CN113779859B - Interpretable time sequence prediction model training method and device and computing equipment - Google Patents
Interpretable time sequence prediction model training method and device and computing equipment Download PDFInfo
- Publication number
- CN113779859B CN113779859B CN202110141485.6A CN202110141485A CN113779859B CN 113779859 B CN113779859 B CN 113779859B CN 202110141485 A CN202110141485 A CN 202110141485A CN 113779859 B CN113779859 B CN 113779859B
- Authority
- CN
- China
- Prior art keywords
- sequence diagram
- interpretable
- prediction
- sequence
- prediction model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention provides a training method, a device, a medium and a computing device of an interpretable time sequence prediction model. The method comprises the following steps: performing data processing on the acquired time sequence data to obtain a sequence diagram corresponding to the time sequence data; modeling the sequence diagram through the interpretable time sequence prediction model to obtain a prediction result corresponding to the sequence diagram; training the interpretable timing prediction model with a view to improving the prediction accuracy of the interpretable timing prediction model based on the true value and the prediction result. According to the invention, the multi-dimensional data in the sequence diagram and the time sequence data with the time sequence can be calculated through the interpretable time sequence prediction model, so that the prediction result is output, and the model can be trained based on the prediction result and the true value, so that the model can output a more accurate prediction result, and the prediction performance of the interpretable time sequence prediction model is also improved.
Description
Technical Field
The embodiment of the invention relates to the technical field of deep learning, in particular to an interpretable time sequence prediction model training method, an interpretable time sequence prediction model training device, an interpretable time sequence prediction model training medium and computing equipment.
Background
This section is intended to provide a background or context to the embodiments of the invention that are recited in the claims. The description herein is not admitted to be prior art by inclusion in this section.
The time-series data refers to data having a predefined time or order, and the time-series data can be widely applied to various application scenarios, such as classification, prediction, and complementation. How to accurately model the influence of various characteristics of time series data on a prediction result in various application scenarios is a significant problem. At present, the prediction of the multi-dimensional time series corresponding to the time series data is usually modeled by some classical explanatory method for modeling the importance of the features. Although the above-described methods are easily explained under corresponding model assumptions, these assumptions also greatly limit the predictive performance of the model.
Disclosure of Invention
In this context, embodiments of the present invention are directed to a method, apparatus, medium, and computing device for training an interpretable timing prediction model.
In a first aspect of the embodiments of the present invention, there is provided a training method of an interpretable time series prediction model, including:
performing data processing on the acquired time sequence data to obtain a sequence diagram corresponding to the time sequence data;
modeling the sequence diagram through the interpretable time sequence prediction model to obtain a prediction result corresponding to the sequence diagram;
training the interpretable timing prediction model with a view to improving the prediction accuracy of the interpretable timing prediction model based on the true value and the prediction result.
In an embodiment of the present invention, a specific way to perform data processing on the acquired time-series data to obtain a sequence diagram corresponding to the time-series data is as follows:
and processing the acquired time sequence data according to a time axis sliding window to obtain a sequence diagram corresponding to the time sequence data.
In an embodiment of the present invention, the interpretable time-series prediction model includes a significance module, and the modeling of the sequence diagram by the interpretable time-series prediction model to obtain a prediction result corresponding to the sequence diagram includes:
determining the sequence diagram as an original sequence diagram;
combining the disturbance area with the original sequence diagram through the saliency module to obtain a first disturbance sequence diagram;
calculating the original sequence diagram and the first disturbance sequence diagram through the significance module to obtain a second disturbance sequence diagram;
and modeling the second disturbance sequence diagram and the original sequence diagram to obtain a prediction result corresponding to the sequence diagram.
In an embodiment of the present invention, the interpretable time series prediction model further includes a deep learning module and a linear regression module, and the modeling of the second disturbance sequence diagram and the original sequence diagram to obtain the prediction result corresponding to the sequence diagram includes:
inputting the second disturbance sequence diagram into the deep learning module to obtain a first prediction result;
inputting the original sequence diagram into the linear regression module to obtain a second prediction result;
and combining the first prediction result and the second prediction result to obtain a prediction result corresponding to the sequence diagram.
In an embodiment of the present invention, training the interpretable timing prediction model based on the true value and the prediction result to improve the prediction accuracy of the interpretable timing prediction model includes:
calculating a real value and the prediction result to obtain a deviation between the real value and the prediction result;
a loss function based on an added norm such that the loss function constrains a complexity of the perturbation region;
training the interpretable timing prediction model based on the disturbance region and the bias with a view to improving a prediction accuracy of the interpretable timing prediction model.
In an embodiment of the present invention, after training the interpretable time series prediction model to improve the prediction accuracy of the interpretable time series prediction model based on the true value and the prediction result, the method further includes:
acquiring an arbitrary sequence diagram;
and updating the attention range of the disturbance region based on each feature in the sequence diagram to obtain a sequence saliency map corresponding to the sequence diagram.
In a second aspect of the embodiments of the present invention, there is provided an interpretable training apparatus for a time series prediction model, including:
the processing unit is used for carrying out data processing on the acquired time sequence data to obtain a sequence diagram corresponding to the time sequence data;
the modeling unit is used for modeling the sequence diagram through the interpretable time sequence prediction model to obtain a prediction result corresponding to the sequence diagram;
a training unit for training the interpretable timing sequence prediction model based on the true value and the prediction result so as to improve the prediction precision of the interpretable timing sequence prediction model.
In an embodiment of the present invention, a way for the processing unit to perform data processing on the acquired time-series data to obtain a sequence diagram corresponding to the time-series data is specifically as follows:
and processing the acquired time sequence data according to a time axis sliding window to obtain a sequence diagram corresponding to the time sequence data.
