CN111598329A - Time sequence data prediction method based on automatic parameter adjustment recurrent neural network - Google Patents
Time sequence data prediction method based on automatic parameter adjustment recurrent neural network Download PDFInfo
- Publication number
- CN111598329A CN111598329A CN202010401266.2A CN202010401266A CN111598329A CN 111598329 A CN111598329 A CN 111598329A CN 202010401266 A CN202010401266 A CN 202010401266A CN 111598329 A CN111598329 A CN 111598329A
- Authority
- CN
- China
- Prior art keywords
- prediction
- industry
- data
- time 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2474—Sequence data queries, e.g. querying versioned data
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/045—Combinations of networks
Abstract
The embodiment of the invention provides a time sequence data prediction method and a training method of a prediction model, wherein the training method comprises the following steps: respectively inquiring time series data of a given industry and a given industry key factor from a previously obtained industry time series data set and an industry key factor information time series data set; and according to the prediction period length, dividing the time sequence data of the key factors of the given industry and the given industry to obtain a training set, and training the time sequence data prediction model by using the training set. The time sequence data prediction method comprises the step of inputting time sequence data of key factors of a given industry and the given industry into a time sequence data prediction model trained in advance by the training method to obtain an industry prediction result. By using the method, the prediction information of the industry can be obtained according to the industry historical information and the selected key factor information, and meanwhile, the workload of the training process is reduced by utilizing automatic parameter adjustment in the characteristic engineering and the super-parameter tuning in the training.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to a time sequence data prediction method based on an automatic parameter adjustment recurrent neural network.
Background
The prediction of future long-term performance is the final target of financial analysis in the financial field, and the prediction of future performance conditions according to a plurality of historical prospective factors of enterprises can be carried out by utilizing a neural network model. Meanwhile, aiming at different enterprises and prospective factors thereof, different industries have different data characteristics and different quantity and data characteristics of the prospective factors, and a unique network and corresponding parameters with generalization capability meeting requirements cannot be found; and a large amount of feature selection and network parameter adjustment are involved in the network model training process, and the adjustment of network model features and hyper-parameters is a tedious but crucial task.
Therefore, a new prediction method is needed to solve the above technical problems.
Disclosure of Invention
The embodiment of the invention provides a time sequence data prediction method, which can acquire industry prediction information according to different industry historical information and selected different key factor information based on a time sequence data prediction model, and provides a training method of the time sequence data prediction model at the same time.
The invention adopts a technical scheme for solving the technical problems that a training method of a time sequence data prediction model is provided, and the method comprises the following steps:
acquiring an industry time sequence data set;
acquiring industry key factor information and a time sequence data set thereof;
acquiring a given industry, and intensively inquiring the time sequence data of the given industry from the industry time sequence data according to the given industry;
acquiring a given industry key factor, and intensively inquiring the industry key factor information and the time sequence data thereof to obtain the time sequence data of the given industry key factor according to the given industry key factor;
setting a prediction period length, and dividing time series data of a given industry and time series data of a given industry key factor according to the prediction period length to obtain a training data set;
and training a pre-established time sequence data prediction model based on the prediction period length by utilizing the training data set.
Preferably, the time series data prediction model is based on a recurrent neural network.
Preferably, the time series data prediction model is based on a long-short term memory neural network or a gated cyclic unit network.
Preferably, the prediction period is composed of an input number and an inference number, the input number is the number of input data for one prediction, and the inference number is the number of prediction results obtained by one prediction.
Preferably, the training, by using the training set, the time series data prediction model pre-established according to the prediction cycle length includes: and (4) carrying out data feature extraction and hyper-parameter selection of the time sequence data prediction model by utilizing automatic parameter adjustment.
Specifically, wherein the time series data prediction model further comprises an optimizer;
the method for extracting the data characteristics and selecting the hyper-parameters of the time sequence data prediction model by utilizing automatic parameter adjustment comprises the following steps:
the optimization device searches a parameter space of the time sequence data prediction model by using the gradient information to obtain a parameter combination in the parameter space, and the time sequence data prediction model is configured by using the parameter combination; the parameter space includes features extracted from the input data by the time series data prediction model and hyper-parameters of the time series data prediction model;
and evaluating the effect of the data prediction model under the parameter combination, and automatically adjusting the characteristics and the hyper-parameters according to the effect until the effect reaches a set expectation.
Preferably, the training, by using the training set, a pre-established time series data prediction model is trained, further comprising:
whether to end the training process is judged by whether the output result meets the set requirement,
if the output result meets the set requirement, ending the training process;
and if the output result does not meet the set requirement, performing automatic parameter adjustment on the time sequence data prediction model according to the output result, and continuing the training process.
