CN106610980A - Equipment and method used for classifying/ predicting spatiotemporal sequence data - Google Patents
Equipment and method used for classifying/ predicting spatiotemporal sequence data Download PDFInfo
- Publication number
- CN106610980A CN106610980A CN201510690684.7A CN201510690684A CN106610980A CN 106610980 A CN106610980 A CN 106610980A CN 201510690684 A CN201510690684 A CN 201510690684A CN 106610980 A CN106610980 A CN 106610980A
- Authority
- CN
- China
- Prior art keywords
- space
- matrix
- serial data
- time
- weight
- 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
Classifications
-
- 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
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The invention provides equipment used for classifying/ predicting spatiotemporal sequence data. The equipment comprises a receiving unit, a modeling unit and a classification/ prediction unit, wherein the receiving unit is configured to receive the spatiotemporal sequence data; the modeling unit is configured to generate a weight parameter related to geographic heterogeneity on the basis of the spatiotemporal sequence data, and constructs a model used for classification/ prediction on the basis of the generated weight parameter; and the classification/ prediction unit is configured to carry out classification/ prediction on the spatiotemporal sequence data by the constructed model used for classification/ prediction. The invention also provides a method used for classifying/ predicting the spatiotemporal sequence data. A geographically weighted extreme learning machine which is put forward by the invention considers the geographic space heterogeneity, and accuracy for carrying out classification or prediction on the spatiotemporal sequence data can be improved.
Description
Technical field
The application is related to data analysis field, and in particular to one kind to Time-space serial data for carrying out point
The apparatus and method of class/prediction.
Background technology
Environment, meteorology, traffic, economic dispatch multiple fields have accumulated the Time-space serial data of magnanimity.When
Empty sequence generally describes a certain variable in different spatial over time.For example, Beijing
35 air monitering websites day PM2.5 concentration values of 24 hours.Time-space serial modeling is intended to description
Dependency relation between Temporal-Spatial Variables, and then some values by this relation pair Time-space serial are classified
Or future value is predicted.
Extreme learning machine (Extreme learning machines, ELM) is that a kind of new neural network is calculated
Method, was proposed earliest by HuangGuangbin in 2004.Compared with traditional neural network, ELM
Training speed it is fast, need Human disturbance less.Bibliography 1 (" Huang G B, Zhu Q Y, Siew
C K.Extreme learning machine：A new learning scheme of feedforward neural
Networks [J] .Proc.int.jointConf.neuralNetw, 2006,2：985--990. ") in describe
The basic procedure of ELM algorithms.
Although however, existing extreme learning machine can be modeled to Time-space serial, can not reflect
Space-time characterisation in data set, does not account for geographic isomerism, therefore the precision classified or predict
It is not good enough.
The content of the invention
The present invention proposes a kind of using geographical weighting extreme learning machine (Geographically Weighted
Extreme Leaning Machine, GWELM) technical scheme, consider geographical in modeling process
On isomerism, geographical position is converted into into local weighted coefficient such that it is able to when being preferably applied for
Null sequence data.
According to an aspect of the invention, there is provided it is a kind of be used to Time-space serial data are carried out classifying/it is pre-
The equipment of survey, including：Receiving unit, is configured to receive Time-space serial data；Modeling unit, quilt
It is configured to Time-space serial data and produces the weight parameter relevant with geographic isomerism, and is based on
Produced weight parameter builds the model for classifying/predicting；And classification/predicting unit, it is configured
It is using the constructed model for classifying/predicting Time-space serial data to be carried out to classify/predict.
In one embodiment, the modeling unit includes：Space length matrix builds subelement, quilt
It is configured to Time-space serial data and builds space length matrix；Spatial weight matrix builds subelement,
It is configured to select at least two different weight parameters respectively for each locus, to build phase
The Spatial weight matrix answered；Model exports weight computing subelement, is configured to based on corresponding space
Weight matrix is calculating corresponding model output weights；And select subelement, be configured to select with
The corresponding output weights of the weight parameter of minimum cost value are produced, as final model output weights.
In one embodiment, the space length matrix is Gaussian spatial distance matrix, the space
Weight matrix is Gaussian spatial weight matrix.