In one embodiment of this embodiment, the interpretable time-series prediction model includes a saliency module, and the modeling unit includes:
a determining subunit, configured to determine the sequence diagram as an original sequence diagram;
a combining subunit, configured to combine the disturbance region with the original sequence diagram through the saliency module to obtain a first disturbance sequence diagram;
the first calculation subunit is configured to calculate the original sequence diagram and the first disturbance sequence diagram through the saliency module to obtain a second disturbance sequence diagram;
and the modeling subunit is used for modeling the second disturbance sequence diagram and the original sequence diagram to obtain a prediction result corresponding to the sequence diagram.
In one embodiment of this embodiment, the interpretable time series prediction model further includes a deep learning module and a linear regression module, and the modeling subunit includes:
the modeling module is used for inputting the second disturbance sequence diagram into the deep learning module to obtain a first prediction result;
the modeling module is further used for inputting the original sequence diagram into the linear regression module to obtain a second prediction result;
and the combination module is used for combining the first prediction result and the second prediction result to obtain the prediction result corresponding to the sequence diagram.
In one embodiment of this embodiment, the training unit includes:
the second calculation subunit is used for calculating a real value and the prediction result to obtain a deviation between the real value and the prediction result;
a constraint subunit, configured to constrain the complexity of the disturbance region by using a loss function based on a norm;
a training subunit, configured to train the interpretable timing prediction model based on the perturbation region and the deviation so as to improve prediction accuracy of the interpretable timing prediction model.
In one embodiment of this embodiment, the apparatus further comprises:
an acquisition unit configured to acquire an arbitrary sequence diagram after the training unit trains the interpretable timing prediction model with a view to improving the prediction accuracy of the interpretable timing prediction model based on a true value and the prediction result;
and the updating unit is used for updating the attention range of the disturbance region based on each feature in the sequence diagram so as to obtain a sequence saliency map corresponding to the sequence diagram.
In a third aspect of embodiments of the present invention, there is provided a computer-readable storage medium storing a computer program enabling, when executed by a processor, the method of any one of the first aspect.
In a fourth aspect of embodiments of the present invention, there is provided a computing device comprising the storage medium of the third aspect.
According to the training method, the training device, the training medium and the computing equipment of the interpretable time sequence prediction model, time sequence data can be processed to obtain a sequence diagram, the sequence diagram can be modeled through the interpretable time sequence prediction model to obtain a prediction result, the interpretable time sequence prediction model can calculate multi-dimensional data and time sequence data with time sequence, the prediction result is output, the interpretable time sequence prediction model can be trained on the basis of the prediction result and a true value, the interpretable time sequence prediction model can output a more accurate prediction result through analyzing the multi-dimensional data and the data with the time sequence, and the prediction performance of the interpretable time sequence prediction model is also improved.
Drawings
The above and other objects, features and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description read in conjunction with the accompanying drawings. Several embodiments of the invention are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which:
FIG. 1 is a schematic flow chart illustrating a method for training an interpretable timing prediction model according to an embodiment of the invention;
FIG. 2 is a flowchart illustrating a method for training an interpretable timing prediction model according to another embodiment of the invention;
FIG. 3 is a flowchart illustrating a method for training an interpretable timing prediction model according to another embodiment of the invention;
FIG. 4 is a schematic structural diagram of a training method of an interpretable timing prediction model according to an embodiment of the present invention;
FIG. 5 schematically shows a schematic of the structure of a medium according to an embodiment of the invention;
fig. 6 schematically shows a structural diagram of a computing device according to an embodiment of the present invention.
In the drawings, the same or corresponding reference numerals indicate the same or corresponding parts.
Detailed Description
The principles and spirit of the present invention will be described with reference to a number of exemplary embodiments. It is understood that these embodiments are given solely for the purpose of enabling those skilled in the art to better understand and to practice the invention, and are not intended to limit the scope of the invention in any way. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
As will be appreciated by one skilled in the art, embodiments of the present invention may be embodied as a system, apparatus, device, method, or computer program product. Accordingly, the present disclosure may be embodied in the form of: entirely hardware, entirely software (including firmware, resident software, micro-code, etc.), or a combination of hardware and software.
According to the embodiment of the invention, a training method, a device, a medium and a computing device of an interpretable time sequence prediction model are provided.
In this document, it is to be understood that any number of elements in the figures are provided by way of illustration and not limitation, and any nomenclature is used for differentiation only and not in any limiting sense.
The principles and spirit of the present invention are explained in detail below with reference to several representative embodiments of the invention.
Exemplary method
Referring to fig. 1, fig. 1 is a flowchart illustrating an interpretable timing prediction model training method according to an embodiment of the invention. It should be noted that the embodiments of the present invention can be applied to any applicable scenarios.
Fig. 1 shows a process 100 of a training method of an interpretable timing prediction model according to an embodiment of the present invention, which includes:
step S101, data processing is carried out on the acquired time sequence data to obtain a sequence diagram corresponding to the time sequence data;
step S102, modeling the sequence diagram through the interpretable time sequence prediction model to obtain a prediction result corresponding to the sequence diagram;
step S103, training the interpretable time sequence prediction model based on the true value and the prediction result so as to improve the prediction precision of the interpretable time sequence prediction model.
The model processing method provided by the application aims at an interpretable time sequence prediction model constructed based on artificial intelligence represented by machine learning, particularly deep learning, can accurately realize tasks such as classification, prediction and completion of time sequence data through the interpretable time sequence prediction model, and can also interpret data output by the interpretable time sequence prediction model, including but not limited to sales prediction, stock analysis, traffic flow prediction, weather forecast and other prediction models suitable for different application scenarios.
