224 results about "Hyper parameters" patented technology

The invention provides a sequential network traffic prediction method and system based on swarm intelligence parameter optimization, and relates to the technical field of network traffic prediction. The method comprises the following steps: acquiring indexes such as throughput packet loss rate and round-trip delay of a target link by using a network probe, and performing deletion filling normalization and multi-scale decomposition to obtain a standardized traffic sequence; calculating information entropy, constructing a traffic complexity feature vector, and dividing a training set and a verification set; constructing a hybrid depth prediction model composed of a one-dimensional convolutional network and a gating cycle unit, and establishing a hyper-parameter search space; using particle swarm optimization and entropy-driven inertia weight adjustment and mutation probability mapping to reconstruct a speed and position updating strategy, and iteratively outputting a global optimal hyper-parameter; and generating a benchmark prediction result according to full-amount training, extracting a residual error, training a nonlinear residual error compensation model to carry out superposition correction and reverse normalization, obtaining a final flow prediction result, and improving prediction precision and generalization ability.

The invention relates to the crossing field of hydraulic engineering hydrological monitoring technology and machine learning prediction technology, discloses a sediment concentration prediction method based on deep learning, and aims to solve the problems that in existing sediment concentration prediction, hyper-parameter manual tuning is low in efficiency, key feature attention is insufficient, local and time sequence information is difficult to consider by a single model and the like. Accurate prediction is realized through five core modules: a data preprocessing module performs missing value filling, abnormal value processing and derivative feature generation on hydrological data; the feature selection module screens key features based on mutual information; the time sequence construction module generates time sequence data through a sliding window; the hyper-parameter automatic optimization module adopts Bayesian optimization iteration to obtain an optimal hyper-parameter; the CNN-LSTM-attention prediction module fuses CNN local feature extraction, bidirectional LSTM time sequence dependence capture and multi-head self-attention mechanism key feature focusing capability, is suitable for scenes such as river channels and channels, and provides efficient decision support for hydrological regulation and control.

The invention relates to the technical field of tunnel engineering and intelligent construction, in particular to a shield tunneling real-time control method based on a random forest and a particle swarm optimization algorithm. The method comprises the steps that initial tunneling parameters are generated through a parameter recommendation random forest model according to geology and tunnel geometric parameters, and model hyper-parameters are optimized through a sparrow optimization algorithm; carrying out settlement prediction by utilizing the settlement prediction random forest model; if the predicted value exceeds the limit, carrying out iterative optimization by adopting a particle swarm optimization algorithm and taking the initial parameter as a starting point, and searching a global optimal tunneling parameter combination meeting the settlement requirement; finally, the optimized parameters are issued to the shield tunneling machine to be executed, the model is continuously updated based on real-time construction data, and closed-loop control is formed. According to the method, intelligent recommendation and real-time optimization of tunneling parameters can be realized, the ground surface settlement control precision and the system response speed are improved, the dependence on artificial experience is effectively reduced, and the self-adaptive capability and the intelligent level of shield construction under complex geological conditions are enhanced.

The invention provides a near-infrared model cross-device application method based on multi-source fusion optimization, and aims to solve the problem that various neural network models constructed by a high-precision near-infrared spectrometer cannot be applied to portable devices when being migrated to the portable devices. And the model performance is reduced due to differences of an optical system structure, a wavelength range, a signal-to-noise ratio, an instrument response characteristic and the like. According to the method, the near infrared spectrum data and the conventional quality parameters are subjected to feature level fusion, and the cross-equipment spectrum deviation is calibrated and compensated by using the stability of the conventional quality parameters, so that efficient knowledge migration from high-precision equipment to portable equipment is realized in the fused feature space. The problem of cross-device feature dimension mismatching is solved through a stable feature importance evaluation-based wave band alignment method; a small sample optimization strategy and a hyper-parameter automatic search technology are combined, so that the transfer learning effect is improved, the performance of a portable equipment detection model is remarkably enhanced, and meanwhile, the modeling cost is greatly reduced.