In one embodiment, described at least two different weight parameters be from minimum space distance to
Select in the scope of maximum space distance.Preferably, described at least two different weight parameters can
With from the scope center line Sexual behavior mode of minimum space distance to maximum space distance.Alternatively, it is described at least
Two different weight parameters can at random be selected from minimum space distance into the scope of maximum space distance
Select.
According to another aspect of the present invention, there is provided one kind is used to Time-space serial data are carried out classifying/
The method of prediction, including：Receive Time-space serial data；Based on the generation of Time-space serial data and geographically
The relevant weight parameter of isomerism；Mould for classifying/predicting is built based on produced weight parameter
Type；And using it is constructed Time-space serial data are carried out classifying for the model classifying/predict/it is pre-
Survey.
In one embodiment, space length matrix is built based on Time-space serial data.It is empty for each
Between position select at least two different weight parameters respectively, to build corresponding Spatial weight matrix.
Corresponding model output weights are calculated based on corresponding Spatial weight matrix.Then, select and produce
The corresponding output weights of the weight parameter of minimum cost value, as final model output weights.
In one embodiment, the space length matrix is Gaussian spatial distance matrix, the space
Weight matrix is Gaussian spatial weight matrix.
In one embodiment, described at least two different weight parameters be from minimum space distance to
Select in the scope of maximum space distance.Preferably, described at least two different weight parameters can
With from the scope center line Sexual behavior mode of minimum space distance to maximum space distance.Alternatively, it is described at least
Two different weight parameters can at random be selected from minimum space distance into the scope of maximum space distance
Select.
Geographical weighting extreme learning machine proposed by the present invention considers the isomerism of geographical space, Neng Gouti
The precision that height is classified to Time-space serial data or predicted.
Description of the drawings
Will be become more by the above and other feature below in conjunction with the detailed description of accompanying drawing, the present invention
Plus substantially, wherein：
Fig. 1 show it is according to an embodiment of the invention for carrying out classifying to Time-space serial data/
The block diagram of the equipment of prediction.
Fig. 2 shows the form of Time-space serial data according to an embodiment of the invention.
Fig. 3 shows the block diagram of the modeling unit shown in Fig. 1.
Fig. 4 shows the schematic diagram of Time-space serial data according to an embodiment of the invention.
Fig. 5 shows the schematic diagram of model according to an embodiment of the invention.
Fig. 6 shows the schematic diagram of Time-space serial data according to an embodiment of the invention.
Fig. 7 shows the schematic diagram of structure space length matrix according to an embodiment of the invention.
Fig. 8 shows the schematic diagram of model according to an embodiment of the invention.
Fig. 9 shows the schematic diagram of model according to an embodiment of the invention.
Figure 10 shows the schematic diagram of model according to an embodiment of the invention.
Figure 11 shows in accordance with another embodiment of the present invention for carrying out to Time-space serial data point
The flow chart of the method for class/prediction.
Specific embodiment
Below, by combine accompanying drawing to the present invention specific embodiment description, the present invention principle and
Realization will become obvious.It should be noted that the present invention should not be limited to concrete reality hereinafter described
Apply example.In addition, for simplicity eliminating the detailed description of known technology unrelated to the invention.
Fig. 1 show it is according to an embodiment of the invention for carrying out classifying to Time-space serial data/
The block diagram of the equipment 10 of prediction.As shown in figure 1, equipment 10 includes receiving unit 110, modeling unit
120 and classification/predicting unit 130.
Receiving unit 110 receives Time-space serial data.In this application, Time-space serial data refer to one
Data of the group with time and space characteristics.Can be by Time-space serial data definition：
Z={ zs(t), s ∈ S, t ∈ T },
Wherein, s representation spaces position, t represents time, zsT () represents the category of t locus s
Property value, S represents the area of space involved by Time-space serial, and T represents the time model involved by Time-space serial
Enclose.Fig. 2 shows the form of Time-space serial data according to an embodiment of the invention, wherein,
With the Grid square of uniform sampling to be illustrated Time-space serial data.For example, as shown in Figure 2,
As s=, (2,4) and during t=1, the value of z=10, i.e. variable z on the column position of the 2nd row of t=1 moment the 4th is
10.Therefore, Time-space serial data are the time that all space cells build in 1 to T moment property value
Arrangement set.
Modeling unit 120 produces the weight ginseng relevant with geographic isomerism based on Time-space serial data
Number, and the model for classifying/predicting is built based on produced weight parameter.Below, with reference to Fig. 3
To describe the operation of modeling unit 120 in detail.