According to the invention, the time sequence data can be processed to obtain a sequence diagram, the sequence diagram can be modeled through the interpretable time sequence prediction model to obtain a prediction result, the interpretable time sequence prediction model can calculate multi-dimensional data and time sequence data with time sequence so as to output the prediction result, and the interpretable time sequence prediction model can be trained on the basis of the prediction result and a true value, so that the interpretable time sequence prediction model can output a more accurate prediction result by analyzing the multi-dimensional data and the data with time sequence, and the prediction performance of the interpretable time sequence prediction model is also improved.
The following describes how to train the interpretable time sequence prediction model by using a suitable training mode to improve the prediction performance of the interpretable time sequence prediction model, with reference to the accompanying drawings:
in the embodiment of the present invention, the time-series data may be data having a predefined time or sequence, and may be widely applied to various application scenarios. The time series data can be subjected to data processing to obtain a sequence diagram corresponding to the time series data, the sequence diagram can contain information such as time, characteristics, characteristic importance and the like, continuous time series data within a preset time length can be obtained from the time series data, characteristics corresponding to the time series data can be obtained, and data such as characteristic importance and the like corresponding to the characteristics can be obtained.
In addition, the interpretable timing prediction model may include a deep learning model, and the deep learning model may be a Long Short-Term Memory network (LSTM), a Gated-Gated Unit (GRU), a Long-and Short-Term Time-series network (LSTNet), or the like, which is not limited in the embodiments of the present invention. The sequence diagram can be input into the interpretable time sequence prediction model, so that a deep learning model in the interpretable time sequence prediction model operates on the time sequence diagram, and a prediction result corresponding to the time sequence diagram is output through the deep learning model.
Furthermore, a true value corresponding to the time series data can be obtained, and the interpretable time series prediction model is trained through the true value and the prediction result, so that the prediction precision of the prediction result output by the interpretable time series prediction model is improved.
For example, the application scenario of the embodiment of the present invention may be an application scenario for predicting the power generated by solar energy, and the interpretable time sequence prediction model may be trained according to the collected power generated by the solar power plant in continuous time periods, so that the interpretable time sequence prediction model may accurately predict the power generated by the solar power plant.
Specifically, the generated power of a plurality of solar power plants in the same area can be obtained, the generated power can be obtained as the obtained time series data, a sequence diagram corresponding to the time series data can be obtained by processing the time series data, the sequence diagram can be modeled by the interpretable time series data to obtain a prediction result corresponding to the sequence diagram, after the prediction result is obtained, the real generated power of the plurality of solar power plants in a period of time in the future can be obtained, and the interpretable time series prediction model can be trained on the basis of the prediction result and the obtained real generated power, so that the prediction result output by the interpretable time series prediction model is more accurate, that is, the generated power of the plurality of power plants in the same area per day can be accurately predicted by the interpretable time series prediction model, the actual value of each time interval every day of the solar power generation system is modeled, the abnormal point of the multichannel time sequence power is detected, the negative influence of abnormal weather on the solar power supply system is reduced, and therefore the interpretable prediction result has important practical significance.
Referring to fig. 2, fig. 2 is a flowchart illustrating a method for training an interpretable timing prediction model according to another embodiment of the present invention, and a flowchart 200 of the method for training the interpretable timing prediction model according to another embodiment of the present invention shown in fig. 2 includes:
step S201, processing the acquired time sequence data according to a time axis sliding window to obtain a sequence diagram corresponding to the time sequence data;
by implementing the step S201, sequence diagrams corresponding to time series data with the same time scale can be acquired according to the time sliding window, and the prediction result can be obtained by calculating different sequence diagrams, so as to reduce the external influence on the prediction result in the interpretation process.
In the embodiment of the present invention, a method for processing the acquired time series data according to the time axis sliding window may specifically be: the preset fixed size of the time axis sliding window can be obtained, the time sequence data can be arranged according to the time axis, the time sequence data corresponding to the time axis sliding window can be obtained by continuously sliding the time axis sliding window with the fixed size along the time axis from the front to the back, the time sequence data corresponding to the time axis sliding window can be obtained by sequentially cutting, the characteristics corresponding to the time sequence data can be obtained, the data such as the characteristic importance corresponding to the characteristics can be obtained, and the sequence diagram corresponding to the time sequence data can be constructed based on the time, the characteristics and the characteristic importance of the associated time sequence data.
Step S202, determining the sequence diagram as an original sequence diagram;
step S203, combining the disturbance area with the original sequence diagram through the significance module to obtain a first disturbance sequence diagram; the interpretable temporal prediction model comprises a saliency module;
step S204, calculating the original sequence diagram and the first disturbance sequence diagram through the significance module to obtain a second disturbance sequence diagram;
step S205, modeling the second disturbance sequence diagram and the original sequence diagram to obtain a prediction result corresponding to the sequence diagram;
by implementing the steps S202 to S205, the sequence diagram can be operated for multiple times by the significance module included in the interpretable time sequence prediction model, so as to obtain a second disturbance sequence diagram corresponding to the sequence diagram, and a prediction result corresponding to the sequence diagram can be modeled based on the second disturbance sequence diagram and the original sequence diagram, so that the processing process of the sequence diagram is more comprehensive, and the accuracy of the prediction result is improved.