The invention discloses a multi-modal data drawing logical relationship analysis method, electronic equipment and a medium, and the method comprises the steps: generating a node set based on drawing image data and text data; generating a cross-modal hyperedge set based on the spatial proximity relationship, the visual feature similarity and the semantic correlation between the node sets; generating a hypergraph embedding input representation based on the node set and the cross-modal hyperedge set; the hypergraph is embedded into the input representation input improved hypergraph self-attention network model, and a hyperedge logic relation type and a corresponding hyperedge confidence coefficient are generated; generating a graph structure result based on the hyperedge logic relationship type and the node set, wherein the graph structure result meets the structure legality requirement; and performing hyper-parameter automatic adjustment and convergence control on the atlas structure result based on hyper-edge confidence, and generating an optimal atlas analysis model and a structured output result. According to the method, the reliability and the quality of analysis of component nodes, logic edge relationships and semantic structures in the drawing are improved.

The invention relates to the technical field of intelligent retrieval, and discloses an intelligent dynamic K value retrieval optimization system and method, and the system comprises a vector retrieval and storage module, a dynamic K value generation module and an adversarial training optimization module. The system introduces a problem multi-dimensional feature library, wherein multi-dimension of the problem comprises a domain label, a complexity level and a candidate result cardinal number of the problem; the system adjusts the top-k value during retrieval through the dynamic K value generation module, and optimizes the performance of the generator in combination with the adversarial training optimization module. According to the intelligent dynamic K value retrieval optimization system and method, multiple technologies of vector database storage, embedded vector caching, GAN model training, dynamic temperature adjustment, multi-index fusion optimization, incremental updating and hyper-parameter automatic tuning are integrated. The method is characterized in that a k value selection problem is converted into a learnable generation task, a generator can predict the number of optimal retrieval results for a specific problem through adversarial training, and the limitation that the k value is fixed in a traditional retrieval system is broken through.

The invention provides an intelligent power distribution room sensor fault early warning method and system based on multi-modal data fusion. The method comprises the following steps: step 1, realizing working condition adaptive acquisition of multi-modal data; step 2, eliminating an alignment scheme of physical contradictions; step 3, noise analysis and filtering parameter adaptive cooperation are realized; 4, performing primary decomposition on the denoised signal, and analyzing a non-stationary signal; step 5, realizing security sharing and weighted aggregation of cross-node features; step 6, constructing a multi-modal sensor fault diagnosis model based on the time consistency fracture field; and step 7, optimizing hyper-parameters of the multi-modal sensor fault diagnosis model based on the time consistency fracture field. According to the invention, early warning of faults is realized, fault traceability and propagation path analysis capability are realized, deep support is provided for operation and maintenance decision, and the overall operation reliability and intelligent operation and maintenance level of the power distribution room are effectively improved.

The invention discloses a two-stage algorithm selection and hyper-parameter joint optimization method, which comprises the following steps of: in the first stage, processing a training set and a test set through row sampling operation and column dimension reduction operation to form a reduced data set; randomly sampling a certain number of configurations in the hyper-parameter space of each candidate algorithm, evaluating the performance of each candidate algorithm by using the reduced data set, and extracting an optimal performance score; in the second stage, a previous algorithm is screened according to the optimal performance score to form a candidate set, and a pruned hyper-parameter search space is formed so as to reduce the calculation complexity of processor hyper-parameter search; and performing hyper-parameter optimization on the pruned hyper-parameter search space by using the original data set, and outputting an optimal algorithm adaptive to the target technical task and hyper-parameter configuration thereof. Algorithm screening and hyper-parameter tuning adaptive to a specific scene are realized through a two-stage optimization strategy, and meanwhile, the method is suitable for a traditional table type dichotomy task and aims at improving the deployment efficiency and performance of a machine learning model.