Fig. 3 shows the block diagram of the modeling unit 120 shown in Fig. 1.As shown in figure 3, modeling
Unit 120 includes that space length matrix builds subelement 1210, Spatial weight matrix and builds subelement
1220th, model output weight computing subelement 1230 and selection subelement 1240.
Space length matrix builds subelement 1210 and builds space length matrix based on Time-space serial data
M.Refer to the attached drawing 4, illustrated therein is the Time-space serial data that number is n=6, it is assumed that each time sequence
The length of row is k.Correspondingly, space length matrix M is：
Wherein,Represent the distance between i and j points
Spatial weight matrix builds subelement 1220 and selects at least two respectively not for each locus
Same weight parameter, to build corresponding Spatial weight matrix.For example, Spatial weight matrix builds son
Unit 1220 can be directed to locus i, build the Gaussian spatial weight under different weight parameters c
Matrix W：
Wherein, wij=exp (- (dij/c))
c∈{c1, c2..., cm, the span of c is in minimum dijWith maximum dijBetween.
Because each locus has carried out k observation, subsequently calculate for convenience, can be further
Define blending space weight matrix
Preferably, above-mentioned space length matrix can be Gaussian spatial distance matrix, above-mentioned space weight
Matrix can be Gaussian spatial weight matrix.
Model output weight computing subelement 1230 calculates corresponding based on corresponding Spatial weight matrix
Model exports weights.First, model output weight computing subelement 1230 obtains the input weights of model
With biasing, these values can be random given.Fig. 5 is showed according to one embodiment of the invention
Model schematic diagram.From figure 5 it can be seen that the input weights of model include a1、a2、…a1,
The biasing of model includes b1、b2、…b1.May be referred to traditional ELM algorithms to obtain above-mentioned input power
Value and biasing.In addition, model output weight computing subelement 1230 calculates hidden layer input matrix H：
Hnk×lβl=onk
On this basis, model exports adding for the output weights of the solving model of weight computing subelement 1230
Power least squares norm solutionMinimize following cost function：
Obtain corresponding so as to solve
Wherein, Z is actual value, and o is model output valve.
Subelement 1240 is selected to select the output power corresponding with the weight parameter for producing minimum cost value
Value, as final model output weights.That is, subelement 1240 is selected to select corresponding with optimal fitting
Weight parameter c corresponding to model, as the final mask of locus i.For example, for space
Position i, calculates different weight parameters c1, c2..., cmIn corresponding minimum CV values：
And select final model with this.
Return to Fig. 1, classification/predicting unit 130 using the model constructed by modeling unit 120, pair when
Null sequence data carries out classifying/predicting.Finally, output category or the result of prediction, such as future time instance
Air quality, weather condition etc..
Below, with reference to accompanying drawing 6-10 describing a concrete application example of the equipment 10 shown in Fig. 1.
Fig. 6 shows the schematic diagram of Time-space serial data according to an embodiment of the invention.For example,
The Time-space serial data can be：
That is, spatially there are six positions of p1 to p6, each position one length of correspondence is 6 time
Sequence (T=1-6).There are two characteristic vectors x1, x2 each position, it is assumed here that they are to all
The value of position is identical：
Therefore, space length matrix builds subelement 1210 and can build according to the mode shown in Fig. 7
Space length matrix M：
Below with locus i=3 (p3) as a example by, illustrate how to set up GWELM models.
For locus p3, weight parameter c chooses respectively different c1, c2..., cm.Preferably,
Weight parameter is in the scope center line Sexual behavior mode from minimum range to ultimate range：
cmin=min (dij), cmax=max (dij) (i=1 ..., n；J=1 ..., n；i≠j)
Wherein, m represents the number of weight parameter.The advantage of the selection strategy can be to obtain quickly
Optimal solution.
It is alternatively possible to select weight parameter c using other strategies.For example, can be according to random
Mode selects weight parameter c from the scope of minimum range to ultimate range.
