CN110059144A - A kind of track owner's prediction technique based on convolutional neural networks - Google Patents
A kind of track owner's prediction technique based on convolutional neural networks Download PDFInfo
- Publication number
- CN110059144A CN110059144A CN201910266737.0A CN201910266737A CN110059144A CN 110059144 A CN110059144 A CN 110059144A CN 201910266737 A CN201910266737 A CN 201910266737A CN 110059144 A CN110059144 A CN 110059144A
- Authority
- CN
- China
- Prior art keywords
- track
- user
- owner
- vec
- latitude
- 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Remote Sensing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Track owner's prediction technique based on convolutional neural networks that the invention discloses a kind of, according to the track of all users formed an oriented no weight graph G=<V, E>, pass through Node2Vec learn track position ID low-dimensional real-valued vectors；Then slicing treatment is carried out to user trajectory, position ID is replaced with position ID corresponding low-dimensional real-valued vectors to the random length track after cutting, and form the fixed dimension matrix of track by interception or filling；Then, one four layers of convolutional neural networks of building and training are as prediction model, then by the track Input matrix of user location longitude and latitude to be detected building into trained prediction model, the probability distribution of track owner classification is obtained, the index of maximum value in probability distribution is finally corresponded to the number of owner labeled as the track.
Description
Technical field
The invention belongs to machine learning techniques fields, more specifically, are related to a kind of rail based on convolutional neural networks
Mark owner's prediction technique.
Background technique
Track owner prediction carries out feature extraction and analysis by the track to some unknown owner, then judges this rail
The owner of mark.Track owner prediction is the basis of many location based services, for improving the quality of location based service
It is of great significance, ISP can use prediction result and carry out personalized recommendation and the path planning based on preference etc..
Existing track owner prediction technique is usually using track as time Series Processing, then using modes such as RNN
The expression of learning time sequence.Although this mode has learnt the context of track, but in a track sets, it may
Some specific position or certain several specific position combination are most important for the classification of track, and existing method but can not be effective
These features of track are captured, and convolutional neural networks can preferably learn to these features, so it is proposed that a kind of
Track owner's prediction technique based on convolutional neural networks.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of track owner based on convolutional neural networks
Prediction technique, by improving track modeling and feature extracting method, the accuracy rate of Lai Tigao trajectory predictions.
For achieving the above object, the present invention proposes a kind of track owner's prediction technique based on convolutional neural networks,
Characterized by comprising the following steps:
(1), data prediction
(1.1), the historical position longitude and latitude of all user trajectories is counted sequentially in time, forms longitude and latitude set,
Wherein, if a certain longitude and latitude repeats, only retain in longitude and latitude set primary；
Each of pair warp and weft degree set longitude and latitude is numbered from 1, gives only one position ID mark；
(1.2), in chronological order, substitution user trajectory historical position longitude and latitude is removed with the position ID given in step (1.1)
Degree, then each user trajectory is indicated with a string of position ID；Meanwhile being identified each user with unique integer ID, thus
User and user trajectory can form the owner track that form is [User ID, track (position ID ..., position ID)]；
(1.3), according to owner track formed an oriented no weight graph G=<V, E>, wherein V is the collection of all position ID
It closes, if there is some user from IDiTo IDj, then < IDi, IDj> indicating a directed edge, all such sides constitute figure G's
Side collection E；
(2), track indicates
(2.1), using the oriented no weight graph G of building as input, each position in G out is learnt by Node2Vec algorithm
The low-dimensional real-valued vectors of ID；
(2.2), the track of each user in step (1.2) is sliced according to Fixed Time Interval, thus by each
A user trajectory is divided into several position ID sequences, then carries out owner's mark to the position ID sequence after cutting with User ID；
(2.3), to the random length track after cutting, position ID is replaced with the corresponding low-dimensional real-valued vectors of position ID,