The invention relates to the technical field of air conditioner load prediction, and provides an air conditioner load prediction method and system, and the method comprises the steps: extracting intra-day meteorological data features and intra-day air conditioner load data features, and forming multi-dimensional data features; performing dimension reduction processing on the multi-dimensional data features to form a comprehensive feature curve; clustering the comprehensive characteristic curve, dividing the air conditioner load data in the historical day into a data set according to a clustering result, and dividing the data set into a training set and a test set; a plurality of prediction models corresponding to different meteorological scenes are constructed, the prediction models are trained through the corresponding training sets, hyper-parameter tuning is conducted on the prediction models through an improved sodat swarm optimization algorithm, and a plurality of air conditioner load prediction models are formed; and inputting the test set corresponding to various meteorological scenes into the corresponding air conditioner load prediction model, and outputting an air conditioner load prediction result. According to the invention, air conditioner load curves with obvious boundaries in different meteorological scenes can be effectively separated, and air conditioner power load prediction errors in extreme weather are reduced.

The invention provides a cutting force prediction method and device based on a multi-target frost ice algorithm optimization model, and relates to the technical field of mechanical cutting machining. The method comprises the following steps: acquiring sensor signal data including cutting force data and vibration signals, extracting target features which are highly related to the cutting force from the vibration signals, constructing target feature vectors, constructing a cutting force prediction model based on SVR, and determining an SVR hyper-parameter range through initialization; the target feature vector is used as an input sample, the cutting force data is used as an output target value, the SVR hyper-parameter range is used as a constraint condition, a multi-target frost ice algorithm and a multi-target optimization mechanism are combined to optimize the SVR hyper-parameter, finally, the SVR is trained based on the optimized hyper-parameter to obtain a prediction model, and the cutting force of the numerical control machine tool is predicted. Efficient and high-precision cutting force prediction can be achieved without a large number of data samples, and the model has good generalization and feature interpretability.

The invention relates to the technical field of computers, and discloses a subject classification model construction method and system, electronic equipment and a storage medium, and the method comprises the steps: obtaining academic paper bibliography data based on an academic database, and constructing an initial training data set; identifying minority category subjects of which the sample quantity is lower than a preset threshold value, generating synthetic data containing chapters and keywords, and labeling corresponding subject categories; mixing the synthetic data with real data in the initial training data set, and constructing a balanced mixed training data set; splicing a text based on the chapter and the keyword of each sample in the mixed training data set, and generating a multi-level fusion feature; and taking the multi-level fusion features as input, accessing a full-connection classification layer to construct a model, carrying out end-to-end training based on a mixed training data set, and adjusting and optimizing model hyper-parameters to obtain a final subject classification model. According to the method, data imbalance can be effectively relieved, existing labeling resources are fully utilized, and the model generalization ability is improved.

The invention relates to a neural network model slope physical parameter inversion method based on data and physics hybrid driving. The problems that in the prior art, a slope parameter inversion method is high in dependence on monitoring data, large in calculation amount, low in efficiency and difficult to deal with complex geological conditions and sparse observation are solved. The method comprises the following steps: S1, selecting a slope and constructing geometric and physical models of the slope; s2, constructing a neural network model loss function according to the geometric and physical models of the slope; s3, defining a neural network model, input and output characteristics of the neural network model and model hyper-parameters; s4, monitoring data preprocessing and sample division; s5, converting the unbounded variable p of the to-be-inverted parameter into a bounded parameter value through a Sigmoid mapping function; S6, designing adaptive sampling of a neural network model; and S7, deploying the model and constructing an inversion-early warning closed-loop system. The method has the advantages that the training time and cost are reduced, and the prediction efficiency and precision are improved; and the engineering availability and stability are improved.

The invention discloses a crop ralstonia solanacearum disease prediction method based on an integrated learning model, and the method comprises the following steps: S1, collecting a published 16s rRNA gene sequence related to solanaceae crop bacterial wilt, and carrying out the preprocessing of original sequencing data based on an EasyAmplicon standardized process; s2, performing data dimension reduction by using a principal component analysis algorithm, and retaining 95% of variance; s3, performing hyper-parameter search based on 5-fold cross validation and grid search on the Light GBM model, the CatBoost model and the XGBoost model respectively, and selecting three groups of optimal hyper-parameters of each model; s4, constructing a model according to three groups of optimal hyper-parameters of each selected model, performing prediction, analyzing model result difference based on a Pearson correlation coefficient, and retaining Pearson correlation coefficient mean < lt > with other eight model prediction values; a model of 0.8; and S5, inputting the screened model prediction result into a second-layer element learner RF for integrated learning to obtain a final prediction result.

