CN113408191A - PM2.5 prediction method based on graph self-supervision learning and storage medium - Google Patents

PM2.5 prediction method based on graph self-supervision learning and storage medium Download PDF

Info

Publication number
CN113408191A
CN113408191A CN202110598978.2A CN202110598978A CN113408191A CN 113408191 A CN113408191 A CN 113408191A CN 202110598978 A CN202110598978 A CN 202110598978A CN 113408191 A CN113408191 A CN 113408191A
Authority
CN
China
Prior art keywords
graph
prediction
self
model
prediction method
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110598978.2A
Other languages
Chinese (zh)
Inventor
张波
吴泽权
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Normal University
University of Shanghai for Science and Technology
Original Assignee
Shanghai Normal University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai Normal University filed Critical Shanghai Normal University
Priority to CN202110598978.2A priority Critical patent/CN113408191A/en
Publication of CN113408191A publication Critical patent/CN113408191A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume or surface-area of porous materials
    • G01N15/06—Investigating concentration of particle suspensions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/18—Network design, e.g. design based on topological or interconnect aspects of utility systems, piping, heating ventilation air conditioning [HVAC] or cabling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/02—CAD in a network environment, e.g. collaborative CAD or distributed simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Tourism & Hospitality (AREA)
  • Biophysics (AREA)
  • Geometry (AREA)
  • Economics (AREA)
  • Computing Systems (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Computer Hardware Design (AREA)
  • Marketing (AREA)
  • Development Economics (AREA)
  • Chemical & Material Sciences (AREA)
  • General Business, Economics & Management (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Quality & Reliability (AREA)
  • Dispersion Chemistry (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Computational Mathematics (AREA)
  • Operations Research (AREA)
  • Mathematical Analysis (AREA)

Abstract

The invention relates to a PM2.5 prediction method based on graph self-supervision learning and a storage medium, wherein the PM2.5 prediction method comprises the following steps: step 1: inputting historical environment space-time data of multiple sites in an area to construct a graph; step 2: constructing a PM2.5 prediction model; and step 3: inputting null graph sequence data, and training the prediction model constructed in the step 2; and 4, step 4: calculating the accuracy of model prediction, if the accuracy exceeds a preset threshold, executing the step 5, otherwise, returning to the step 3; and 5: inputting the multi-site data in the area into the trained prediction model to obtain the PM2.5 predicted concentration value of the multiple sites in the area. Compared with the prior art, the method has the advantages of good prediction effect, good practicability and the like.