In the embodiment of the present invention, the interpretable time sequence prediction model may include a saliency module, the sequence diagram may be determined as an original sequence diagram X, the saliency module may determine a disturbance region, and the original sequence diagram X may be combined based on the determined disturbance region M to obtain a first disturbance sequence diagram
First perturbation sequence chart
The specific generation mode of (2) may be:
the time series data corresponding to the disturbance region M in the original sequence diagram X may be added with a mean value μ and a variance σ2Gaussian noise of (u, σ)2) To obtain a first disturbance sequence chart after disturbance
Specifically, a first disturbance sequence chart is calculated
May be:
ε~N(μ,σ2)
wherein M (x) in the disturbance region Mt,i)→[0,1]When m (x)t,i) =1 feature x representing original time series datat,iE X perturbed feature
Alternatively, when m (x)t,i) =0 represents the feature x using the original time-series datat,i∈X。
Still further, the saliency module may also pair the original sequence diagram X and the first perturbed sequence diagram
Calculating to obtain a second disturbance sequence chart
Specifically, calculating the second disturbance sequence chart
May be:
wherein, the operation is a dot-by-dot operation, and the perturbed second perturbation sequence diagram
Input to the deep learning model to make the deep learning model output the first prediction y0The original sequence diagram X can also be input into the linear autoregressive model, so that the linear autoregressive model outputs the second prediction yrSecond predictionyrMay be:
where b is a constant, second prediction yrAnd length k of history data xt-k，xt-k+1，…，xt-1On, for historical data by WkLinear weighting, i.e. y0And yrThe final prediction result can be obtained by combination
。
Step S206, training the interpretable time sequence prediction model based on the true value and the prediction result so as to improve the prediction precision of the interpretable time sequence prediction model.
Referring to fig. 3, fig. 3 is a flowchart illustrating a method for training an interpretable timing prediction model according to another embodiment of the present invention, and a flowchart 300 of the method for training the interpretable timing prediction model according to another embodiment of the present invention illustrated in fig. 3 includes:
step S301, processing the acquired time sequence data according to a time axis sliding window to obtain a sequence diagram corresponding to the time sequence data;
step S302, determining the sequence diagram as an original sequence diagram;
step S303, combining the disturbance area with the original sequence diagram through the significance module to obtain a first disturbance sequence diagram; the interpretable temporal prediction model comprises a saliency module;
step S304, calculating the original sequence diagram and the first disturbance sequence diagram through the significance module to obtain a second disturbance sequence diagram;
step S305, inputting the second disturbance sequence diagram into the deep learning module to obtain a first prediction result; the interpretable time series prediction model further comprises a deep learning module and a linear regression module;
step S306, inputting the original sequence diagram into the linear regression module to obtain a second prediction result;
step S307, combining the first prediction result and the second prediction result to obtain a prediction result corresponding to the sequence diagram;
by implementing the steps S305 to S307, the second disturbance sequence diagram can be calculated by the deep learning module of the interpretable time sequence prediction model to obtain the first prediction result, and the original sequence diagram can be calculated by the linear regression module of the interpretable time sequence prediction model to obtain the second prediction result, so that the prediction result corresponding to the sequence diagram formed by combining the first prediction result and the second prediction result can be obtained, and the obtained prediction result is more accurate and comprehensive.
In the embodiment of the invention, the interpretable time sequence prediction model can also comprise a deep learning module and a linear regression module, and different disturbance sequence charts can be processed through different modules to obtain different prediction results.
Wherein the deep learning module can be used for processing the input second disturbance sequence chart
And then the second disturbance sequence chart can be output through the deep learning module
Corresponding first prediction result y0(ii) a The linear regression module can be used for processing the input original sequence diagram X, and then the second prediction result y corresponding to the original sequence diagram X can be output through the linear regression moduler(ii) a The first prediction y can be further processed0And a second prediction result yrCombining to obtain final prediction result output by interpretable time sequence prediction model
。
Step S308, calculating a true value and the prediction result to obtain a deviation between the true value and the prediction result;
step S309, based on the loss function added with norm, making the loss function constrain the complexity of the disturbance area;
step S310, training the interpretable time sequence prediction model based on the disturbance area and the deviation so as to improve the prediction precision of the interpretable time sequence prediction model.
By implementing the steps S308 to S310, the deviation between the true value and the prediction result can be calculated, the complexity of the disturbance area can be constrained based on the loss function, and the interpretable time sequence prediction model is trained based on the disturbance area and the deviation, so that the prediction precision of the interpretable time sequence prediction model is improved.
In the embodiment of the invention, the real value can be y, and the deviation between the real value and the prediction result can be calculated by the mean square error
And calculating the deviation between the predicted value and the true value.
As an alternative embodiment, based on the loss function added with the norm, the way that the loss function constrains the complexity of the disturbance region M can be to use the norm lm=║M║p0And lrA loss function is added to achieve a constraint on the complexity of the perturbation region M, wherein,
the loss function may be:
the matrix norm can be constrained, for example, in the model algorithm, the norm is 2, a loss function is added to the matrix norm to serve as a regular term for optimization, and the complexity of the disturbance region M can be optimized.
In addition, the method of training the interpretable time sequence prediction model to improve the prediction accuracy of the interpretable time sequence prediction model may be: optimizing an objective function L in a gradient descent mode in a training process1And iteratively optimizing the parameters of each module in the model, wherein the objective function L1It may also be a loss function, an objective function L1May be:
L1=lp+λ1lm+λ2lr
wherein lpMay be the prediction error, λ1And λ2Can be balance coefficients among different loss functions, and the loss function L defined in the deep learning model can be a multi-task learning objective function by simultaneously optimizing Lp、lmAnd lrThe model can improve the prediction precision and can explain the prediction result of the model to calculate the target function L1Each part of the network architecture of the interpretable time sequence prediction model can be automatically derived through the pyrrch framework, and each layer of parameter theta is updated in sequence, wherein the updating formula of each layer of parameter theta can be as follows:
wherein, the deep learning model can optimize the loss function by a gradient descent method, γ is the learning rate, t is the number of iterations of multiple rounds,. L1(θ(t)) Represents the loss function L1For each parameter theta of the model in the t-th iteration(t)The updating amount of the time sequence prediction model is optimized and iterated for multiple rounds by means of a gradient descent method until the model converges, and parameters of the interpretable time sequence prediction model used for time sequence prediction gradually reach an optimal solution after the multiple rounds of iteration to obtain the optimal model performance.