The invention discloses a multi-modal depth sensing and grabbing system based on a transparent object, which relates to the field of robot operation and comprises a multispectral sensing module, a depth correction module, a grabbing posture generation module and a control module. The visual information and the thermal radiation information of a transparent object are comprehensively obtained by combining two perception modes of an RGB-D image and a thermal imaging (TIR) image, systematic error analysis is performed on a depth map, and error sources of an RGB-D camera and a TIR camera are detected. The system adopts an encoder-decoder model as a depth correction core, the model extracts complementary features of an RGB-D image and a TIR image through a modal exclusive encoder, feature alignment and integration are completed by using a feature fusion module, and the depth estimation precision on the transparent surface is effectively improved. And meanwhile, a Bayesian optimization method is adopted to carry out hyper-parameter optimization on the depth correction model, through hyper-parameter optimization and model fitting processing, the system effectively avoids the problems of over-fitting and under-fitting, and the robustness of the depth correction model and the transparent object grabbing precision are further improved.

The invention discloses a ship engine fault diagnosis method based on IWOA-CNN-Transform, and the method comprises the steps: collecting the operation data of a ship engine in a normal state and a fault state, carrying out the preprocessing, constructing a data set, and dividing the data set into a training set and a verification set; the whale optimization algorithm is improved by introducing optimal neighborhood disturbance, adaptive weight and a variable spiral position updating strategy, and the improved whale optimization algorithm is constructed; constructing a CNN-Transform model, and optimizing hyper-parameters of the model by using an improved whale optimization algorithm to obtain an optimal hyper-parameter combination; and the CNN-Transform model is improved based on the optimal hyper-parameter combination, the training set is used for training, the verification set is used for evaluation, and prediction and diagnosis of the improved CNN-Transform model on engine fault classification are realized. According to the method, the accuracy and robustness of ship engine fault diagnosis can be effectively improved, and intelligent prediction and diagnosis of fault categories are realized.

The invention relates to the technical field of coal processing and utilization, and discloses a clean coal yield prediction method based on a support vector machine, which comprises a data acquisition module used for analyzing factors influencing the clean coal yield, acquiring related data and integrating the data into a data set, and a data preprocessing module connected with the data acquisition module and used for preprocessing the data. The data preprocessing module is used for randomly dividing a data set according to a 70% training set and a 30% test set and carrying out standardization and normalization preprocessing, the parameter optimization module is connected with the data preprocessing module and optimizes hyper-parameters of a support vector machine through an improved grey wolf algorithm, and the model building module is connected with the parameter optimization module and is used for building a model. The model building module is used for building and training a support vector regression model based on the optimized hyper-parameters, and the model verification module is connected with the model building module and uses a test set to verify the performance of the model. According to the method, the hyper-parameters of the support vector machine are optimized through the improved grey wolf algorithm, the problem that a traditional optimization method is prone to falling into local optimum is effectively avoided, and the precision of clean coal yield prediction is remarkably improved.

The invention discloses a thin-wall shell plastic part injection molding process parameter multi-objective optimization method, which is based on an RIME-RF-MOGWO framework, takes a simulation sample as a research object, selects a volume shrinkage rate and a buckling deformation amount as optimization objectives, and firstly adopts SMOTE to process a data imbalance problem; establishing a nonlinear mapping relation between the process parameters and the quality target by using RF; an RIME is introduced to optimize the hyper-parameter of the RF; multi-objective optimization of process parameters is realized in combination with MOGWO, and a Pareto frontier solution set is obtained through non-dominated sorting and a congestion degree control mechanism. A multi-round optimization and simulation verification result shows that the multi-objective optimization method for the injection molding process parameters of the thin-wall shell plastic part can effectively obtain an optimal process parameter combination, the volume shrinkage rate is reduced by 19.02%, the buckling deformation amount is reduced by 50.63%, and the molding quality of the thin-wall plastic part can be remarkably improved.