Description

PM2.5 prediction method based on graph self-supervision learning and storage medium
Technical Field
The invention relates to the technical field of air quality monitoring, in particular to a PM2.5 multi-site joint prediction method based on graph self-supervision learning and a storage medium.
Background
With the continuous development of our society and the continuous expansion of urban scale, the air pollution problem in cities is receiving social attention in recent years. Air pollution causes many problems, such as causing diseases of human respiratory tract and cardiopulmonary system, affecting social production efficiency, etc. Particularly, the fine particles such as PM2.5, once formed, have a large area and are difficult to be digested, and the like, and therefore, the important attention needs to be paid. In order to improve social problems caused by air pollution, timely and accurately know the PM2.5 propagation and diffusion trend, and establish an accurate PM2.5 concentration prediction model, the method can help people to know the pollutant concentration variation trend in advance and help management departments to make decisions in advance. Currently, air pollutant concentration prediction involves multiple departments, multiple space-time data and multiple regions, and accurate prediction often faces massive data and complex nonlinear propagation dependence.
Many researchers at home and abroad put forward a plurality of prediction methods and technologies, but most of the methods still use the traditional machine learning method and only can extract shallow features. Furthermore, most of the current research methods do not take into account the non-euclidian space where the contaminant data is collected from the monitored site. The spatial correlation under the non-Euclidean space is less researched, and although a PM2.5 prediction method exists in the prior art, the prediction accuracy and precision are low.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a PM2.5 prediction method and a storage medium based on graph self-supervision learning, which have good prediction effect and good practicability.
The purpose of the invention can be realized by the following technical scheme:
a PM2.5 prediction method based on graph self-supervision learning, wherein the PM2.5 prediction method comprises the following steps:
step 1: carrying out graph construction;
step 2: constructing a PM2.5 prediction model;
and step 3: training the prediction model constructed in the step 2;
and 4, step 4: calculating the accuracy of model prediction, if the accuracy exceeds a preset threshold, executing the step 5, otherwise, returning to the step 3;
and 5: inputting the multi-site data into the trained prediction model to obtain the PM2.5 predicted concentration values of all sites.
Preferably, the step 1 specifically comprises:
the method comprises the steps of collecting air pollutants and environment space-time monitoring data of a plurality of sites in an area, preprocessing the data, regarding the sites in the area as nodes in a graph, regarding the environment space-time data collected in the sites as node characteristic vectors, and regarding the distance between two nodes as the connecting edge when the distance between the two nodes is smaller than a preset threshold value L.
More preferably, the environmental spatiotemporal monitoring data comprise pollutant concentration values, meteorological monitoring values and spatiotemporal condition values.
Preferably, the step 2 specifically comprises:
the method comprises the steps of constructing a graph self-coding network ST-GAE and a long and short term memory network LSTM based on a deep learning principle to establish a spatio-temporal joint prediction model, wherein the graph self-coding network ST-GAE is used for extracting a spatial dependency relationship, the long and short term memory network LSTM is used for extracting a temporal dependency relationship, and the graph self-coding network ST-GAE is connected with the long and short term memory network LSTM.
More preferably, the graph self-coding network ST-GAE uses an Encoder-Decoder architecture, an Encoder portion of the ST-GAE uses GCN to fuse multi-level neighbor information to obtain a new feature matrix, and the Encoder portion can be represented by the following formula:
Z＝GCN(X,A)
wherein the content of the first and second substances,
Figure BDA0003092223030000021
is a feature matrix of the node;
Figure BDA0003092223030000022
is a adjacency matrix of the graph;
Figure BDA0003092223030000023
representing the graph fused with the neighbor information;
the GCN calculation method comprises the following steps:
Figure BDA0003092223030000024
wherein the content of the first and second substances,
Figure BDA0003092223030000025
W0and W1Is a parameter to be learned; d is the degree matrix of the graph.
More preferably, the graph self-encoding network ST-GAE reconstructs an original graph structure by using a feed-forward neural network as a Decoder, and the specific method is as follows:
Figure BDA0003092223030000026
wherein the content of the first and second substances,
Figure BDA0003092223030000027
is the reconstructed adjacency matrix; g is a sigmod activation function; wdAnd bdAre parameters to be learned.
More preferably, the step 3 specifically includes:
the ST-GAE part is first self-supervised trained, and loss function is used in the training process
Figure BDA0003092223030000037
Training is carried out;
and after the pre-training model is obtained, inputting the reconstructed feature matrix into an LSTM network, extracting a timing sequence dependency relationship, training by using a loss function MSE, and finely tuning the model to finish the training of the model.
More preferably, said loss function
Figure BDA0003092223030000031
The cross entropy is specifically:
Figure BDA0003092223030000032
wherein y represents a certain element value in the adjacency matrix a and is 0 or 1;
Figure BDA0003092223030000033
representing reconstructed adjacency matrices
Figure BDA0003092223030000034
The value of the corresponding element in (1) is 0 or 1; and N is the number of nodes.
More preferably, the loss function MSE is a mean square error, specifically:
Figure BDA0003092223030000035
wherein, yiActual values representing PM2.5 concentrations;
Figure BDA0003092223030000036
representing the predicted value of model PM2.5 concentration.