As an alternative embodiment, after step S310, the following steps may be further performed:
acquiring an arbitrary sequence diagram;
and updating the attention range of the disturbance region based on each feature in the sequence diagram to obtain a sequence saliency map corresponding to the sequence diagram.
By implementing the implementation mode, the attention range of the disturbance region can be updated based on each feature in the sequence diagram, so that the sequence saliency map corresponding to the sequence diagram is obtained, the sequence saliency map can clearly explain the interpretable time sequence prediction model, and the interpretability of the interpretable time sequence prediction model is improved.
In the embodiment of the invention, the dynamic influence of each feature on the prediction result can be explained through the sequence saliency map corresponding to the generated sequence map, wherein any sequence map X can be sampled from a test set containing a plurality of sequence maps0Finding the characteristic causing the maximum performance reduction of the interpretable time sequence prediction model by disturbing different areas, and aiming at a multivariable time sequence regression task, hopefully finding the most critical disturbance area to maximize the prediction error l when explaining the interpretable time sequence prediction modelpDuring interpretation, perturbing the region of interest of the field by optimizing the loss function L2The adjustment is carried out, and a gradient descent method is adopted to optimize the loss function L through multiple rounds of iteration2Determining a disturbance region, an objective function L2May be:
L2=-lp+λ1lm+λ2lr
wherein lmAnd lrCan be defined in the training process to restrict the complexity of the disturbance area. In the process of explaining the sample, limiting other part parameters of the interpretable time sequence prediction model to be fixed, only updating the attention range of the disturbance area to obtain X0Corresponding sequence saliency map.
According to the technical scheme, the interpretable time sequence prediction model can output a more accurate prediction result by analyzing the data with multiple dimensions and time sequence, and the prediction performance of the interpretable time sequence prediction model is also improved. In addition, the external influence on the prediction result in the interpretation process can be reduced. In addition, the processing process of the sequence diagram can be more comprehensive, so that the accuracy of the prediction result is improved. In addition, the obtained prediction result can be more accurate and comprehensive. In addition, the prediction accuracy of the interpretable time sequence prediction model can be improved. In addition, the interpretability of the interpretable time sequence prediction model can be improved.
Exemplary devices
Having described the method of the exemplary embodiment of the present invention, next, a training apparatus of an interpretable timing prediction model of the exemplary embodiment of the present invention is described with reference to fig. 4, the apparatus including:
a processing unit 401, configured to perform data processing on the acquired time-series data to obtain a sequence diagram corresponding to the time-series data;
a modeling unit 402, configured to model the sequence diagram obtained by the processing unit 401 through the interpretable time-series prediction model, so as to obtain a prediction result corresponding to the sequence diagram;
a training unit 403, configured to train the interpretable time sequence prediction model based on the true value and the prediction result obtained by the modeling unit 402, so as to improve the prediction accuracy of the interpretable time sequence prediction model.
As an optional implementation manner, the processing unit 401 performs data processing on the acquired time-series data to obtain a sequence diagram corresponding to the time-series data specifically may be:
and processing the acquired time sequence data according to a time axis sliding window to obtain a sequence diagram corresponding to the time sequence data.
By implementing the implementation mode, the sequence diagrams corresponding to the time sequence data with the same time scale can be acquired according to the time sliding window, and the prediction result can be obtained by calculating different sequence diagrams, so that the external influence on the prediction result in the interpretation process is reduced.
As an alternative implementation, the interpretable time-series prediction model includes a significance module, and the modeling unit 402 may include:
a determining subunit, configured to determine the sequence diagram as an original sequence diagram;
a combining subunit, configured to combine the disturbance region with the original sequence diagram through the saliency module to obtain a first disturbance sequence diagram;
the first calculation subunit is configured to calculate the original sequence diagram and the first disturbance sequence diagram through the saliency module to obtain a second disturbance sequence diagram;
and the modeling subunit is used for modeling the second disturbance sequence diagram and the original sequence diagram to obtain a prediction result corresponding to the sequence diagram.
By implementing the implementation mode, the sequence diagram can be operated for multiple times through a significance module included in the interpretable time sequence prediction model to obtain a second disturbance sequence diagram corresponding to the sequence diagram, and the prediction result corresponding to the sequence diagram can be modeled based on the second disturbance sequence diagram and the original sequence diagram, so that the processing process of the sequence diagram is more comprehensive, and the accuracy of the prediction result is improved.
As an alternative implementation, the interpretable time series prediction model further includes a deep learning module and a linear regression module, and the modeling unit may include:
the modeling module is used for inputting the second disturbance sequence diagram into the deep learning module to obtain a first prediction result;
the modeling module is further used for inputting the original sequence diagram into the linear regression module to obtain a second prediction result;
and the combination module is used for combining the first prediction result and the second prediction result to obtain the prediction result corresponding to the sequence diagram.
By implementing the implementation mode, the second disturbance sequence diagram can be calculated through a deep learning module of the interpretable time sequence prediction model to obtain a first prediction result, the original sequence diagram can be calculated through a linear regression module of the interpretable time sequence prediction model to obtain a second prediction result, and then the prediction result corresponding to the sequence diagram formed by combining the first prediction result and the second prediction result can be obtained, so that the obtained prediction result is more accurate and comprehensive.