The invention relates to the field of electric energy metering device fault identification, and discloses a communication loop internal topology defect fault diagnosis and early warning method based on multi-dimensional feature fusion. Comprising the following steps: acquiring original data of an electric energy meter sample with a communication fault caused by an internal topology defect and a normal operation electric energy meter sample, and cleaning to remove invalid samples to obtain intermediate data; constructing a sample set, and extracting a multi-dimensional feature vector for each sample; initializing an XGBoost classification model, optimizing hyper-parameters of the XGBoost classification model by adopting an improved sparrow search algorithm to obtain an optimal hyper-parameter combination, and training on the training set by utilizing the optimal hyper-parameter combination to obtain a final XGBoost fault diagnosis model; and deploying the final XGBoost fault diagnosis model to a monitoring system to obtain a fault risk score. According to the method, the accuracy and robustness of fault identification are remarkably improved, and misjudgment caused by fluctuation of a single feature is effectively avoided.

The invention relates to an iron-based high-temperature alloy design method based on multi-model machine learning and an alloy thereof, and the method comprises the following steps: 1, obtaining iron-based high-temperature alloy components, carrying out high-throughput calculation, and constructing a data set; 2, correlation analysis is carried out on the data set, feature importance sorting is carried out based on a random forest model, and key components are screened out; 3, constructing a machine learning algorithm model, training by using the key components, and adjusting and optimizing hyper-parameters of the machine learning algorithm model through an optimization algorithm to obtain a trained prediction model; 4, constraint conditions are constructed and optimized, an optimal solution set is obtained, the optimal solution set is input into the trained prediction model, and optimal alloy components are obtained; and 5, laser powder bed melting forming is conducted according to the optimal alloy components, and the iron-based high-temperature alloy is obtained. A data-driven machine learning method is adopted to replace a traditional trial and error method, the alloy research and development period is greatly shortened, and the research and development cost is reduced.

The invention discloses a power transmission line fault positioning method, device and equipment and a computer readable storage medium, and the method comprises the steps: collecting an electrical transient signal when a power transmission line breaks down, and carrying out the preprocessing of the electrical transient signal, and obtaining a standardized signal; performing multi-scale decomposition on the standardized signal, extracting energy features of each sub-band, and constructing a time-frequency-energy three-dimensional feature vector representing a fault; constructing a gating circulation unit network model, and performing global optimization on key hyper-parameters of the GRU network model by adopting a sparrow search algorithm SSA; and processing the time-frequency-energy three-dimensional feature vector by using a GRU network model subjected to SSA optimization so as to position the fault of the power transmission line. According to the method, the problems of low positioning precision and poor adaptability caused by noise interference, model mismatch and manual parameter adjustment limitation in a complex power grid environment can be solved, and millisecond-level, high-precision and high-robustness precise positioning of the power transmission line fault is realized.

The invention discloses an FISTA plug-and-play sparse radar imaging method based on adaptive regulation and control, which combines FISTA with deep learning prior, introduces reinforcement learning to carry out adaptive parameter regulation and control, and aims to significantly improve the automation degree, convergence rate and cross-scene robustness of an imaging process. The method is especially suitable for processing challenges of coexistence of different sampling rates, noise levels and scene complexity, and can effectively solve the problems that a traditional method depends on manual parameter adjustment, the calculation overhead is high, and the generalization ability is weak. According to the method, selection modeling of key hyper-parameters (such as step length, de-noising intensity and momentum coefficient) and a stop criterion is used as a Markov decision process, and an intelligent agent makes a decision adaptively according to a current reconstruction state, so that the problem is solved, and an advanced solution is provided for efficient and reliable radar imaging processing.