To generate the track matrix of each user；
Then by interception or filling mode, the track matrix of each user's fixed dimension is constructed, so that data set is formed,
Wherein, the vector of filling is the average value of the corresponding low-dimensional real-valued vectors of all position ID；
(3), prediction model is constructed
Four layers of convolutional neural networks are constructed, input layer is the track matrix of fixed dimension；Convolutional layer is arranged three
Convolution kernel m*embedding_size, wherein m is constant, embedding size be Node2vec output low-dimensional real value to
The dimension of amount；Pond layer is k-max pooling, and k indicates the preceding k maximum value after convolution；The output input of full articulamentum
To softmax function, the probability distribution of track owner is obtained；
(4), training prediction model
(4.1), training set is constructed
By the track matrix stack X of data set partial user fixed dimension and corresponding one-hot categorization vector collection Y
As training set, wherein X=[Vec_x1,Vec_x2,...,Vec_xn], Vec_xnIndicate the track of nth user's fixed dimension
Matrix, Y=[Vec_t1,Vec_t2,...,Vec_tn], Vec_tnIndicate the corresponding one-hot categorization vector of n-th of owner, if
Nth position is 1, then remaining position is all 0；
(4.2), prediction model is initialized
Initialize the weight matrix W of each convolution kernel in convolutional layerpValue be normal distribution, mean value 0, variance is
0.1；The bias vector B of each convolution kernel in convolutional layer is initialized simultaneouslypIt is 0.1, the element number of each bias vector is
The number of respective layer neuron；The weight matrix for initializing full articulamentum is W, and dimension is [batch_size*k* convolution kernel
Number, classification number], while the bias vector B value for initializing full articulamentum is 0.1, element number is classification number；Wherein, batch_
Size is constant, and p=1,2,3, p indicate which convolution kernel in convolutional layer；
(4.3) training set is input in the prediction model after initialization, using Adam algorithm optimization loss function, then
Error is transmitted to preceding layer using error back propagation BP algorithm, updates the weight matrix W of convolutional layerp, bias vector BpAnd it is complete
Articulamentum weight matrix W, bias vector B obtain convergent neural network model after iteration several times, to be instructed
Practice the prediction model completed；
(5), track owner predicts
By user location longitude and latitude to be detected according to step (1), (2) the method, user's fixed dimension is constructed
Track matrix it is corresponding all to obtain the track then by the track Input matrix of building into trained prediction model
The probability distribution of owner's classification, the index of maximum value is then the corresponding owner's number in the track in probability distribution.
Goal of the invention of the invention is achieved in that
The present invention is based on track owner's prediction techniques of convolutional neural networks, and forming one according to the track of all users has
To no weight graph G=<V, E>, pass through the low-dimensional real-valued vectors that Node2Vec learns track position ID；Then user trajectory is carried out
Slicing treatment replaces position ID with the corresponding low-dimensional real-valued vectors of position ID to the random length track after cutting, and passes through interception
Or filling forms the fixed dimension matrix of track；Then, it constructs and one four layers of convolutional neural networks of training is as prediction model,
Then the track Input matrix of user location longitude and latitude to be detected building is obtained into rail into trained prediction model
The index of maximum value in probability distribution is finally corresponded to the number of owner by the probability distribution of mark owner classification labeled as the track.
Meanwhile the present invention is based on track owner's prediction techniques of convolutional neural networks also to have the advantages that
(1), owner's track sets are built into network, i.e., oriented no weight graph, by Node2Vec algorithm learning network
The low-dimensional real-valued vectors of each node；
(2), the vector of track filling for institute's directed quantity average value, compared to being stuffed entirely with being 0, owner's predictablity rate
There is very big promotion；
(3), it is improved relative to traditional convolutional neural networks, the accuracy rate for predicting track owner, which has, further to be mentioned
It is high.
Detailed description of the invention
Fig. 1 is track owner's prediction technique flow chart the present invention is based on convolutional neural networks；
Fig. 2 is the configuration diagram of prediction model.