As an alternative embodiment, the training unit 403 may include:
the second calculation subunit is used for calculating a real value and the prediction result to obtain a deviation between the real value and the prediction result;
a constraint subunit, configured to constrain the complexity of the disturbance region by using a loss function based on a norm;
a training subunit, configured to train the interpretable timing prediction model based on the perturbation region and the deviation so as to improve prediction accuracy of the interpretable timing prediction model.
By implementing the implementation mode, the deviation between the true value and the prediction result can be calculated, the complexity of the disturbance area can be restrained based on the loss function, and the interpretable time sequence prediction model is trained based on the disturbance area and the deviation, so that the prediction precision of the interpretable time sequence prediction model is improved.
As an optional implementation, the apparatus may further include:
an acquisition unit configured to acquire an arbitrary sequence diagram after the training unit trains the interpretable timing prediction model with a view to improving the prediction accuracy of the interpretable timing prediction model based on a true value and the prediction result;
and the updating unit is used for updating the attention range of the disturbance region based on each feature in the sequence diagram so as to obtain a sequence saliency map corresponding to the sequence diagram.
By implementing the implementation mode, the attention range of the disturbance region can be updated based on each feature in the sequence diagram, so that the sequence saliency map corresponding to the sequence diagram is obtained, the sequence saliency map can clearly explain the interpretable time sequence prediction model, and the interpretability of the interpretable time sequence prediction model is improved.
Exemplary Medium
Having described the method and apparatus of the exemplary embodiment of the present invention, next, a computer-readable storage medium of the exemplary embodiment of the present invention is described with reference to fig. 5, please refer to fig. 5, which illustrates a computer-readable storage medium being an optical disc 50 having a computer program (i.e., a program product) stored thereon, where the computer program, when executed by a processor, implements the steps described in the above method embodiment, for example, performs data processing on the acquired time-series data to obtain a sequence diagram corresponding to the time-series data; modeling the sequence diagram through the interpretable time sequence prediction model to obtain a prediction result corresponding to the sequence diagram; training the interpretable time sequence prediction model with the aim of improving the prediction precision of the interpretable time sequence prediction model based on a real value and the prediction result; the specific implementation of each step is not repeated here.
It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory, or other optical and magnetic storage media, which are not described in detail herein.
Exemplary computing device
Having described the methods, media, and apparatus of exemplary embodiments of the present invention, a computing device for training of interpretable temporal prediction models of exemplary embodiments of the present invention is next described with reference to FIG. 6.
FIG. 6 illustrates a block diagram of an exemplary computing device 60 suitable for use in implementing embodiments of the present invention, the computing device 60 may be a computer system or server. The computing device 60 shown in FIG. 6 is only one example and should not be taken to limit the scope of use and functionality of embodiments of the present invention.
As shown in fig. 6, components of computing device 60 may include, but are not limited to: one or more processors or processing units 601, a system memory 602, and a bus 603 that couples various system components including the system memory 602 and the processing unit 601.
The system memory 602 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM) 6021 and/or cache memory 6022. Computing device 60 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, ROM6023 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 6, but typically referred to as a "hard disk drive"). Although not shown in FIG. 6, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 603 by one or more data media interfaces. At least one program product may be included in system memory 602 with a set (e.g., at least one) of program modules configured to perform the functions of embodiments of the present invention.
A program/utility 6025 having a set (at least one) of program modules 6024 may be stored, for example, in the system memory 602, and such program modules 6024 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment. Program modules 6024 generally carry out the functions and/or methodologies of embodiments of the invention as described herein.
The processing unit 601 executes various functional applications and data processing by running a program stored in the system memory 602, for example, performs data processing on the acquired time-series data to obtain a sequence diagram corresponding to the time-series data; modeling the sequence diagram through the interpretable time sequence prediction model to obtain a prediction result corresponding to the sequence diagram; training the interpretable timing prediction model with a view to improving the prediction accuracy of the interpretable timing prediction model based on the true value and the prediction result. The specific implementation of each step is not repeated here. It should be noted that although in the above detailed description several units/modules or subunits/sub-modules of the training apparatus of the interpretable temporal prediction model are mentioned, such partitioning is merely exemplary and not mandatory. Indeed, the features and functionality of two or more of the units/modules described above may be embodied in one unit/module according to embodiments of the invention. Conversely, the features and functions of one unit/module described above may be further divided into embodiments by a plurality of units/modules.
In the description of the present invention, it should be noted that the terms "first", "second", and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the embodiments provided in the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one logical division, and there may be other divisions when actually implemented, and for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or units through some communication interfaces, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present invention, which are used for illustrating the technical solutions of the present invention and not for limiting the same, and the protection scope of the present invention is not limited thereto, although the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Moreover, while the operations of the method of the invention are depicted in the drawings in a particular order, this does not require or imply that the operations must be performed in this particular order, or that all of the illustrated operations must be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step execution, and/or one step broken down into multiple step executions.