The invention provides a spinning machine fault detection method and system based on machine learning, and relates to the technical field of fault detection.The method comprises the steps that in the running process of a spinning machine, process parameter data and result parameter data are collected and preprocessed to obtain comprehensive processing data, and fault sensitive features are extracted from the comprehensive processing data; and constructing a support vector machine model, and optimizing hyper-parameters of the support vector machine model according to a Genethink shark optimization algorithm to obtain an enhanced vector machine model. And inputting the fault sensitive features into the reinforcement vector machine model, outputting a fault prediction probability, and judging whether a preset fault judgment threshold value is reached or not, and if so, positioning a fault source according to the current fault sensitive features. And formulating a maintenance scheme according to the fault source, and regularly maintaining the spinning machine according to the fault detection interval. According to the method, the fault prediction precision can be improved, the error and omission ratio is reduced, the fault detection timeliness is improved, and the burst loss is reduced.

The invention discloses an underwater robot battery temperature rise health state prediction method based on a PO-LSSVM hybrid algorithm, and relates to the field of underwater robot battery health state prediction, and the method comprises the steps: collecting the battery working condition data in the operation process of an underwater robot battery; performing adaptive optimization on the hyper-parameters of the LSSVM by adopting an improved parrot optimization algorithm to obtain optimized hyper-parameters, and training an LSSVM model based on a kernel function to obtain a PO-LSSVM hybrid prediction model; using a PO-LSSVM hybrid prediction model to obtain an initial prediction value of the battery health state; inputting the preliminary prediction value into a constructed temperature-pressure-salinity compensation system for secondary correction to obtain a battery health state accurate estimation value suitable for the underwater complex environment; the degradation rule of the battery performance in the temperature rise process can be effectively captured, high-precision migration prediction is achieved, and the generalization ability of the model under the unknown temperature working condition is remarkably improved.

The invention discloses a water nutritive salt remote sensing inversion method based on multi-source fusion and a machine learning model. According to the method, non-optical environment factors such as optical remote sensing features, meteorological features and land utilization data are subjected to unified formatting and structured processing, so that multi-source data are input into an inversion model in a unified feature vector form. A self-adaptive screening strategy based on feature importance is adopted, the contribution degree of each feature can be dynamically calculated in the model training process, and variables with low contribution or large noise are automatically deleted. Temperature attenuation, a Metropolis criterion and disturbance generation are introduced, and a globally optimal model structure and hyper-parameter combination is automatically searched. Through joint design of data fusion, feature screening and an optimization algorithm, a stable, generalizable and easy-to-deploy nutritive salt inversion system is constructed, the problems that an existing method depends on a single spectrum, generalization is poor and parameter optimization is difficult can be effectively solved, and inversion of the concentration of the nutritive salt in the water body can be directly carried out.

The invention provides a multi-modal large-model low-resource modal adaptive learning method, and relates to the technical field of data processing, and the method comprises the steps: dividing a polygon in a reference topological unit into a convex region and a concave region based on a concavity and convexity identification result, and segmenting the reference topological unit to generate multi-scale topological partitions, mapping the low-resource modal data stream to a corresponding domain of the multi-scale topological partition and extracting a multi-resolution structural feature; calculating geometric feature parameters based on each polygon in the multi-scale topological partition, and matching the multi-resolution structural features with the high-resource modal data stream to generate a heterogeneous feature topology; performing optimization operation on the heterogeneous feature topology to obtain deeply optimized model parameter configuration; and verifying the deeply optimized model parameter configuration, and adjusting a model hyper-parameter set to obtain a low-resource modal adaptive multi-modal large model. According to the method, the feature extraction precision and the self-adaptive adaptation capability of the multi-mode large model to the low-resource mode are effectively improved.

The invention discloses an intelligent music recommendation method based on emotional perception and acoustic features, and relates to the technical field of intelligent recommendation systems and emotional computation.The method comprises the steps that physiological signals, music acoustic features and historical behavior data of a user are collected; preprocessing the multi-source features and mapping the multi-source features to the same dimension to construct a fusion matrix; building a double-branch deep learning model, fusing features through an attention mechanism and training parameters; an evolutionary algorithm is adopted to optimize hyper-parameter screening optimal combination; generating a recommendation list matched with the real-time emotion and preference; and continuously collecting user interaction data, and regularly and incrementally training the dynamic update model. According to the method, emotion and behavior dual-drive recommendation is achieved by fusing physiological signals and acoustic features, emotion perception is accurate, recommended content fits the real-time mood, model optimization is efficient, recommendation precision and diversity are remarkably improved, and the music consumption experience of a user is greatly improved.

