realized, and the method has the advantages of high timeliness, high prediction precision and comprehensive risk assessment.

The invention relates to an intelligent computing resource allocation method based on reinforcement learning. According to the technical scheme, the method comprises the steps that monitoring data from different sources including hardware monitoring, software logs and network bandwidths are fused, and a multi-dimensional time sequence state space is formed; carrying out dimensionality reduction and de-noising processing on the multi-modal data by using a depth auto-encoder; multi-task learning MTL is introduced into time sequence modeling, and resource requirements and state evolution of multiple tasks are predicted at the same time; generating a plurality of predictive resource scheduling strategies by using a GAN (Generative Adversarial Network), and dynamically selecting a strategy scheme when a load changes; each strategy is realized by an independent sub-network, and part of core knowledge is shared; through a strategy evolution mechanism, according to a historical feedback optimization strategy combination including task completion time and resource consumption, in a heterogeneous resource environment including a plurality of cloud computing platforms and edge computing nodes, based on difference of resource types, fine-grained scheduling of strategies is carried out; and scheduling the decision by using a distributed Q-learning mechanism in the reinforcement learning model.

The invention provides a personalized fitness system and method based on multi-modal data and an adaptive large model. According to the method, multi-modal data are integrated, and heterogeneous data collaborative analysis is realized by using a space-time alignment algorithm and a confidence coefficient weighting mechanism. A multi-task learning framework is adopted, physiological prediction, exercise risk early warning and psychological incentive strategy generation tasks are synchronously processed, a user stage target is adapted through a dynamic attention mechanism, and federal learning and transfer learning technologies are combined. A nonlinear periodic planning engine based on reinforcement learning dynamically adjusts training load and action difficulty according to real-time biomechanical simulation results and recovery state evaluation, and meanwhile, a lightweight model compression technology and end-side reverse dynamics calculation are adopted. Based on the scheme, the exercise performance prediction precision and the early warning timeliness are improved, and the exercise loss risk caused by overtraining is reduced; the cold start data bottleneck is broken through; low-delay virtual coach interaction and privacy protection are realized; and dynamically adjusting the training intensity and the incentive strategy.

The invention relates to the technical field of robot training. The robot self-adaptive training method based on reinforcement learning comprises the steps that task sub-target information is generated through a high-level strategy network, the task sub-target information is input into a low-level execution network, an action control instruction is generated according to the task sub-target information, interaction feedback information is collected in the execution process, and the action control instruction is sent to a robot through a robot. Calculating a reward value according to the interaction feedback information, carrying out association processing on the reward value and the scene complexity parameter, executing a dynamic reward shaping operation, generating an adjusted reward signal, generating a strategy model optimized by meta-learning based on the adjusted reward signal, loading the strategy model in a simulation environment, and carrying out dynamic reward shaping. A target strategy model optimized through simulation training is generated, the target strategy model is loaded to the robot, and the robot is controlled to execute task operation in the actual interaction scene. The method has the effect of realizing adaptive task learning of the robot in a multi-interaction scene.

The invention belongs to the field of voice noise reduction processing, and discloses an intelligent communication equipment voice noise reduction method and system based on artificial intelligence. Comprising intelligent communication equipment, and a microphone array is deployed on the intelligent communication equipment to record real-time noisy voice data of a user; inputting the recorded noisy voice data into the edge layer, and obtaining the voice data after noise reduction through real-time processing; the method comprises the following steps: pre-constructing a dialect voice database, adding a scene evolution layer, a voice change layer and a noise mixing layer in a conditional adversarial network cGAN to expand the dialect voice database, forming a data sample, and performing three-dimensional marking on each piece of voice data in the data sample; an input, output and multi-task learning mechanism is designed based on an end-to-end voice enhancement model, pre-training is carried out by using a data sample, and a shared encoder is arranged in the multi-task learning mechanism and is used for combining a task R1, a task R2 and a task R3 to realize voice noise reduction of intelligent communication equipment.