Claims (10)
1. An interpretable training method of a time sequence prediction model, which is applied to an application scene of predicting the generated power of a solar power station, and comprises the following steps:
performing data processing on the acquired time sequence data to obtain a sequence diagram corresponding to the time sequence data, wherein the time sequence data is the power generation power of the solar power station in continuous time periods;
modeling the sequence diagram through the interpretable time sequence prediction model to obtain a prediction result corresponding to the sequence diagram;
training the interpretable time sequence prediction model with the aim of improving the prediction precision of the interpretable time sequence prediction model based on a real value and the prediction result; wherein the prediction accuracy indicates the predicted generated power of the solar power plant;
the interpretable time sequence prediction model comprises a significance module, the interpretable time sequence prediction model models the sequence diagram to obtain a prediction result corresponding to the sequence diagram, and the method comprises the following steps:
determining the sequence diagram as an original sequence diagram;
combining the disturbance area with the original sequence diagram through the saliency module to obtain a first disturbance sequence diagram;
calculating the original sequence diagram and the first disturbance sequence diagram through the significance module to obtain a second disturbance sequence diagram;
modeling the second disturbance sequence diagram and the original sequence diagram to obtain a prediction result corresponding to the sequence diagram;
after training the interpretable time sequence prediction model for the purpose of improving the prediction accuracy of the interpretable time sequence prediction model based on the true value and the prediction result, the method further comprises:
acquiring an arbitrary sequence diagram;
updating the attention range of the disturbance region based on each feature in the sequence diagram to obtain a sequence saliency map corresponding to the sequence diagram; and the sequence saliency map corresponding to the sequence map is used for explaining the dynamic influence of each feature on the prediction result.
2. The method for training an interpretable time series prediction model according to claim 1, wherein the manner of obtaining a sequence diagram corresponding to the time series data by performing data processing on the acquired time series data is specifically as follows:
and processing the acquired time sequence data according to a time axis sliding window to obtain a sequence diagram corresponding to the time sequence data.
3. The method for training the interpretable time-series prediction model according to claim 2, wherein the interpretable time-series prediction model further comprises a deep learning module and a linear regression module, and the modeling of the second disturbance sequence diagram and the original sequence diagram results in a prediction result corresponding to the sequence diagram comprises:
inputting the second disturbance sequence diagram into the deep learning module to obtain a first prediction result;
inputting the original sequence diagram into the linear regression module to obtain a second prediction result;
and combining the first prediction result and the second prediction result to obtain a prediction result corresponding to the sequence diagram.
4. The method of claim 3, wherein training the interpretable timing prediction model with a view to improving the prediction accuracy of the interpretable timing prediction model based on the true value and the prediction result comprises:
calculating a real value and the prediction result to obtain a deviation between the real value and the prediction result;
a loss function based on an added norm such that the loss function constrains a complexity of the perturbation region;
training the interpretable timing prediction model based on the disturbance region and the bias with a view to improving a prediction accuracy of the interpretable timing prediction model.
5. An interpretable training device of a time sequence prediction model, which is applied to an application scene of predicting the generated power of a solar power station, and comprises:
the processing unit is used for carrying out data processing on the acquired time sequence data to obtain a sequence diagram corresponding to the time sequence data, wherein the time sequence data are the power generation power of the solar power station in continuous time periods;
the modeling unit is used for modeling the sequence diagram through the interpretable time sequence prediction model to obtain a prediction result corresponding to the sequence diagram;
a training unit for training the interpretable timing prediction model with a view to improving the prediction accuracy of the interpretable timing prediction model based on a true value and the prediction result; wherein the prediction accuracy indicates the predicted generated power of the solar power plant;
the interpretable temporal prediction model includes a saliency module, the modeling unit including:
a determining subunit, configured to determine the sequence diagram as an original sequence diagram;
a combining subunit, configured to combine the disturbance region with the original sequence diagram through the saliency module to obtain a first disturbance sequence diagram;
the first calculation subunit is configured to calculate the original sequence diagram and the first disturbance sequence diagram through the saliency module to obtain a second disturbance sequence diagram;
the modeling subunit is used for modeling the second disturbance sequence diagram and the original sequence diagram to obtain a prediction result corresponding to the sequence diagram;
the device further comprises:
an acquisition unit configured to acquire an arbitrary sequence diagram after the training unit trains the interpretable timing prediction model with a view to improving the prediction accuracy of the interpretable timing prediction model based on a true value and the prediction result;
an updating unit, configured to update the attention range of the disturbance region based on each feature in the sequence diagram, so as to obtain a sequence saliency map corresponding to the sequence diagram; and the sequence saliency map corresponding to the sequence map is used for explaining the dynamic influence of each feature on the prediction result.
6. The apparatus for training an interpretable time series prediction model according to claim 5, wherein the processing unit performs data processing on the acquired time series data to obtain a sequence diagram corresponding to the time series data in a specific manner:
and processing the acquired time sequence data according to a time axis sliding window to obtain a sequence diagram corresponding to the time sequence data.
7. Training apparatus of the interpretable temporal prediction model of claim 6, the interpretable temporal prediction model further comprising a deep learning module and a linear regression module, the modeling subunit comprising:
the modeling module is used for inputting the second disturbance sequence diagram into the deep learning module to obtain a first prediction result;
the modeling module is further used for inputting the original sequence diagram into the linear regression module to obtain a second prediction result;
and the combination module is used for combining the first prediction result and the second prediction result to obtain the prediction result corresponding to the sequence diagram.
8. Training device of the interpretable timing prediction model of claim 7, the training unit comprising:
the second calculation subunit is used for calculating a real value and the prediction result to obtain a deviation between the real value and the prediction result;
a constraint subunit, configured to constrain the complexity of the disturbance region by using a loss function based on a norm;
a training subunit, configured to train the interpretable timing prediction model based on the perturbation region and the deviation so as to improve prediction accuracy of the interpretable timing prediction model.
9. A storage medium storing a program, wherein the storage medium stores a computer program which, when executed by a processor, implements a method of training an interpretable timing prediction model according to any one of claims 1-4.