Specific embodiment
A specific embodiment of the invention is described with reference to the accompanying drawing, preferably so as to those skilled in the art
Understand the present invention.Requiring particular attention is that in the following description, when known function and the detailed description of design perhaps
When can desalinate main contents of the invention, these descriptions will be ignored herein.
Embodiment
Fig. 1 is track owner's prediction technique flow chart the present invention is based on convolutional neural networks.
In the present embodiment, as shown in Figure 1, a kind of track owner's prediction technique based on convolutional neural networks of the present invention,
The following steps are included:
S1, data prediction
S1.1, the historical position longitude and latitude of all user trajectories is counted sequentially in time, forms longitude and latitude set,
In, if a certain longitude and latitude repeats, only retain in longitude and latitude set primary；
As shown in table 1, each of pair warp and weft degree set longitude and latitude is numbered from 1, gives only one position ID
Mark；
Table 1 is the position ID mark table of longitude and latitude；
| Number | Longitude and latitude |
| 1 | 39.747652-104.99251 |
| 2 | 39.891383-105.070814 |
| 3 | 39.891077-105.068532 |
| 4 | 39.750469-104.999073 |
| …… | …… |
Table 1
S1.2, in chronological order removes substitution user trajectory historical position longitude and latitude with the position ID given in step S1.1,
Then each user trajectory is indicated with a string of position ID；Meanwhile being identified each user with unique integer ID, thus user
And user trajectory can form the owner track that form is [User ID, track (position ID ..., position ID)]；In this implementation
In example, such as a certain owner track: [0, (622,474,474,474,481,482,482,83,83,270,487,270,270,83,
83,471 ... ...)], wherein 0 is User ID, and subsequent is position ID sequence；
S1.3, according to owner track formed an oriented no weight graph G=<V, E>, wherein V is the set of all position ID,
If there is some user from IDiTo IDj, then < IDi, IDj> indicate a directed edge, it is all such in the while collection for constituting figure G
E；
S2, track indicate
S2.1, using the oriented no weight graph G of building as input, each position in G out is learnt by Node2Vec algorithm
The low-dimensional real-valued vectors of ID；Node2Vec is a kind of network representation learning method, and detailed process belongs to the prior art, herein
Just repeat no more.
In the present embodiment, the low-dimensional real-valued vectors for the position ID for being 60 such as number indicate are as follows:
[60 (- 0.383389, -0.826315, -1.379363 ... ..., -1.839076,1.930556,0.502587)]；
S2.2, the track of user each in step S1.2 is sliced according to Fixed Time Interval, thus by each
User trajectory is divided into several position ID sequences, then carries out owner's mark to the position ID sequence after cutting with User ID；?
In the present embodiment, dicing effect is as follows:
[0, (622,474,474,474,481,482,482,83)]
[0, (83,270,487,270,270,83,83,471)]
……
Wherein, 0 is User ID, and subsequent is position ID sequence；
S2.3, position ID is replaced with position ID corresponding low-dimensional real-valued vectors to the random length track after cutting, from
And generate the track matrix of each user；
Then by interception or filling mode, the track matrix of each user's fixed dimension is constructed, wherein the vector of filling
For the average value of the corresponding low latitude real-valued vectors of all position ID；In the present embodiment, it is every to construct that preceding 30 position ID are intercepted
The track matrix of a user's fixed dimension；
S3, building prediction model
Four layers of convolutional neural networks are constructed, as shown in Fig. 2, its input layer is the track matrix of fixed dimension；Convolution
Three convolution kernel m*embedding_size of layer setting, wherein the value of m is 2,3,4, embedding size be Node2vec
The dimension of the low latitude real-valued vectors of output, the number of every kind of convolution kernel are 64；Pond layer is k-max pooling, and k indicates convolution
Preceding k maximum value later, in the present embodiment k=3；The output of full articulamentum is input to softmax function, obtains track category
Main probability distribution.