This specification describes systems and methods, implemented as computer programs on one or more computers in one or more locations, for controlling an agent to perform multiple different tasks in an environment. The described techniques partition the architecture of a controller into a dispatcher that understands the environment and an executor that understands how to control the agent, with a control channel between them that structures the partitioning. This allows implementations of the controller to generalize better.

The invention relates to a multi-element load combined prediction method fusing a decomposition reconstruction technology and multi-task learning, and the method comprises the steps: firstly obtaining a historical load sequence of an electric load, a cold load and a heat load and a data sequence of influence factor characteristics, and forming a historical multi-element load sequence library and an influence factor characteristic library; performing preliminary decomposition on data in the historical multivariate load sequence library through a variational mode decomposition method, performing reconstruction by using a mixed characteristic evaluation method, and updating the historical multivariate load sequence library; the historical multivariate load sequence library and the influence factor feature library are remodeled into three-dimensional feature tensors, and a CNN-ECA-MMoE multi-task learning model is constructed and used for outputting feature sharing and feature extraction results; and constructing a CNN-ECA-MMoE-LSTM multi-task learning model, carrying out model training by using the three-dimensional feature tensor, and outputting prediction results of an electric load, a cold load and a thermal load through the trained model. According to the method, the prediction precision of the multi-element load is improved.

The invention belongs to the technical field of multi-modal information processing, and particularly relates to a multi-modal sentiment analysis method based on hierarchical feature fusion and multi-task learning, which comprises the following steps: S1, establishing a multi-modal sentiment analysis model framework; s2, preparing a data set and preprocessing multi-modal audio-visual data; s3, inputting the single-mode data into a single-mode hierarchical feature extraction fusion network (UFEN), and obtaining a single-mode multilevel space-time attention fusion feature and an emotion analysis subtask result; s4, inputting different single-mode emotion representations into a multi-mode cross fusion network (MCFN) to obtain complementary features of different-mode emotion information; s5, inputting the emotion information complementary features into a feature reconstruction network and a prediction layer for emotion prediction; and S6, performing multi-task learning in combination with a plurality of sub-task results, and constructing a multi-task dynamic loss function training model. The spatial-temporal features of the multi-modal data are effectively extracted, multi-modal complementary feature interaction is achieved, and the accuracy and generalization of multi-modal sentiment analysis are improved.

The invention discloses a small sample learning method and system based on data enhancement. Collecting electric power operation and maintenance text data from a plurality of data sources, preprocessing and labeling the data, and constructing an electric power field corpus; performing fine tuning on a predefined pre-training language model by using the power field corpus, training power field word vectors and entity vectors from the power field corpus, and fusing the power field word vectors and entity vectors into the pre-training language model; performing named entity recognition and multi-task learning of power operation and maintenance knowledge based on a pre-training language model, constructing a multi-task adversarial learning model in combination with an adversarial learning algorithm, training the multi-task adversarial learning model by adopting multiple data enhancement strategies, and performing data prediction through integrated learning and self-training by utilizing the multi-task adversarial learning model; and evaluating a model prediction result, and dynamically adjusting the weight of each data enhancement strategy based on an evaluation result. The small sample learning performance is improved.

The invention relates to the field of data processing, and particularly provides a big data mining method and system based on digital enterprise management, and the method comprises the steps: obtaining multi-modal session data generated in the process of carrying out digital service interaction between a target enterprise and a user, extracting a user demand feature set from the multi-modal session data, and based on a preset multi-task learning framework, carrying out joint training on the user demand feature set, generating a demand analysis model, calling the demand analysis model to analyze the user demand feature set, generating an optimization strategy set associated with the target enterprise service process, and sending the optimization strategy set to the server. And updating the digital service execution logic of the target enterprise according to the optimization strategy set, and performing incremental training on the demand analysis model based on user feedback data. According to the invention, the response efficiency and the function adaptability of the digital service process can be improved.