In another aspect, a method for predicting time series data is provided, the method including:
acquiring time series data of a given industry;
acquiring time series data of key factors of a given industry;
obtaining a data set according to the time series data of the given industry and the time series data of the key factors of the given industry;
after the prediction data set is divided, a time sequence data prediction model trained in advance according to the method of claim 1 is input to obtain an industry prediction result.
Specifically, the inputting of the time series data prediction model trained in advance according to the method of claim 1 to obtain the industry prediction result includes:
obtaining the pre-trained time sequence data prediction model, wherein the prediction period set in the pre-training process comprises input quantity and inference quantity; the maximum quantity of input data of one-time prediction and the maximum quantity of prediction results obtained by one-time prediction of a pre-trained time sequence data prediction model are limited by the setting of the length of a prediction period in the method of claim 1;
and inputting any industry data which is less than or equal to the input quantity into the pre-trained time sequence data prediction model to obtain any industry prediction result which is less than or equal to the inference quantity.
By using the time sequence data prediction method and the training method of the time sequence data prediction model provided by the embodiment of the invention, the prediction information of the industry can be obtained according to different industry history information and selected different key factor information, and meanwhile, automatic parameter adjustment is introduced in characteristic engineering and super-parameter optimization in training, so that no additional manual operation is needed except for initial setting, and the workload consumed in the training process is reduced.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a flowchart of a training method of a time series data prediction model according to an embodiment of the present invention;
FIG. 2 is a diagram illustrating a process of training and predicting a time series data prediction model according to an embodiment of the present invention;
FIG. 3 is a block diagram of a recurrent neural network provided in accordance with an embodiment of the present invention;
FIG. 4 is a schematic diagram of an automated parameter adjustment process provided by one embodiment of the present invention;
FIG. 5 is a flowchart of a method for predicting time series data according to an embodiment of the present invention;
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In summary, the prediction of future long-term performance is the final objective of financial analysis, the future performance situation is predicted according to multiple historical prospective factors of enterprises, and the method belongs to time series prediction and is suitable for a recurrent neural network model. Meanwhile, aiming at different enterprises and their look-ahead factors, different industries have different data characteristics, and different look-ahead factor quantities and data characteristics, so that a unique network and corresponding parameters with generalization capability meeting requirements cannot be found, and therefore different network models are required to be used for training to obtain a result suitable for the relevant look-ahead factors of each enterprise machine; and a large amount of feature selection and network parameter adjustment are involved in the network model training process, and the adjustment of network model features and hyper-parameters is a tedious but crucial task.
In view of the above problems, embodiments of the present invention provide a time series data prediction method and a training method for a prediction model, and for a part which is most time-consuming in training of a network model, the time series data prediction method is just characteristic engineering and hyper-parameter tuning, and automated parameter adjustment is introduced, aiming at finding an optimal hyper-parameter in a shorter time by using heuristic search with a policy, and without additional manual operation except for initial setting.
Fig. 1 is a flowchart illustrating a training method of a time series data prediction model according to an embodiment of the present invention. As shown in fig. 1, the training process at least comprises the following steps:
and 11, acquiring an industry time series data set.
In one embodiment, the set of industry time series data is stored in a database.
In another embodiment, the industry time series data includes revenue data, price data, index data, and the like.
And step 12, acquiring the industry key factor information and a time sequence data set thereof.
In one embodiment, the business critical factor information and its time series data set are stored in a database.
In another embodiment, the business key factor information and its time series data comprise information of economic factors (macro economic factors, business factors, etc.) and time series data.
In one example, the business key factor information and its time series dataset and the business time series dataset are stored in two databases. In another example, the business key factor information and its time series dataset and the business time series dataset are stored in a database.
And step 13, acquiring a given industry, and intensively inquiring the time sequence data of the given industry from the industry time sequence data according to the given industry.
In one embodiment, time series data for a given industry is looked up from a database, which stores industry critical factor information and its time series data set, based on the given industry.
In one embodiment, time series data for a given business critical factor is looked up from a database that holds business critical factor information and its time series data set based on the given business critical factor. In one example, the given industry key factor is determined based on the given industry obtained in step 13.
And step 15, setting the length of a prediction period, and dividing the time series data of the given industry and the time series data of the key factors of the given industry according to the length of the prediction period to obtain a training set.
In one embodiment, the prediction period is composed of a training number and an inference number, the training number is the number of training data used for prediction in one prediction, and the inference number is the number of prediction results in one prediction.