S4, training prediction model
S4.1, building training set
By the track matrix stack X of data set partial user fixed dimension and corresponding one-hot categorization vector collection Y
As training set, wherein X=[Vec_x1,Vec_x2,...,Vec_xn], Vec_xnIndicate the track of nth user's fixed dimension
Matrix, Y=[Vec_t1,Vec_t2,...,Vec_tn], Vec_tnIndicate the corresponding one-hot categorization vector of n-th of owner, if
Nth position is 1, then remaining position is all 0；
S4.2, initialization prediction model
Initialize the weight matrix W of each convolution kernel in convolutional layerpValue be positive and be distributed very much, mean value 0, variance is
0.1；The bias vector B of each convolution kernel in convolutional layer is initialized simultaneouslypIt is 0.1, the element number of each bias vector is
The number of respective layer neuron；The weight matrix for initializing full articulamentum is W, and dimension is [batch_size*k* convolution kernel
Number, classification number], while the bias vector B value for initializing full articulamentum is 0.1, element number is classification number；Wherein, batch_
Size is constant, and value 64, p=1,2,3, p indicates which convolution kernel in convolutional layer；
Training set is input in the prediction model after initialization by S4.3, using Adam algorithm optimization loss function, and it is excellent
The loss function of change is as follows:
Wherein, N=batch_size, yjFor Vec_tjIn all elements, ajFor the output valve of softmax function；
Then error is transmitted to preceding layer using error back propagation BP algorithm, updates the weight matrix W of convolutional layerp, partially
Set vector BpAnd full articulamentum weight matrix W, bias vector B obtain convergent neural network after iteration several times
Model, to obtain the prediction model of training completion；
S5, track owner prediction
By user location longitude and latitude to be detected according to step S1, S2 the method, user's fixed dimension is constructed
Track matrix, then by the track Input matrix of building into trained prediction model, obtain the track and correspond to all categories
The probability distribution of major category, the index of maximum value is then the corresponding owner's number in the track in probability distribution.
Although the illustrative specific embodiment of the present invention is described above, in order to the technology of the art
Personnel understand the present invention, it should be apparent that the present invention is not limited to the range of specific embodiment, to the common skill of the art
For art personnel, if various change the attached claims limit and determine the spirit and scope of the present invention in, these
Variation is it will be apparent that all utilize the innovation and creation of present inventive concept in the column of protection.
Claims (2)
1. a kind of track owner's prediction technique based on convolutional neural networks, which comprises the following steps:
(1), data prediction
(1.1), the historical position longitude and latitude of all user trajectories is counted sequentially in time, forms longitude and latitude set, wherein
If a certain longitude and latitude repeats, only retain in longitude and latitude set primary；
Each of pair warp and weft degree set longitude and latitude is numbered from 1, gives only one position ID mark；
(1.2), substitution user trajectory historical position longitude and latitude in chronological order, is removed with the position ID given in step (1.1), then
Each user trajectory is indicated with a string of position ID；Meanwhile each user is identified with unique integer ID, thus user and
User trajectory can form the owner track that form is [User ID, track (position ID ..., position ID)]；
(1.3), according to owner track formed an oriented no weight graph G=<V, E>, wherein V is the set of all position ID, such as
There is some user from ID in fruitiTo IDj, then < IDi, IDj> indicate a directed edge, all such collection E when composition schemes G；
(2), track indicates
(2.1), using the oriented no weight graph G of building as input, each position ID in G out is learnt by Node2Vec algorithm
Low-dimensional real-valued vectors；
(2.2), the track of each user in step (1.2) is sliced according to Fixed Time Interval, so that each be used
Family track is divided into several position ID sequences, then carries out owner's mark to the position ID sequence after cutting with User ID；
(2.3), position ID is replaced with position ID corresponding low-dimensional real-valued vectors to the random length track after cutting, thus raw