The invention relates to the technical field of computer vision, and particularly discloses an adaptive pyramid and multi-stage path aggregation-based lightweight multi-task small target detection algorithm, which comprises an adaptive feature pyramid network, a multi-stage path aggregation module, a lightweight Transform module, an optimized channel attention mechanism, a multi-task learning head and an algorithm training method. According to the lightweight multi-task small target detection algorithm based on adaptive pyramid and multi-stage path aggregation, dynamic weight adjustment of different levels of features is realized by combining an adaptive FPN, and the expression ability of multi-scale features is enhanced. The MPAM module introduces a lightweight Transform module on the original basis, global feature modeling is performed by using an axial attention mechanism, the perception ability of the model to a small target is improved, and a channel attention mechanism is optimized to reduce calculation overhead.

The invention discloses a single air conditioner load regulation potential evaluation method based on transfer learning and multi-task learning, and relates to the technical field of air conditioner management.The method comprises the steps that a regional association model is built based on air conditioner operation data in a general scene, and the regional association model is analyzed to obtain a cooperative regulation relation between air conditioner equipment; constructing an air conditioner load regulation potential evaluation model by using a deep residual neural network; a multi-task learning framework is formulated according to the cooperative adjustment relation between the air conditioning equipment, and the air conditioning load adjustment potential evaluation model in the general scene is optimized by using the multi-task learning framework; and adjusting the weight of the optimized air conditioner load regulation potential evaluation model in combination with a transfer learning technology to obtain an air conditioner load regulation potential evaluation model in a specific scene. According to the method, the adaptability and generalization ability of the air conditioner load regulation potential evaluation model are remarkably improved by constructing the regional correlation model and introducing self-supervised learning and adversarial training.

The invention discloses a power equipment fault prediction system based on big data analysis. The method comprises the following steps: acquiring initial equipment multi-dimensional data; constructing a dynamic topology network of the power equipment, including a dependency relationship between the equipment and a fault propagation path, performing embedded learning on the dynamic topology network by using a GNN graph neural network, and extracting equipment collaboration features in the initial equipment multi-dimensional data; a multi-task learning framework is constructed in combination with the equipment cooperation features to predict the equipment fault probability and the remaining service life, and an equipment health index is obtained; and acquiring environmental parameters, dynamically adjusting a fault judgment threshold based on the equipment health index and the environmental parameters, generating a prediction result, integrating the prediction result with an SCADA system, and triggering graded early warning. And the influence of environmental factors on the operation state of the equipment is fully considered. Under different environmental conditions, the equipment fault risk can be judged timely and accurately.

The invention discloses a machine generated text detection method, a terminal, a medium and a program product in the field of natural language processing, and the method comprises the steps: inputting a to-be-detected target text into a trained text detection model, and obtaining a detection result outputted by the text detection model; the text detection model comprises a first semantic feature extraction network, a second semantic feature extraction network and a fusion classification network; the first semantic feature extraction network is used for extracting semantic features of keywords in a target text to generate a first representation vector; the second semantic feature extraction network is used for extracting overall semantic features of the target text and generating a second representation vector; the fusion classification network processes the first representation vector and the second representation vector to generate a detection result about whether the target text is a machine-generated text; the training of the text detection model adopts a multi-task learning strategy. According to the method, multi-task training is introduced, so that the accuracy and robustness of machine generated text detection are effectively improved.

The invention relates to the technical field of natural language processing, and discloses a natural language processing system based on Transform, comprising: a neural architecture search assembly for automatically searching and optimizing a Transform model architecture for natural language processing; the multi-task learning component is used for simultaneously training a plurality of natural language processing tasks on the basis of the shared model layer; the meta-learning component is used for optimizing parameter initialization of the model, so that the model can quickly adapt to a new task through a small amount of data and training steps; and the dynamic data enhancement component is used for generating diversified training data and enhancing the generalization ability and robustness of the model. According to the method, knowledge sharing and resource efficiency are enhanced through the multi-task learning and meta-learning component, dependence on a large amount of data is reduced, the generalization ability and rapid adaptability of the model are improved, the adaptability to a new scene is enhanced through the dynamic data enhancement component, and the robustness of the model is improved.