The invention relates to a mixed oil viscosity prediction method based on a residual physical guidance neural network, and relates to the technical field of oil and gas gathering and transportation. The method comprises the following steps: acquiring crude oil samples mixed in different proportions, acquiring a component characteristic factor and viscosity high-correlation data set, and performing sample enhancement through a generative adversarial network (GAN) to generate synthetic data conforming to a physical law; processing time sequence data by adopting a gating cycle unit (GRU), and optimizing network hyper-parameters by utilizing an adaptive genetic algorithm (AGA); constructing a residual physical guidance neural network (PGNN) model, and combining physical model output with deep learning residual correction; and finally, dynamic prediction and visual output of the viscosity of the mixed oil are realized. The method effectively solves the problems that a traditional empirical formula is insufficient in precision, a pure data driving model is poor in generalization ability and inconsistent in physics, is suitable for real-time viscosity soft measurement and closed-loop control under the complex oil mixing working condition, and is beneficial to efficient transportation of mixed crude oil.

The invention provides a multi-branch sequence recommendation method based on dynamic channel fusion. The method aims at solving the problems that in sequence recommendation, data have a large amount of noise and are sparse, and existing model prediction is smooth. The method comprises the following steps: on one hand, enabling a user sequence to pass through an embedding layer, introducing position information by utilizing RoPE rotation position coding, and extracting long-term dependency of a user in the sequence through a multi-feature channel feature network to obtain a new embedded Eglobal, a short-term interest and long-term dependency interactive embedded Ecross, an embedded Eema after exponential smoothing filtering and an embedded Efreq after frequency domain modulation; and on the other hand, in order to fully fuse the characteristics of different channels, a traditional hyper-parameter fusion mode is abandoned, and a pooling network layered adaptive fusion mode is adopted. Firstly, four channel features are divided into two groups including a self-attention group and a trend group, features of different channels are learned through Squeeze average pooling, learned parameters are activated through Excitation by adopting sigmoid to obtain channel weights, and feature representations of different groups are sent into a gating network for weighted summation loss calculation.

The invention provides a small sample tunnel low-temperature asphalt optimization design method based on machine learning. According to the method, an asphalt sample is prepared through orthogonal test design, performance indexes are measured, and an initial small sample data set is established; the SMOTER technology is used for data enhancement, and the sample scale is effectively expanded; constructing an AR-CatBoost machine learning model, and introducing an adaptive regularization and gradient weighting mechanism to enhance the learning ability of the nonlinear relationship between the asphalt component mixing amount and the performance; carrying out hyper-parameter automatic optimization by adopting an improved reptile search algorithm fusing Cauchy variation and dynamic boundary adjustment; based on the optimization model, large-scale virtual ratio performance prediction is generated in a component mixing amount range, and the optimal ratio is accurately screened by integrating a scoring function. The method effectively solves the problem of insufficient model generalization ability under the small sample condition, realizes high-precision and low-cost automatic design of the tunnel low-temperature asphalt mix proportion, and is suitable for rapid research and development and performance optimization of tunnel asphalt materials in cold regions.

Systems and methods include a computer-implemented method for predicting hydraulic fracture initiation. A fracking operations dataset is prepared using historical field information for fracking wells. A set of hyper-parameters is tuned for use in a machine learning algorithm configured to predict fracture initiation for new fracturing wells. The dataset is divided into training and test datasets. A regression algorithm is applied to train the training dataset and to validate with the test dataset. A target variable of a breakdown pressure for a new hydraulic fracturing treatment is determined. A prediction dataset is updated using at least the target variable. The training dataset is trained using a classifier of the machine learning algorithm. A prediction is made using the prediction dataset whether the new hydraulic fracturing treatment can be initiated or not. The breakdown pressure is incrementally adjusted, and the method is repeated until successful hydraulic fracture initiation is predicted.