At the track matrix of each user；
Then by interception or filling mode, the track matrix of each user's fixed dimension is constructed, so that data set is formed,
In, the vector of filling is the average value of the corresponding low latitude real-valued vectors of all position ID；
(3), prediction model is constructed
Four layers of convolutional neural networks are constructed, input layer is the track matrix of fixed dimension；Three convolution are arranged in convolutional layer
Core m*embedding_size, wherein m is constant, and embedding size is the low-dimensional real-valued vectors of Node2vec output
Dimension；Pond layer is k-max pooling, and k indicates the preceding k maximum value after convolution；The output of full articulamentum is input to
Softmax function obtains the probability distribution of track owner；
(4), training prediction model
(4.1), training set is constructed
Using the track matrix stack X of data set partial user fixed dimension and corresponding one-hot categorization vector collection Y as
Training set, wherein X=[Vec_x1,Vec_x2,...,Vec_xn], Vec_xnIndicate the track square of nth user's fixed dimension
Battle array, Y=[Vec_t1,Vec_t2,...,Vec_tn], Vec_tnIndicate the corresponding one-hot categorization vector of n-th of owner, if the
N position is 1, then remaining position is all 0；
(4.2), prediction model is initialized
Initialize the weight matrix W of each convolution kernel in convolutional layerpValue be normal state normal distribution, mean value 0, variance is
0.1；The bias vector B of each convolution kernel in convolutional layer is initialized simultaneouslypIt is 0.1, the element number of each bias vector is
The number of respective layer neuron；The weight matrix for initializing full articulamentum is W, and dimension is [batch_size*k* convolution kernel
Number, classification number], while the bias vector B value for initializing full articulamentum is 0.1, element number is classification number；Wherein, batch_
Size is constant, and p=1,2,3, p indicate which convolution kernel in convolutional layer；
(4.3) training set is input in the prediction model after initialization, using Adam algorithm optimization loss function, is then utilized
Error is transmitted to preceding layer by error back propagation BP algorithm, updates the weight matrix W of convolutional layerp, bias vector BpAnd full connection
Layer weight matrix W, bias vector B obtain convergent neural network model after iteration several times, to obtain having trained
At prediction model；
(5), track owner predicts
By user location longitude and latitude to be detected according to step (1), (2) the method, the rail of user's fixed dimension is constructed
Mark matrix, then by the track Input matrix of building into trained prediction model, obtain the track and correspond to all owners
The probability distribution of classification, the index of maximum value is then the corresponding owner's number in the track in probability distribution.
2. a kind of track owner's prediction technique based on convolutional neural networks according to claim 1, which is characterized in that institute
The loss function for the Adam algorithm optimization stated is as follows:
Wherein, N=batch_size, yjFor Vec_tjIn all elements, ajFor the output valve of softmax function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910266737.0A CN110059144B (en) | 2019-04-03 | 2019-04-03 | Trajectory owner prediction method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910266737.0A CN110059144B (en) | 2019-04-03 | 2019-04-03 | Trajectory owner prediction method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110059144A true CN110059144A (en) | 2019-07-26 |
| CN110059144B CN110059144B (en) | 2021-12-10 |
Family
ID=67318263
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910266737.0A Expired - Fee Related CN110059144B (en) | 2019-04-03 | 2019-04-03 | Trajectory owner prediction method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110059144B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110991751A (en) * | 2019-12-06 | 2020-04-10 | 讯飞智元信息科技有限公司 | User life pattern prediction method and device, electronic equipment and storage medium |
| CN112405542A (en) * | 2020-11-17 | 2021-02-26 | 中国科学院自动化研究所 | Musculoskeletal robot control method and system based on brain inspiring multitask learning |
| WO2021051224A1 (en) * | 2019-09-16 | 2021-03-25 | Intel Corporation | Stochastic trajectory prediction using social graph networks |