Assume m=6, then corresponding weight parameter c is calculated as follows：
cmin=1.4, cmax=3.6
C1=1.40；C2=1.84；C3=2.28；C4=2.72；C5=3.16；C6=3.60
Spatial weight matrix builds subelement 1220 and is directed to locus i=3, builds different weight parameters
Under Gaussian spatial weight matrix：
WI, c=diag (wi1, wi2..., wi6)
Wherein, wij=exp (- (dij/c)2)
As c1=1.40, W3,1.40=diag (0.08,0.36,1,0.13,0.36,0.36)
As c2=1.84, W3,1.84=diag (0.23,0.55,1,0.31,0.55,0.55)
As c3=2.28, W3,2.28=diag (0.38,0.68,1,0.46,0.68,0.68)
As c4=2.72, W3,2.72=diag (0.51,0.76,1,0.58,0.76,0.76)
As c5=3.16, W3,3.16=diag (0.61,0.82,1,0.67,0.82,0.82)
As c6=3.60, W3,3.60=diag (0.68,0.86,1,0.73,0.86,0.86)
Fig. 8 shows the schematic diagram of model according to an embodiment of the invention, wherein, give at random
Go out input weights and the biasing of model：
a1=[0.8147；0.9058]
a2=[0.1270；0.9134]
a3=[0.6324；0.0975]
B=[0.2785；0.5469；0.9575]
On this basis, model output weight computing subelement 1230 calculates hidden layer input matrix H：
Result of calculation is
Due to the variable x all sames of each location point, therefore
Then, model output weight computing subelement 1230 solves the defeated of the model under different weight parameters
Go out weights：
As c1=1.40,
Cv=3.96
As c2=1.84,
Cv=4.33
As c3=2.28,
Cv=4.59
As c4=2.72,
Cv=4.77
As c5=3.16,
Cv=4.89
As c6=3.60,
Cv=4.97
It can be seen that, for locus p3, when c takes 1.40, corresponding CV values are minimum.Therefore, select
Select the output weight that subelement 1240 selects the estimate corresponding with minimum CV as forecast model：
Thus, it is possible to obtain as shown in Figure 9 for locus p3GWELM models.
For locus p3, when needing to predict the value at t=7 moment, it is necessary first to obtain variable
Values of the x1 and x2 at the t=7 moment.Assume x1 (7)=4, x2 (7)=6, in substituting into the model of Fig. 9,
Obtain final predicted valueAs shown in Figure 10.
It can be seen that, the geographical weighting extreme learning machine model proposed using the above embodiment of the present invention, energy
Enough improve the precision classified to Time-space serial data or predicted.
Figure 11 shows in accordance with another embodiment of the present invention for carrying out to Time-space serial data point
The flow chart of the method for class/prediction.As shown in figure 11, method 1100 starts at step S1110.
In step S1120, Time-space serial data are received.
In step S1130, based on Time-space serial data the weight ginseng relevant with geographic isomerism is produced
Number.Then, in step S1340, the mould for classifying/predicting is built based on produced weight parameter
Type.It is preferably based on Time-space serial data and builds space length matrix, and for each locus
At least two different weight parameters are selected respectively, to build corresponding Spatial weight matrix.Then,
Corresponding model output weights are calculated based on corresponding Spatial weight matrix.Finally, select and produce
The corresponding output weights of the weight parameter of minimum cost value, as final model output weights.
Above-mentioned at least two different weight parameters can be from minimum space distance to maximum space distance
Select in scope.Preferably, above-mentioned at least two different weight parameters can be from minimum space distance
To the scope center line Sexual behavior mode of maximum space distance.Alternatively, above-mentioned at least two different weight ginseng
Number can be minimum space distance to maximum space distance scope in randomly choose.
In step S1150, Time-space serial data are entered using the constructed model for classifying/predicting
Row classification/prediction.
Finally, method 1100 terminates at step S1160.
It should be understood that the above embodiment of the present invention can pass through software, hardware or software and hardware
Both is implemented in combination in.For example, the various assemblies in the system in above-described embodiment can pass through many
Plant device to realize, these devices are included but is not limited to：Analog circuit, digital circuit, general procedure
Device, Digital Signal Processing (DSP) circuit, programmable processor, special IC (ASIC),
Field programmable gate array (FPGA), PLD (CPLD), etc..
In addition, it will be understood to those skilled in the art that the initial parameter described in the embodiment of the present invention
Can store in the local database, it is also possible to be stored in distributed data base or can be stored in
In remote data base.
Additionally, embodiments of the invention disclosed herein can be realized on computer program.