The training quantity and the inference quantity refer to the data length of the time sequence required for performing each network training by using the divided data, and can be said to be how long a time sequence result is predicted by using how long historical time sequence data.
In one embodiment, the time series data prediction model is based on a recurrent neural network.
In another embodiment, the time series data prediction model is based on a long-short term memory neural network, a gated cyclic unit network.
Fig. 3 is a diagram illustrating a recurrent neural network architecture provided by an embodiment of the present invention. As shown in fig. 3, the dark gray portion is training data composed of industry data and look-ahead factor data, the light gray portion is an intermediate state vector, and the number of the intermediate state vectors is determined according to the training number (training number) in the prediction period; the middle gray part is a prediction result, and the number of the prediction results is determined according to the number of inferences (prediction number) in the prediction period. Because the long-short time memory (LSTM) neural network structure is complex and may generate an overfitting phenomenon in the case of a small data volume, in an example, a GRU network model with a relatively simple network structure may also be used, the network model is input as a plurality of time series, the network model is output as a prediction result of a plurality of time points, and the network model simultaneously uses the time series characteristics of look-ahead factor data and industry data.
In another embodiment, training a time series data prediction model pre-established according to the prediction cycle length by using the training set includes: and (4) carrying out data feature extraction and hyper-parameter selection of the time sequence data prediction model by utilizing automatic parameter adjustment.
In a particular embodiment, the time series data prediction model further comprises an optimizer;
the method comprises the following steps of utilizing automatic parameter adjustment to extract data characteristics and select hyper-parameters of a time sequence data prediction model, and realizing the following steps:
the optimization device searches a parameter space of the time sequence data prediction model by using the gradient information to obtain one configuration in the parameter space, and configures the time sequence data prediction model by using the configuration; the parameter space comprises characteristics extracted from input data by the time series data prediction model and hyper-parameters of the time series data prediction model;
and evaluating the effect of the data prediction model under the parameter configuration, and automatically adjusting the characteristics and the hyper-parameters according to the effect until the effect reaches a set expectation.
In another embodiment, the training set is used to train a pre-established time series data prediction model, and the method further includes the following steps:
whether to end the training process is judged by whether the output result meets the set requirement,
if the output result meets the set requirement, ending the training process;
and if the output result does not meet the set requirement, performing automatic parameter adjustment on the time sequence data prediction model according to the output result, and continuing the training process.
Further, the method shown in fig. 1 is explained, and essentially, the training of the neural network is to search for the optimal parameters, but the search for the parameters can be assisted by using gradient information (the network structure is fixed, and is actually a multi-objective parameter optimization process). FIG. 4 is a diagram illustrating an automated parameter adjustment process provided by one embodiment of the present invention. As shown in fig. 4, the parameter search space includes features and hyper-parameters, one of the parameter spaces is obtained through configuration of the optimizer, the effect of the parameter configuration is obtained through evaluation, and iteration is performed until the threshold requirement is met. Automated feature engineering aims at one way of training by automatically creating candidate features from a dataset and selecting from them a number of best features; the selection of the hyper-parameters has great influence on the final effect of the model, for example, a complex model may have better expression capability to process different types of data, but the gradient may disappear due to too many layers and the training may not be performed, for example, the learning rate is too high, the convergence effect may be poor, the convergence speed may be too low, and the automatic hyper-parameter selection aims to automatically perform the combination of the hyper-parameters from the hyper-parameters.
Based on the time sequence data prediction model trained by the method shown in fig. 1, the embodiment of the invention also provides a time sequence data prediction method. Fig. 5 is a flowchart illustrating a method for predicting time series data according to an embodiment of the present invention. As shown in fig. 5, the prediction process at least includes the following steps:
And 53, obtaining a prediction data set according to the time series data of the given industry and the time series data of the key factors of the given industry.
And step 54, dividing the prediction data set, inputting a time sequence data prediction model pre-trained according to the method of the figure 1, and obtaining an industry prediction result.
In one embodiment, any industry prediction result less than or equal to the inference number in the prediction period is obtained by inputting any industry data less than or equal to the input number in the prediction period into a pre-trained time series data prediction model.
In the prediction using the trained prediction model, the longest usable data length and the longest inferable data length are defined, and the sizes of the input and output data lengths can be selected according to the prediction requirement within the longest range of the usable data length and the inferable data length.
It is understood that there may be different specific implementations of the architecture provided by the present invention, and the different specific implementations do not depart from the spirit and scope of the present invention, and the technical effects obtained by the implementation of the method of the present invention are all within the scope of the present invention.
It can be seen from the foregoing embodiments that, with the adoption of the time series data prediction method and the training method of the prediction model provided by the embodiments of the present invention, the training method includes: acquiring an industry time sequence data set; acquiring industry key factor information and a time sequence data set thereof; acquiring a given industry, and intensively inquiring the time sequence data of the given industry from the industry time sequence data according to the given industry; acquiring a given industry key factor, and intensively inquiring the industry key factor information and the time sequence data thereof to obtain the time sequence data of the given industry key factor according to the given industry key factor; setting a prediction period length, and dividing time series data of a given industry and time series data of a given industry key factor according to the prediction period length to obtain a training set; and training a pre-established time sequence data prediction model by utilizing the training set. The time sequence data prediction method comprises the steps of obtaining time sequence data of a given industry; acquiring time series data of key factors of a given industry; obtaining a data set according to the time series data of the given industry and the time series data of the key factors of the given industry; and inputting the data set into a time sequence data prediction model pre-trained according to the training method to obtain an industry prediction result. By using the method, on one hand, the prediction information of the industry can be obtained according to different industry historical information and the selected different key factor information. On the other hand, automatic parameter adjustment is introduced in characteristic engineering and hyper-parameter tuning in the predictive model training process, no additional manual operation is needed except for initial setting, and the workload consumed in the training process is reduced.
The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
Those of skill would further appreciate that the various illustrative components and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied in hardware, a software module executed by a processor, or a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are merely exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (9)
1. A method of training a time series data prediction model, the method comprising:
acquiring an industry time sequence data set;
acquiring industry key factor information and a time sequence data set thereof;
acquiring a given industry, and intensively inquiring the time sequence data of the given industry from the industry time sequence data according to the given industry;
acquiring a given industry key factor, and intensively inquiring the industry key factor information and the time sequence data thereof to obtain the time sequence data of the given industry key factor according to the given industry key factor;
setting a prediction period length, and dividing time series data of a given industry and time series data of a given industry key factor according to the prediction period length to obtain a training data set;
and training a pre-established time sequence data prediction model based on the prediction period length by utilizing the training data set.
2. The method of claim 1, wherein the time series data prediction model is based on a recurrent neural network.
3. The method of claim 2, wherein the recurrent neural network comprises a long-short term memory neural network or a gated recurrent cell network.
4. The method according to claim 1, wherein the prediction period is composed of an input number and an inference number, the input number is the number of input data for one prediction, and the inference number is the number of prediction results obtained by one prediction.
5. The method of claim 1, wherein the training, using the training set, a time series data prediction model pre-established according to the prediction cycle length comprises: and (4) carrying out data feature extraction and hyper-parameter selection of the time sequence data prediction model by utilizing automatic parameter adjustment.
6. The method of claim 5, wherein the time series data prediction model further comprises an optimizer;
the method for extracting the data characteristics and selecting the hyper-parameters of the time sequence data prediction model by utilizing automatic parameter adjustment comprises the following steps:
the optimization device searches a parameter space of the time sequence data prediction model by using the gradient information to obtain a parameter combination in the parameter space, and the time sequence data prediction model is configured by using the parameter combination; the parameter space includes features extracted from the input data by the time series data prediction model and hyper-parameters of the time series data prediction model;
and evaluating the effect of the data prediction model under the parameter combination, and automatically adjusting the characteristics and the hyper-parameters according to the effect until the effect reaches a set expectation.
7. The method of claim 1, training a pre-established time series data prediction model using the training set, further comprising:
whether to end the training process is judged by whether the output result meets the set requirement,
if the output result meets the set requirement, ending the training process;
and if the output result does not meet the set requirement, performing automatic parameter adjustment on the time sequence data prediction model according to the output result, and continuing the training process.
8. A method of predicting time series data, the method comprising:
acquiring time series data of a given industry;
acquiring time series data of key factors of a given industry;
obtaining a prediction data set according to the time series data of the given industry and the time series data of the key factors of the given industry;
after the prediction data set is divided, a time sequence data prediction model trained in advance according to the method of claim 1 is input to obtain an industry prediction result.