| CN113222229A (en) * | 2021-04-28 | 2021-08-06 | 南京航空航天大学 | Non-cooperative unmanned aerial vehicle trajectory prediction method based on machine learning |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180032868A1 (en) * | 2016-07-26 | 2018-02-01 | University Of Connecticut | Early prediction of an intention of a user's actions |
| CN108153867A (en) * | 2017-12-25 | 2018-06-12 | 山东大学 | User trajectory Forecasting Methodology and device based on temporal regularity |
-
2019
- 2019-04-03 CN CN201910266737.0A patent/CN110059144B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180032868A1 (en) * | 2016-07-26 | 2018-02-01 | University Of Connecticut | Early prediction of an intention of a user's actions |
| CN108153867A (en) * | 2017-12-25 | 2018-06-12 | 山东大学 | User trajectory Forecasting Methodology and device based on temporal regularity |
Non-Patent Citations (2)
| Title |
|---|
| QIANG GAO等: "Identifying Human Mobility via Trajectory Embeddings", 《PROCEEDINGS OF THE TWENTY-SIXTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE (IJCAI-17)》 * |
| 乔少杰等: "一种基于高斯混合模型的轨迹预测算法", 《软件学报》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021051224A1 (en) * | 2019-09-16 | 2021-03-25 | Intel Corporation | Stochastic trajectory prediction using social graph networks |
| CN110991751A (en) * | 2019-12-06 | 2020-04-10 | 讯飞智元信息科技有限公司 | User life pattern prediction method and device, electronic equipment and storage medium |
| CN112405542A (en) * | 2020-11-17 | 2021-02-26 | 中国科学院自动化研究所 | Musculoskeletal robot control method and system based on brain inspiring multitask learning |
| CN113222229A (en) * | 2021-04-28 | 2021-08-06 | 南京航空航天大学 | Non-cooperative unmanned aerial vehicle trajectory prediction method based on machine learning |
| CN113222229B (en) * | 2021-04-28 | 2024-02-13 | 南京航空航天大学 | Non-cooperative unmanned aerial vehicle track prediction method based on machine learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110059144B (en) | 2021-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110059144A (en) | A kind of track owner's prediction technique based on convolutional neural networks | |
| Yao et al. | Hierarchically structured meta-learning | |
| Yuan et al. | Parameter extraction of solar cell models using chaotic asexual reproduction optimization | |
| Gill et al. | Training back propagation neural networks with genetic algorithm for weather forecasting | |
| CN109948029A (en) | Based on the adaptive depth hashing image searching method of neural network | |
| CN107563567A (en) | Core extreme learning machine Flood Forecasting Method based on sparse own coding | |
| CN111119282B (en) | Pressure monitoring point optimal arrangement method for water supply pipe network | |
| CN108009674A (en) | Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks | |
| CN109214599A (en) | The method that a kind of pair of complex network carries out link prediction | |
| CN108701210A (en) | Method and system for CNN Network adaptations and object online tracing | |
| CN104834748A (en) | Image retrieval method utilizing deep semantic to rank hash codes | |
| CN112487805B (en) | Small sample Web service classification method based on meta-learning framework | |
| CN106503867A (en) | A kind of genetic algorithm least square wind power forecasting method | |
| CN113190688B (en) | Complex network link prediction method and system based on logical reasoning and graph convolution | |
| CN107609648B (en) | Genetic algorithm combined with stacking noise reduction sparse automatic encoder | |
| CN108460461A (en) | Mars earth shear parameters prediction technique based on GA-BP neural networks | |
| Zhang et al. | Simulation and prediction of soybean growth and development under field conditions | |
| Wu et al. | The establishment of plants following long-distance dispersal | |
| CN114169649A (en) | Traffic flow prediction method based on space-time data embedding | |
| CN113033786A (en) | Fault diagnosis model construction method and device based on time convolution network | |
| CN108399470A (en) | A kind of indoor PM2.5 prediction techniques based on more example genetic neural networks | |
| CN111950784A (en) | Productivity prediction method integrating attention mechanism | |
| CN111709519A (en) | Deep learning parallel computing architecture method and hyper-parameter automatic configuration optimization thereof | |
| CN114584406A (en) | Industrial big data privacy protection system and method for federated learning | |
| CN107026755B (en) | A kind of service recommendation method based on sequence study |
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 | ||
| 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: 20211210 |