10. A computing device comprising the storage medium of claim 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110141485.6A CN113779859B (en) | 2021-02-02 | 2021-02-02 | Interpretable time sequence prediction model training method and device and computing equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110141485.6A CN113779859B (en) | 2021-02-02 | 2021-02-02 | Interpretable time sequence prediction model training method and device and computing equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113779859A CN113779859A (en) | 2021-12-10 |
| CN113779859B true CN113779859B (en) | 2022-04-05 |
Family
ID=78835583
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110141485.6A Active CN113779859B (en) | 2021-02-02 | 2021-02-02 | Interpretable time sequence prediction model training method and device and computing equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113779859B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114519610A (en) * | 2022-02-16 | 2022-05-20 | 支付宝(杭州)信息技术有限公司 | Information prediction method and device |
| CN115049067A (en) * | 2022-05-12 | 2022-09-13 | 支付宝(杭州)信息技术有限公司 | Model interpretation and training method, device, equipment, medium and program product |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2002228814A1 (en) * | 2000-11-09 | 2002-07-25 | International Business Machines Corporation | System and method for building a time series model |
| CN101541030A (en) * | 2009-05-06 | 2009-09-23 | 华为技术有限公司 | Method for predicting data based on support vector machine and equipment thereof |
| CN106777874A (en) * | 2016-11-18 | 2017-05-31 | 中国科学院自动化研究所 | The method that forecast model is built based on Recognition with Recurrent Neural Network |
| CN110889560A (en) * | 2019-12-06 | 2020-03-17 | 西北工业大学 | Express delivery sequence prediction method with deep interpretability |
| CN110942101A (en) * | 2019-11-29 | 2020-03-31 | 湖南科技大学 | Rolling bearing residual life prediction method based on depth generation type countermeasure network |
| CN111209370A (en) * | 2019-12-27 | 2020-05-29 | 同济大学 | Text classification method based on neural network interpretability |
| CN111898318A (en) * | 2020-08-04 | 2020-11-06 | 西安建筑科技大学 | Industrial building concrete carbonization depth prediction method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102011011929A1 (en) * | 2011-02-18 | 2012-08-23 | Hella Kgaa Hueck & Co. | Method for detecting target objects in a surveillance area |
-
2021
- 2021-02-02 CN CN202110141485.6A patent/CN113779859B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2002228814A1 (en) * | 2000-11-09 | 2002-07-25 | International Business Machines Corporation | System and method for building a time series model |
| CN101541030A (en) * | 2009-05-06 | 2009-09-23 | 华为技术有限公司 | Method for predicting data based on support vector machine and equipment thereof |
| CN106777874A (en) * | 2016-11-18 | 2017-05-31 | 中国科学院自动化研究所 | The method that forecast model is built based on Recognition with Recurrent Neural Network |
| CN110942101A (en) * | 2019-11-29 | 2020-03-31 | 湖南科技大学 | Rolling bearing residual life prediction method based on depth generation type countermeasure network |
| CN110889560A (en) * | 2019-12-06 | 2020-03-17 | 西北工业大学 | Express delivery sequence prediction method with deep interpretability |
| CN111209370A (en) * | 2019-12-27 | 2020-05-29 | 同济大学 | Text classification method based on neural network interpretability |
| CN111898318A (en) * | 2020-08-04 | 2020-11-06 | 西安建筑科技大学 | Industrial building concrete carbonization depth prediction method |
Non-Patent Citations (1)
| Title |
|---|
| 长短期记忆网络预测混沌时间序列;熊有成 等;《中国科学：物理学 力学 天文学》;20190531;正文第2-3节 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113779859A (en) | 2021-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Russo et al. | Air quality prediction using optimal neural networks with stochastic variables | |
| EP3504666A1 (en) | Asychronous training of machine learning model | |
| CN113779859B (en) | Interpretable time sequence prediction model training method and device and computing equipment | |
| CN112232495B (en) | Prediction model training method, device, medium and computing equipment | |
| Oo et al. | Time series prediction based on Facebook Prophet: a case study, temperature forecasting in Myintkyina | |
| EP3200038A1 (en) | Model evaluation device, model evaluation method, and program recording medium | |
| CN112668238B (en) | Rainfall processing method, rainfall processing device, rainfall processing equipment and storage medium | |
| Girimurugan et al. | Application of deep learning to the prediction of solar irradiance through missing data | |
| JPWO2019069865A1 (en) | Parameter estimation system, parameter estimation method and parameter estimation program | |
| CN116522467B (en) | Prediction method and device for service life of spacecraft component | |
| Kong et al. | Multi-step short-term solar radiation prediction based on empirical mode decomposition and gated recurrent unit optimized via an attention mechanism | |
| Stephenson | Bayesian inference for extreme value modelling | |
| Schneider et al. | Opinion: Optimizing climate models with process knowledge, resolution, and artificial intelligence | |
| Nandhini et al. | Predictive analytics for climate change detection and disease diagnosis | |
| JP6467953B2 (en) | Temperature prediction system, temperature prediction method and program | |
| CN117131654A (en) | Target observation method based on nonlinear optimal disturbance of pre-analysis initial guess condition | |
| Çetiner et al. | Analysis of different regression algorithms for the estimate of energy consumption | |
| CN116613732A (en) | Multi-element load prediction method and system based on SHAP value selection strategy | |
| van der Herten et al. | Multi-objective Bayesian optimization for engineering simulation | |
| Suhartono et al. | Hybrid Machine Learning for Forecasting and Monitoring Air Pollution in Surabaya | |
| Vakilzadeh | Stochastic model updating and model selection with application to structural dynamics | |
| Shiri et al. | Data-driven segmentation of long term condition monitoring data in the presence of heavy-tailed distributed noise with finite-variance | |
| Laurini et al. | Smoothing sample extremes: The mixed model approach | |
| Ntotsios et al. | Optimal sensor location methodology for structural identification and damage detection. | |
| CN116611022B (en) | Intelligent campus education big data fusion method and platform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |