CN103294601A - Software reliability forecasting method based on selective dynamic weight neural network integration - Google Patents
Software reliability forecasting method based on selective dynamic weight neural network integration Download PDFInfo
- Publication number
- CN103294601A CN103294601A CN2013102769339A CN201310276933A CN103294601A CN 103294601 A CN103294601 A CN 103294601A CN 2013102769339 A CN2013102769339 A CN 2013102769339A CN 201310276933 A CN201310276933 A CN 201310276933A CN 103294601 A CN103294601 A CN 103294601A
- Authority
- CN
- China
- Prior art keywords
- neural network
- individuality
- model
- mistake
- pass
- 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.)
- Granted
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the field of software reliability forecasting, and particularly relates to a software reliability forecasting method based on selective dynamic weight neural network integration. The software reliability forecasting method mainly includes the steps: A, generating neural network individuals: selecting Elman neural networks as network individuals and generating n neural network individuals by a Bagging algorithm; B, optimizing the individuals: firstly, determining the cluster number of the generated neural network individuals by a K value optimization algorithm, secondly, clustering the neural network individuals according to a K-mean clustering algorithm to increase individual difference, and finally, integrating the clustered individuals; C, building a dynamic model: building a dynamic weight model based on a fuzzy neural network by the aid of errors of fitting data of the optimized individuals; and D, performing integrated output: combining forecasting results of the optimized individuals with weights generated by the dynamic weight model to generate final forecasting results. A neural network integration algorithm is applied to software reliability forecasting, and the software reliability forecasting method has the advantages of high precision and fine stability.
Description
Technical field
The invention belongs to the software reliability prediction field, can be used for the prediction of software reliability aspect, specifically is a kind of based on the integrated software reliability prediction method of selectivity changeable weight neural network.
Background technology
Along with the continuous development of infotech, computer software has obtained significant progress and has used widely.But because there is uncontrollability in computer system environment of living in, usually can be owing to user's faulty operation is brought out mistake potential in the software and then caused thrashing, serious threat people's safety of life and property.Therefore people are more and more urgent to the demand of high-quality software, and as the software reliability of software quality important indicator, become the focus that people pay close attention to.
The software reliability theoretical developments so far, be born a large amount of models and mutation.Traditional model is often set up based on hypothesis, and this just causes model only to be only applicable to specific occasions.Continuous development along with artificial intelligence technology, neural network is as a kind of important method in machine learning field, owing to have self-adaptation and learning functionality, and only need given abundant training data to train in use, just can obtain a forecast model more satisfactory than conventional model, these advantages make it become the new lover in software reliability prediction field.When yet any single model is predicted, unhealthy and strong characteristic under cover often, single model is understood the slight change of incompatibility software engineering project and predicated error is sharply increased in some cases, and since the shortcomings such as over-fitting that neural network self exists make when using single model to predict, often to have problems.
Neural network is integrated formally to be proposed in nineteen ninety by Hansen and Salamon.Hansen and Salamon experiment showed, by using a plurality of neural networks that same problem is learnt, and then their predicting the outcome are synthesized, and can improve the generalization ability of neural network effectively.Therefore the present invention proposes a kind of based on the integrated software reliability prediction method of selectivity changeable weight neural network, solve the problem that the cluster number need manually be set in the clustering algorithm by optimizing algorithm based on the k value of distance cost function, and solve the integrated output problem of neural network by the changeable weight model of setting up based on fuzzy neural network.
Summary of the invention
The present invention proposes the integrated software reliability prediction method of a kind of selectivity changeable weight neural network, the problem such as not robustness and over-fitting etc. that exists when overcoming single Neural Network model predictive.
Selectivity changeable weight neural network integrated software reliability prediction method mainly comprises following four steps:
A. generate the neural network individuality: choose the Elman neural network as the network individuality, produce n neural network individuality by the Bagging algorithm;
B. individual preferred: to the neural network individuality that generates, at first optimize algorithm by the K value and determine the cluster number, by the K-means clustering algorithm neural network individuality is carried out cluster then and increase individual difference, the individuality after the cluster carries out integrated the most at last;
C. dynamic model is set up: utilize the error of preferred individual fitting data, make up the changeable weight model based on fuzzy neural network;
D. integrated output: will be preferred individual predict the outcome and weight that the changeable weight model generates combines to generate and finally predicts the outcome;
Further, in above-mentioned steps A, in conjunction with the Elman neural network, and utilize the Bagging algorithm to carry out individual generation, specifically comprise following key step:
1) given original training set X=(x
1, y
1), (x
2, y
2), (x
3, y
3) ... (x
n, y
n), wherein n is the capacity of original training set, and the maximum iteration time of training is T, and the initial integrated E of neural network is empty set;
2) the initialization training set adopts m training sample of Bootstrap method for resampling extraction to form sub-training set X
t
3) by training set X
tTrain t neural network h
t
Further, in above-mentioned steps B, utilize the individual following steps of carrying out of the neural network that generates:
1) with the output result of each neural network individuality on training set
(wherein m is the number of training output sample point) forms matrix Y (y
1, y
2..., y
n), then matrix Y is carried out cluster analysis;
2) to k=1,2 ..., the T circulation:
(1) by the K-means clustering algorithm matrix Y is divided into the k group;
(2) mistake under the current k value of calculating! Can not pass through edit field code establishing object.；
3) search 1 to T between apart from cost minimum value mistake! Can not pass through edit field code establishing object.；
4) use the k value that matrix Y is carried out cluster, in every cluster data, select the highest individuality of fitting precision as candidate's individuality, produce k neural network individuality altogether.
Further, the individual following steps of carrying out of neural network after in above-mentioned steps C, utilizing preferably:
1) with k the neural network that trains the training sample data is predicted, is obtained mistake respectively! Can not pass through edit field code establishing object., mistake! Can not pass through edit field code establishing object., mistake! Can not pass through edit field code establishing object.；
2) foundation is based on the changeable weight model of fuzzy neural network, and it is e at t relative error absolute value constantly that three inputs are respectively i forecast model
i(t), i forecast model t constantly and preceding k-1 relative error absolute value constantly on average be designated as E
i(t), i forecast model is at the rate of change c of t relative error absolute value constantly
i(t), be output as e
i(t+1), to the model training;
3) use test data are imported as changeable weight, and prediction network individuality is in difference relative error absolute value mistake constantly! Can not pass through edit field code establishing object., by mistake! Can not pass through edit field code establishing object.The weight that just can calculate t moment i model is: mistake! Can not pass through edit field code establishing object., wherein, a mistake! Can not pass through edit field code establishing object.。
4) with m model training test data is predicted, is used the 3rd) weights that obtain of step are weighted integratedly with predicting the outcome, and finally neural network integrated prediction model prediction result is: mistake! Can not pass through edit field code establishing object.。
The present invention contrasts prior art and has following remarkable advantage:
1. solve the selected problem of the integrated individual amount of neural network by the distance cost function, solved the blindness problem that the preferred number of neural network is selected.
2. use the changeable weight model to carry out the prediction of weight, solved the drawback of fixed weight model, make it better to reflect the situation of change of predicted value under the dynamic environment.
Description of drawings
Fig. 1 is overall integrating process synoptic diagram of the present invention;
Fig. 2 is Bagging integrated technology synoptic diagram;
Fig. 3 is K-means clustering algorithm process flow diagram;
Fig. 4 is based on the individual preferred process synoptic diagram of neural network of K-means clustering algorithm;
Fig. 5 is structure of fuzzy neural network figure;
Fig. 6 is based on the changeable weight integrating process synoptic diagram of fuzzy neural network.
Embodiment
Below in conjunction with the description of drawings embodiments of the present invention.
Fig. 1 is overview flow chart of the present invention, and this localization method is divided into four-stage, specifically comprises:
A. generate the neural network individuality: choose the Elman neural network as the network individuality, produce n neural network individuality by the Bagging algorithm, for next step neural network individuality preferably provides accurate data;
B. individual preferred: to the neural network individuality that generates, at first optimize algorithm by the K value and determine the cluster number, by the K-means clustering algorithm neural network individuality is carried out cluster then and increase individual difference, the individuality after the cluster carries out integrated the most at last;
C. dynamic model is set up: utilize the error of preferred individual fitting data, make up the changeable weight model based on fuzzy neural network;
D. integrated output: will be preferred individual predict the outcome and weight that the changeable weight model generates combines to generate and finally predicts the outcome.
As shown in Figure 2, in above-mentioned steps A, in conjunction with the Elman neural network, and utilize the Bagging algorithm to carry out individual generation, specifically comprise following key step:
1) given original training set X=(x
1, y
1), (x
2, y
2), (x
3, y
3) ... (x
n, y
n), wherein n is the capacity of original training set, and the maximum iteration time of training is T, and the initial integrated E of neural network is empty set;
2) the initialization training set adopts m training sample of Bootstrap method for resampling extraction to form sub-training set X
t
3) by training set X
tTrain t neural network h
t
To t=1,2 ..., the T circulation:
(1) according to training set X
tThe individual h of neural network training
t
(2) with the neural network h that trains
t, list candidate's individuality in.
Further, in above-mentioned steps B, utilize the individual following steps of carrying out of the neural network that generates:
1) with the output result on training set of each neural network individuality
(wherein m is the number of training output sample point) forms matrix Y (y
1, y
2..., y
n), then matrix Y is carried out cluster analysis;
2) to k=1,2 ..., the T circulation:
(1) by the K-means clustering algorithm matrix Y is divided into the k group, the process flow diagram of K-means clustering algorithm as shown in Figure 3;
(2) mistake under the current k value of calculating! Can not pass through edit field code establishing object.；
3) search 1 to T between apart from cost minimum value mistake! Can not pass through edit field code establishing object.；
4) use k value to the cluster of carrying out of matrix Y, as candidate's individuality, produce k neural network individuality from the selection fitting precision is the highest every cluster data individuality altogether, whole individual preferred process as shown in Figure 4:
(s, k): suppose that n spatial object is divided into k bunch, cluster D is respectively in definition class border distance L and the class apart from cost function F
Wherein L represent all cluster centres to the universe center apart from sum, D represents the summation of clustering cluster inner distance, k is the cluster number, m is the average of whole samples, m
iBe a bunch C
iThe average of interior sample, p is arbitrary spatial object.Then apart from cost function F (s is that the distance B sum is namely in class border distance L and the class k):
(s, k) value are chosen and are made F (s, the k) k under the minimum can think and carry out the cluster result optimum by this k value this moment by the F of contrast under the different value of K.
Further, the individual following steps of carrying out of neural network after in above-mentioned steps C, utilizing preferably:
1) with k the neural network that trains the training sample data is predicted, is obtained mistake respectively! Can not pass through edit field code establishing object., mistake! Can not pass through edit field code establishing object., mistake! Can not pass through edit field code establishing object.；
2) foundation is based on the changeable weight model of fuzzy neural network, and Fig. 5 is the structural drawing of fuzzy neural network, and it is e at t relative error absolute value constantly that three inputs are respectively i forecast model
i(t), i forecast model t constantly and preceding k-1 relative error absolute value constantly on average be designated as E
i(t) (before the expression k constantly the whole estimated performance of model), i forecast model is at the rate of change c of the relative error absolute value in the t moment
i(t), be output as e
i(t+1), to model training, e
i(t), E
i(t), c
i(t) be expressed as respectively:
c
i(t)=| e
i(t)-e
i(t-1) |, Y (t) be t constantly (t=1,2 ..., n) Shi Ke actual value, f
i(t) be i(i=1,2 ..., m) the kind method is in t predicted value constantly.
3) use test data are imported as changeable weight, and prediction network individuality is in difference relative error absolute value mistake constantly! Can not pass through edit field code establishing object., by mistake! Can not pass through edit field code establishing object.The weight that just can calculate t moment i model is: mistake! Can not pass through edit field code establishing object.Wherein, a mistake! Can not pass through edit field code establishing object.；
4) with m the model that trains test data is predicted, use the 3rd) weights that obtain of step with predict the outcome be weighted integrated, Fig. 6 is the changeable weight integrating process synoptic diagram of fuzzy neural network, and final neural network integrated prediction model prediction result is: mistake! Can not pass through edit field code establishing object.。
Embodiment 1
The method of choose BP, Elman respectively in experiment, traditional neural network is integrated and the present invention proposes compares, here software reliability data Data13 tests before choosing, and choose preceding 130 groups as training data, the 40 groups of data in back are as verification msg, and last 43 groups of data are as test data.
In the experiment, produce 25 neural network individualities altogether and experimentize.For method proposed by the invention, the upper limit of K-means clustering algorithm k value is made as 5, calculating k then respectively is 1,2,3,4,5 o'clock F (s, value k), F (s, 1)=514.45, F (s, 2)=476.63, F (s, 3)=385.27, F (s, 4)=420.34, F (s, 5)=433.18.Select 3 individualities to carry out integrated thus.
Table 1 BP, Elman, NNE, KF-NNE under the Data13 data comparative analysis with actual value that predicts the outcome
Can find that by analyzing the predicted value of the integrated algorithm of selectivity dynamic neural network that the present invention proposes and the gap minimum between the actual value are also relatively stable.So the method that the present invention proposes has higher precision and stability.
Claims (1)
1. one kind based on the integrated software reliability prediction method of selectivity changeable weight neural network, it is characterized in that, may further comprise the steps:
A. generate the neural network individuality: choose the Elman neural network as the network individuality, produce n neural network individuality by the Bagging algorithm;
B. individual preferred: to the neural network individuality that generates, at first optimize algorithm by the K value and determine the cluster number, by the K-means clustering algorithm neural network individuality is carried out cluster then and increase individual difference, the individuality after the cluster carries out integrated the most at last;
C. dynamic model is set up: utilize the error of preferred individual fitting data, make up the changeable weight model based on fuzzy neural network;
D. integrated output: will be preferred individual predict the outcome and weight that the changeable weight model generates combines to generate and finally predicts the outcome;
Further, in above-mentioned steps A, in conjunction with the Elman neural network, and utilize the Bagging algorithm to carry out individual generation, specifically comprise following key step:
1) given original training set X=(x
1, y
1), (x
2, y
2), (x
3, y
3) ... (x
n, y
n), wherein n is the capacity of original training set, and the maximum iteration time of training is T, and the initial integrated E of neural network is empty set;
2) the initialization training set adopts m training sample of Bootstrap method for resampling extraction to form sub-training set X
t
3) by training set X
tTrain t neural network h
t
Further, in above-mentioned steps B, utilize the individual following steps of carrying out of the neural network that generates:
1) with the output result of each neural network individuality on training set
(wherein m is the number of training output sample point) forms matrix Y (y
1, y
2..., y
n), then matrix Y is carried out cluster analysis;
2) to k=1,2 ..., the T circulation:
(1) by the K-means clustering algorithm matrix Y is divided into the k group;
(2) mistake under the current k value of calculating! Can not pass through edit field code establishing object;
3) search 1 to T between apart from cost minimum value mistake! Can not pass through edit field code establishing object;
4) use the k value that matrix Y is carried out cluster, in every cluster data, select the highest individuality of fitting precision as candidate's individuality, produce k neural network individuality altogether;
Further, the individual following steps of carrying out of neural network after in above-mentioned steps C, utilizing preferably:
1) with k the neural network that trains the training sample data is predicted, is obtained mistake respectively! Can not pass through edit field code establishing object, mistake! Can not pass through edit field code establishing object, mistake! Can not pass through edit field code establishing object;
2) foundation is based on the changeable weight model of fuzzy neural network, and it is e at t relative error absolute value constantly that three inputs are respectively i forecast model
i(t), i forecast model t constantly and preceding k-1 relative error absolute value constantly on average be designated as E
i(t), i forecast model is at the rate of change c of t relative error absolute value constantly
i(t), be output as e
i(t+1), to the model training;
3) use test data are imported as changeable weight, and prediction network individuality is in difference relative error absolute value mistake constantly! Can not pass through edit field code establishing object., by mistake! Can not pass through edit field code establishing object.The weight that just can calculate t moment i model is: mistake! Can not pass through edit field code establishing object., wherein, a mistake! Can not pass through edit field code establishing object;
4) with m model training test data is predicted, is used the 3rd) weights that obtain of step are weighted integratedly with predicting the outcome, and finally neural network integrated prediction model prediction result is: mistake! Can not pass through edit field code establishing object.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310276933.9A CN103294601B (en) | 2013-07-03 | 2013-07-03 | A kind of based on the integrated software reliability prediction method of selective changeable weight neutral net |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310276933.9A CN103294601B (en) | 2013-07-03 | 2013-07-03 | A kind of based on the integrated software reliability prediction method of selective changeable weight neutral net |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103294601A true CN103294601A (en) | 2013-09-11 |
| CN103294601B CN103294601B (en) | 2016-05-04 |
Family
ID=49095519
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310276933.9A Expired - Fee Related CN103294601B (en) | 2013-07-03 | 2013-07-03 | A kind of based on the integrated software reliability prediction method of selective changeable weight neutral net |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103294601B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105260613A (en) * | 2015-10-26 | 2016-01-20 | 中国石油天然气股份有限公司 | Method for obtaining robustness of compressor unit |
| CN106651020A (en) * | 2016-12-16 | 2017-05-10 | 燕山大学 | Short-term power load prediction method based on big data reduction |
| CN108549597A (en) * | 2018-03-05 | 2018-09-18 | 浙江大学 | A kind of fuzzy detection seed set generation method and generator based on WGAN models |
| CN108563565A (en) * | 2018-04-08 | 2018-09-21 | 中国人民解放军海军工程大学 | Flight landing guides system reliability Quantitative Analysis Model method for building up |
| CN108985386A (en) * | 2018-08-07 | 2018-12-11 | 北京旷视科技有限公司 | Obtain method, image processing method and the corresponding intrument of image processing model |
| CN110322933A (en) * | 2019-06-20 | 2019-10-11 | 浙江工业大学 | A kind of polypropylene melt index hybrid modeling method based on dynamic error compensation mechanism |
| CN110888798A (en) * | 2019-10-14 | 2020-03-17 | 西安理工大学 | Software defect prediction method based on graph convolution neural network |
| CN111337833A (en) * | 2020-02-25 | 2020-06-26 | 北京航空航天大学 | Lithium battery capacity integrated prediction method based on dynamic time-varying weight |
| CN116319025A (en) * | 2023-03-22 | 2023-06-23 | 云南电网有限责任公司信息中心 | Zero-trust network trust evaluation method based on machine learning |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110288199A (en) * | 2019-05-29 | 2019-09-27 | 北京航空航天大学 | The method of product quality forecast |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040267499A1 (en) * | 2003-06-27 | 2004-12-30 | Barnes Arthur E. | Measuring discontinuity in seismic data |
| CN101008917A (en) * | 2007-01-29 | 2007-08-01 | 中南大学 | Fault diagnostic system of mobile robot software based on artificial immune system |
| CN102073586A (en) * | 2010-12-23 | 2011-05-25 | 北京航空航天大学 | Gray generalized regression neural network-based small sample software reliability prediction method |
-
2013
- 2013-07-03 CN CN201310276933.9A patent/CN103294601B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040267499A1 (en) * | 2003-06-27 | 2004-12-30 | Barnes Arthur E. | Measuring discontinuity in seismic data |
| CN101008917A (en) * | 2007-01-29 | 2007-08-01 | 中南大学 | Fault diagnostic system of mobile robot software based on artificial immune system |
| CN102073586A (en) * | 2010-12-23 | 2011-05-25 | 北京航空航天大学 | Gray generalized regression neural network-based small sample software reliability prediction method |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105260613B (en) * | 2015-10-26 | 2018-06-01 | 中国石油天然气股份有限公司 | Method for obtaining robustness of compressor unit |
| CN105260613A (en) * | 2015-10-26 | 2016-01-20 | 中国石油天然气股份有限公司 | Method for obtaining robustness of compressor unit |
| CN106651020B (en) * | 2016-12-16 | 2020-09-11 | 燕山大学 | Short-term power load prediction method based on big data reduction |
| CN106651020A (en) * | 2016-12-16 | 2017-05-10 | 燕山大学 | Short-term power load prediction method based on big data reduction |
| CN108549597A (en) * | 2018-03-05 | 2018-09-18 | 浙江大学 | A kind of fuzzy detection seed set generation method and generator based on WGAN models |
| CN108563565A (en) * | 2018-04-08 | 2018-09-21 | 中国人民解放军海军工程大学 | Flight landing guides system reliability Quantitative Analysis Model method for building up |
| CN108985386A (en) * | 2018-08-07 | 2018-12-11 | 北京旷视科技有限公司 | Obtain method, image processing method and the corresponding intrument of image processing model |
| CN110322933A (en) * | 2019-06-20 | 2019-10-11 | 浙江工业大学 | A kind of polypropylene melt index hybrid modeling method based on dynamic error compensation mechanism |
| CN110888798A (en) * | 2019-10-14 | 2020-03-17 | 西安理工大学 | Software defect prediction method based on graph convolution neural network |
| CN110888798B (en) * | 2019-10-14 | 2022-11-04 | 西安理工大学 | Software defect prediction method based on graph convolution neural network |
| CN111337833A (en) * | 2020-02-25 | 2020-06-26 | 北京航空航天大学 | Lithium battery capacity integrated prediction method based on dynamic time-varying weight |
| CN116319025A (en) * | 2023-03-22 | 2023-06-23 | 云南电网有限责任公司信息中心 | Zero-trust network trust evaluation method based on machine learning |
| CN116319025B (en) * | 2023-03-22 | 2024-01-26 | 云南电网有限责任公司信息中心 | Zero-trust network trust evaluation method based on machine learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103294601B (en) | 2016-05-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103294601A (en) | Software reliability forecasting method based on selective dynamic weight neural network integration | |
| US20210383041A1 (en) | In-situ thermodynamic model training | |
| Tian et al. | Data driven parallel prediction of building energy consumption using generative adversarial nets | |
| Khalil et al. | Machine Learning, Deep Learning and Statistical Analysis for forecasting building energy consumption—A systematic review | |
| Zhang et al. | Incorporating rich features into deep knowledge tracing | |
| CN109142171B (en) | Urban PM10 concentration prediction method based on feature expansion and fusing with neural network | |
| Hong et al. | SVR with hybrid chaotic genetic algorithms for tourism demand forecasting | |
| JP5888640B2 (en) | Photovoltaic power generation prediction apparatus, solar power generation prediction method, and solar power generation prediction program | |
| US11501153B2 (en) | Methods and apparatus for training a neural network | |
| Ma et al. | Deep-esn: A multiple projection-encoding hierarchical reservoir computing framework | |
| CN111723523B (en) | Estuary surplus water level prediction method based on cascade neural network | |
| CN111242351A (en) | Tropical cyclone track prediction method based on self-encoder and GRU neural network | |
| Alfred | A genetic-based backpropagation neural network for forecasting in time-series data | |
| CN105184398A (en) | Power maximum load small-sample prediction method | |
| CN104850677A (en) | Deep learning cellular automaton simulation analysis method based on geographic information system | |
| CN112634992A (en) | Molecular property prediction method, training method of model thereof, and related device and equipment | |
| CN109583588A (en) | A kind of short-term wind speed forecasting method and system | |
| CN110795522A (en) | Method and device for predicting track position of mobile user | |
| Chen et al. | Offline training for improving online performance of a genetic algorithm based optimization model for hourly multi-reservoir operation | |
| Adeli et al. | An advanced spatio-temporal convolutional recurrent neural network for storm surge predictions | |
| CN109408896B (en) | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production | |
| CN102663493A (en) | Delaying nerve network used for time sequence prediction | |
| CN112183721B (en) | Construction method of combined hydrological prediction model based on self-adaptive differential evolution | |
| JP2020091171A (en) | Weather forecasting system, weather forecasting method, and weather forecasting program | |
| Ajagunsegun et al. | Machine learning-based system for managing energy efficiency of public buildings: An approach towards smart cities |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20160504 Termination date: 20190703 |