The invention relates to the field of big data analysis, in particular to an intelligent recommendation system based on user behaviors. The method comprises the following steps: acquiring user explicit behavior data and user implicit behavior data in a system, acquiring interaction depth index data of a user in the system, adding noise to sensitive behaviors in the data by using a differential privacy technology, establishing a variable time attenuation function, segmenting a behavior sequence according to sessions by using a hierarchical Transform encoder, and obtaining a variable time attenuation function; establishing a GNN user intelligent recommendation model based on a heterogeneous graph neural network, connecting the multi-task learning framework with the heterogeneous graph neural network to obtain a target GNN user intelligent recommendation model, optimizing a recommendation result of the model by using a multi-target optimization algorithm, inputting feature multi-granularity behavior data into the user intelligent recommendation model for recognition, and performing recommendation on the target GNN user intelligent recommendation model. And obtaining a user recommendation result. The problem of recommendation homogenization caused by a single target can be avoided, and intelligent recommendation efficiency and recommendation accuracy are improved.

The invention relates to the technical field of language learning, and discloses an English auxiliary pronunciation training method and system based on multi-task learning, and the method comprises the following steps: constructing a multi-task learning framework which comprises at least two subtask models related to pronunciation, comprising the steps of pronunciation accuracy scoring, pronunciation error type positioning and intonation fluency analysis. A voice signal, a lip movement video and a text content of a user are collected through a multi-modal data input module; and performing joint coding on the multi-modal data by using the shared feature extraction layer to generate fused acoustics. According to the English auxiliary pronunciation training method and system based on multi-task learning, the pronunciation training path of the user can be adaptively adjusted, the requirements of different users can be met, the English pronunciation level can be more effectively improved, and through strategies such as multi-modal data input, shared feature extraction, inter-task attention mechanism and adaptive training path adjustment, the English pronunciation training efficiency can be improved. Accurate evaluation and personalized training of English pronunciation are realized.

The invention provides a city calculation basic model and equipment based on a Mama time sequence, and relates to the technical field of data processing. In the model provided by the embodiment of the invention, a Mama framework is combined with multi-task time sequence analysis. Aiming at the non-stationarity of urban time series data, a frequency prompt network is designed for decoupling in combination with a Koopman operator theory and dynamic mode decomposition, a steady mode and non-stationary fluctuation are analyzed from the angles of time and frequency, and a frequency mode memory pool (FPMP) is introduced for realizing cross-scene feature reuse. The advantage of a Mama model in capturing long-term time dependence is utilized, and an attention mechanism similar to Mama is designed to simulate channel dependence in a multivariable time sequence. A learnable supplementary sequence and a multi-task prompt mark are introduced to support dynamic multi-task adaptation, and the contradiction between urban data isomerism and task diversity is relieved. Shared knowledge and specific task knowledge are balanced through shared experts and specific task experts of the multi-task attention module, the negative migration problem in multi-task learning is relieved, and the generalization ability of the model is enhanced.

The invention discloses a coordinated optimization multi-step prediction method based on target correlation space-time coding-task decoding, and the method comprises the steps: constructing a space-time encoder formed by stacking a plurality of space attention enhanced time sequence convolution modules, and carrying out the complete adaptive noise set empirical mode decomposition of a target value historical sequence; constructing a parameter sharing mechanism, and training a multi-step prediction model formed by a space-time encoder and a task specific decoder by using a multi-task learning normal form; training input is preprocessed historical industrial process production data, and gradient conflicts among multiple tasks are reduced by using a gradient adjustment mechanism and a dynamic weight distribution method; and preprocessing real-time industrial process production data, and inputting the data into the trained prediction model to obtain a prediction result. According to the method, the relationship among different prediction tasks can be effectively coordinated, the overall model performance is optimized, and challenges encountered in long-term prediction are more effectively processed by dynamically adjusting weight distribution and gradient flow among multiple tasks.