More specifically, the computer program is a kind of following product：With computer-readable medium,
Coding has computer program logic on computer-readable medium, when performing on the computing device, the meter
Calculate machine program logic to provide related operation to realize the above-mentioned technical proposal of the present invention.When in calculating system
When performing at least one processor of system, computer program logic causes the computing device present invention real
Apply the operation (method) described in example.This set of the present invention is typically provided as arranging or encoding in example
Software, code such as on the computer-readable medium of optical medium (such as CD-ROM), floppy disk or hard disk
And/or consolidating on other data structures or such as one or more ROM or RAM or PROM chips
It is Downloadable software image in other media or one or more modules of part or microcode, shared
Database.Software or firmware or this configuration may be installed in computing device, so that computing device
In one or more processors perform the embodiment of the present invention described by technical scheme.
Although showing the present invention already in connection with the preferred embodiments of the present invention above, this area
Technical staff will be understood that, without departing from the spirit and scope of the present invention, can be to this
It is bright to carry out various modifications, replacement and change.Therefore, the present invention should not be limited by above-described embodiment,
And should be limited by claims and its equivalent.
Claims (12)
1. a kind of equipment for carrying out classifying/predicting to Time-space serial data, including：
Receiving unit, is configured to receive Time-space serial data；
Modeling unit, is configured to relevant with geographic isomerism based on the generation of Time-space serial data
Weight parameter, and the model for classifying/predicting is built based on produced weight parameter；And
Classification/predicting unit, be configured to using it is constructed for classify/predict model pair when
Null sequence data carries out classifying/predicting.
2. equipment according to claim 1, wherein, the modeling unit includes：
Space length matrix build subelement, be configured to based on Time-space serial data build space away from
From matrix；
Spatial weight matrix build subelement, be configured to for each locus select respectively to
Few two different weight parameters, to build corresponding Spatial weight matrix；
Model exports weight computing subelement, is configured to be counted based on corresponding Spatial weight matrix
Calculate corresponding model output weights；And
Subelement is selected, the weight parameter for being configured to select with produce minimum cost value is corresponding
Output weights, as final model output weights.
3. equipment according to claim 2, wherein, the space length matrix is that Gauss is empty
Between distance matrix, the Spatial weight matrix is Gaussian spatial weight matrix.
4. equipment according to claim 2, wherein, described at least two different weights ginsengs
Number is selected into the scope of maximum space distance from minimum space distance.
5. equipment according to claim 4, wherein, described at least two different weights ginsengs
Number is the scope center line Sexual behavior mode from minimum space distance to maximum space distance.
6. equipment according to claim 4, wherein, described at least two different weights ginsengs
Number is from minimum space distance to randomly selected in the scope of maximum space distance.
7. a kind of method for carrying out classifying/predicting to Time-space serial data, including：
Receive Time-space serial data；
The weight parameter relevant with geographic isomerism is produced based on Time-space serial data；
Model for classifying/predicting is built based on produced weight parameter；And
Using it is constructed Time-space serial data are carried out classifying for the model classifying/predict/it is pre-
Survey.
8. method according to claim 7, wherein,
Space length matrix is built based on Time-space serial data；
At least two different weight parameters are selected respectively for each locus, it is corresponding to build
Spatial weight matrix；
Corresponding model output weights are calculated based on corresponding Spatial weight matrix；And
The output weights corresponding with the weight parameter for producing minimum cost value are selected, as final
Model exports weights.
9. method according to claim 8, wherein, the space length matrix is that Gauss is empty
Between distance matrix, the Spatial weight matrix is Gaussian spatial weight matrix.
10. method according to claim 8, wherein, described at least two different weights
Parameter is selected into the scope of maximum space distance from minimum space distance.
11. methods according to claim 10, wherein, described at least two different weights
Parameter is the scope center line Sexual behavior mode from minimum space distance to maximum space distance.
12. methods according to claim 10, wherein, described at least two different weights
Parameter is from minimum space distance to randomly selected in the scope of maximum space distance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510690684.7A CN106610980B (en) | 2015-10-22 | 2015-10-22 | Apparatus and method for classifying/predicting spatio-temporal sequence data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510690684.7A CN106610980B (en) | 2015-10-22 | 2015-10-22 | Apparatus and method for classifying/predicting spatio-temporal sequence data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106610980A true CN106610980A (en) | 2017-05-03 |
| CN106610980B CN106610980B (en) | 2022-03-18 |
Family
ID=58610345
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510690684.7A Active CN106610980B (en) | 2015-10-22 | 2015-10-22 | Apparatus and method for classifying/predicting spatio-temporal sequence data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106610980B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101127029A (en) * | 2007-08-24 | 2008-02-20 | 复旦大学 | Method for training SVM classifier in large scale data classification |
| CN102110365A (en) * | 2009-12-28 | 2011-06-29 | 日电（中国）有限公司 | Road condition prediction method and road condition prediction system based on space-time relationship |
| JP2014041547A (en) * | 2012-08-23 | 2014-03-06 | Nippon Telegr & Teleph Corp <Ntt> | Time series data analysis device, method and program |
| CN103700255A (en) * | 2013-12-30 | 2014-04-02 | 复旦大学 | Time and space related data mining-based traffic flow prediction method |
-
2015
- 2015-10-22 CN CN201510690684.7A patent/CN106610980B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101127029A (en) * | 2007-08-24 | 2008-02-20 | 复旦大学 | Method for training SVM classifier in large scale data classification |
| CN102110365A (en) * | 2009-12-28 | 2011-06-29 | 日电（中国）有限公司 | Road condition prediction method and road condition prediction system based on space-time relationship |
| JP2014041547A (en) * | 2012-08-23 | 2014-03-06 | Nippon Telegr & Teleph Corp <Ntt> | Time series data analysis device, method and program |
| CN103700255A (en) * | 2013-12-30 | 2014-04-02 | 复旦大学 | Time and space related data mining-based traffic flow prediction method |
Non-Patent Citations (2)
| Title |
|---|
| TORSTEN HOTHORN等: "Decomposing environmental,spatial,and spatiotemporal components of species distributions", 《ECOLOGICAL MONOGRAPHS》 * |
| 郭龙等: "基于空间自回归和地理加权回归模型的不同尺度下土地利用程度研究", 《华中农业大学学报》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106610980B (en) | 2022-03-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106485353B (en) | Air pollutant concentration forecasting procedure and system | |
| CN109214581A (en) | A kind of Along Railway wind speed forecasting method considering wind direction and confidence interval | |
| CN110519128A (en) | A kind of operating system recognition methods based on random forest | |
| CN110427654A (en) | A kind of predictive model of landslide construction method and system based on sensitiveness | |
| CN107515842B (en) | A kind of urban population density dynamic prediction method and system | |
| CN107392389A (en) | Taxi dispatching processing method based on ARIMA models | |
| CN107301475A (en) | Load forecast optimization method based on continuous power analysis of spectrum | |
| Kaveh et al. | Ground motion record selection using multi-objective optimization algorithms: A comparative study | |
| CN109948791A (en) | Utilize the method for genetic algorithm optimization BP neural network and its application in positioning | |
| CN112819237B (en) | Precipitation multi-scale fusion forecasting method and device | |
| Yang et al. | Prediction of significant wave heights based on CS-BP model in the south China sea | |
| CN105956252A (en) | Generative deep belief network-based multi-scale forecast modeling method for ultra-short-term wind speed | |
| CN103942251A (en) | Method and system for inputting high altitude meteorological data into database based on multiple quality control methods | |
| CN115032720A (en) | Application of multi-mode integrated forecast based on random forest in ground air temperature forecast | |
| CN109543879A (en) | Load forecasting method and device neural network based | |
| CN109116300A (en) | A kind of limit learning position method based on non-abundant finger print information | |
| CN106503793B (en) | A kind of neural network short-term wind speed forecasting method based on improvement difference algorithm | |
| CN110570046B (en) | Site selection method and system for environment monitoring site | |
| CN112307672A (en) | BP neural network short-term wind power prediction method based on cuckoo algorithm optimization | |
| CN108334957A (en) | Power grid primary equipment O＆M cost of overhaul prediction technique and system | |
| CN109886316B (en) | Transformer state parameter combination prediction method based on cloud system similarity weight distribution | |
| CN106610980A (en) | Equipment and method used for classifying/ predicting spatiotemporal sequence data | |
| CN106652443B (en) | Method for predicting short-time traffic volume of expressway with similar longitudinal and transverse dimensions | |
| CN108509537A (en) | A kind of transmission line galloping probability forecast system and method | |
| Michaelides et al. | Investigation of trends in synoptic patterns over Europe with artificial neural networks |
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 |