9. The method of claim 8, wherein after the prediction data set is divided, inputting a time series data prediction model trained in advance according to the method of claim 1 to obtain an industry prediction result, and the method comprises:
obtaining the pre-trained time sequence data prediction model, wherein the prediction period set in the pre-training process comprises input quantity and inference quantity; the maximum quantity of input data of one-time prediction and the maximum quantity of prediction results obtained by one-time prediction of a pre-trained time sequence data prediction model are limited by the setting of the length of a prediction period in the method of claim 1;
and inputting any industry data which is less than or equal to the input quantity into the pre-trained time sequence data prediction model to obtain any industry prediction result which is less than or equal to the inference quantity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010401266.2A CN111598329A (en) | 2020-05-13 | 2020-05-13 | Time sequence data prediction method based on automatic parameter adjustment recurrent neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010401266.2A CN111598329A (en) | 2020-05-13 | 2020-05-13 | Time sequence data prediction method based on automatic parameter adjustment recurrent neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111598329A true CN111598329A (en) | 2020-08-28 |
Family
ID=72188658
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010401266.2A Pending CN111598329A (en) | 2020-05-13 | 2020-05-13 | Time sequence data prediction method based on automatic parameter adjustment recurrent neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111598329A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112215696A (en) * | 2020-09-28 | 2021-01-12 | 北京大学 | Personal credit evaluation and interpretation method, device, equipment and storage medium based on time sequence attribution analysis |
| CN113312497A (en) * | 2021-06-01 | 2021-08-27 | 中国科学院计算机网络信息中心 | Prospective factor screening method and system based on knowledge graph |
| CN115081586A (en) * | 2022-05-19 | 2022-09-20 | 中国科学院计算机网络信息中心 | Short-term time sequence prediction method and system based on time and space attention |
-
2020
- 2020-05-13 CN CN202010401266.2A patent/CN111598329A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112215696A (en) * | 2020-09-28 | 2021-01-12 | 北京大学 | Personal credit evaluation and interpretation method, device, equipment and storage medium based on time sequence attribution analysis |
| CN113312497A (en) * | 2021-06-01 | 2021-08-27 | 中国科学院计算机网络信息中心 | Prospective factor screening method and system based on knowledge graph |
| CN115081586A (en) * | 2022-05-19 | 2022-09-20 | 中国科学院计算机网络信息中心 | Short-term time sequence prediction method and system based on time and space attention |
| CN115081586B (en) * | 2022-05-19 | 2023-03-31 | 中国科学院计算机网络信息中心 | Photovoltaic power generation time sequence prediction method and system based on time and space attention |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Automating feature subspace exploration via multi-agent reinforcement learning | |
| CN111598329A (en) | Time sequence data prediction method based on automatic parameter adjustment recurrent neural network | |
| CN110110858B (en) | Automatic machine learning method based on reinforcement learning | |
| US11366806B2 (en) | Automated feature generation for machine learning application | |
| CN110378382A (en) | Novel quantization transaction system and its implementation based on deeply study | |
| CN109191276A (en) | A kind of P2P network loan institutional risk appraisal procedure based on intensified learning | |
| Wu et al. | Optimized deep learning framework for water distribution data-driven modeling | |
| Lima et al. | Artificial intelligence applied to software testing: A literature review | |
| CN114548591A (en) | Time sequence data prediction method and system based on hybrid deep learning model and Stacking | |
| Wimmer et al. | Leveraging vision-language models for granular market change prediction | |
| CN117439053A (en) | Method, device and storage medium for predicting electric quantity of Stacking integrated model | |
| Weytjens et al. | Learning uncertainty with artificial neural networks for improved remaining time prediction of business processes | |
| CN113282747A (en) | Text classification method based on automatic machine learning algorithm selection | |
| Zhou et al. | Deep learning-based intelligent multilevel predictive maintenance framework considering comprehensive cost | |
| Li et al. | Research on the application of multimedia entropy method in data mining of retail business | |
| CN116703607A (en) | Financial time sequence prediction method and system based on diffusion model | |
| Jadli et al. | A Novel LSTM-GRU-Based Hybrid Approach for Electrical Products Demand Forecasting. | |
| CN116542701A (en) | Carbon price prediction method and system based on CNN-LSTM combination model | |
| Demidova et al. | Optimization of hyperparameters with constraints on time and memory for the classification model of the hard drives states | |
| Spiliotis | Time Series Forecasting with Statistical, Machine Learning, and Deep Learning Methods: Past, Present, and Future | |
| Rahman et al. | Implementation of artificial neural network on regression analysis | |
| KR102433431B1 (en) | Improved prediction system and prediction method using heterogeneous data | |
| Meng et al. | Application of EEMD+ BI_GRU hybrid model for intelligent service area traffic flow forecasting. | |
| Pulkkinen | Forecasting emergency department arrivals with neural networks | |
| Gicić et al. | Proposal of a model for credit risk prediction based on deep learning methods and SMOTE techniques for imbalanced dataset |
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 |