A storage medium having stored therein the PM2.5 prediction method of any one of the above.
Compared with the prior art, the invention has the following beneficial effects:
firstly, the prediction effect is good: the PM2.5 prediction method is based on the environment space-time big data and the deep learning theory, utilizes a large amount of accumulated air and meteorological monitoring data, and organizes the unstructured data by using a topological structure of a graph to form space-time graph sequence data; then, the novel deep learning model based on the graph self-supervision learning provided by the invention is used for carrying out deep extraction on the complex nonlinear space-time dependence in the PM2.5 propagation process, and finally, the predicted values of a plurality of stations in the region are output, and the predicted values have more excellent prediction effect than the predicted values obtained by the traditional method, so that the deep learning model has practical application value.
Secondly, the practicability is good: the PM2.5 prediction method can directly predict the PM2.5 of all the sites in the current area, does not need to respectively process data of each site, and has the advantages of high data processing speed, wide range and good practicability.
Drawings
FIG. 1 is a flow chart of a PM2.5 prediction method of the present invention.
FIG. 2 is a schematic diagram of spatiotemporal map sequence data constructed in an embodiment of the present invention.
Fig. 3 is a structure diagram of a PM2.5 concentration multi-site joint prediction model based on graph self-supervision learning in the embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
The graph neural network is concerned by researchers with strong non-European data processing capacity, and the multi-site prediction model is established by utilizing the strong learning capacity of the graph neural network on non-European data and combining the ideas of an autoencoder and self-supervision learning. The model can consider the non-linear dependence of pollutant propagation from two aspects of space and time, and simultaneously output the predicted values of a plurality of stations, and the following provides a specific implementation mode:
a PM2.5 prediction method based on graph self-supervised learning, the flow of which is shown in fig. 1, comprising:
step 1: carrying out graph construction;
acquiring air pollutants and environment space-time monitoring data of a plurality of sites in an area, preprocessing the data, regarding the sites in the area as nodes in a graph, regarding the environment space-time data acquired in the sites as node characteristic vectors, and regarding that a connecting edge exists between two nodes if the distance between the two nodes is smaller than a preset threshold value L;
the environment space-time monitoring data comprise a pollutant concentration value, a meteorological monitoring value and a space-time condition value, and the Euclidean distance L between two nodes is used as the basis for the existence of the edge;
the distance threshold L adopted in this embodiment is 200km, that is, when the distance between two stations is less than or equal to 200km, a connecting edge is established between two nodes.
The contaminants include: AQI, PM2.5, PM10, SO2, NO2, O3, CO;
the weather monitoring values include: air temperature, air pressure, wind direction, wind speed, precipitation and cloud cover;
the time condition values include: year, month, day, hour, week;
the spatial condition values include: longitude, latitude.
And finally, forming 20-dimensional feature vectors, collecting pollutants and meteorological features once per hour, and finally obtaining time-space diagram sequence data after missing value filling and abnormal value processing. That is, in the T period, the time-space diagram sequence data is obtained
Figure BDA0003092223030000041
Step 2: and (3) constructing a PM2.5 prediction model, and setting a model structure, a hyper-parameter and a loss function, wherein the specific structure of the model is shown in FIG. 2. The model mainly comprises two parts, wherein a graph self-Encoder part is responsible for extracting the spatial dependence relation of pollutant propagation diffusion in a graph, implicit graph representation of the graph is obtained through an Encoder-Decoder self-Encoder framework, finally, the implicit graph representation containing multi-order neighbor node information is input into a long-short term memory network to extract the time dependence relation of pollutant propagation diffusion, and PM2.5 predicted values of all stations in an area are output;
the graph self-coding network ST-GAE is used for extracting spatial dependency, the long-short term memory network LSTM is used for extracting temporal dependency, and the two parts are connected through operations of vector stretching, filling and the like.
The graph self-coding network ST-GAE uses an Encoder-Decoder architecture, an Encoder part of the ST-GAE uses GCN to fuse multi-order neighbor information to obtain a new characteristic matrix, and the Encoder part can be expressed by the following formula:
Z＝GCN(X,A)
wherein the content of the first and second substances,
Figure BDA0003092223030000051
is a feature matrix of the node;
Figure BDA0003092223030000052
is a adjacency matrix of the graph;
Figure BDA0003092223030000053
representing the graph fused with the neighbor information;
regarding the GCN as a function, inputting a feature matrix X and an adjacency matrix A of a graph and outputting a potential representation of the graph, wherein the GCN is calculated by the following steps:
Figure BDA0003092223030000054
wherein the content of the first and second substances,
Figure BDA0003092223030000055
W0and W1Is a parameter to be learned; d is a degree matrix of the graph;
the graph self-coding network ST-GAE adopts a feedforward neural network as a Decoder to reconstruct an original graph structure, the reconstruction aims at enabling a reconstructed adjacent matrix and an original adjacent matrix to be similar as much as possible, because the adjacent matrix determines the topological structure of the graph, and the calculation formula of the Decoder part is as follows:
Figure BDA0003092223030000056
wherein the content of the first and second substances,
Figure BDA0003092223030000057
is the reconstructed adjacency matrix; g is a sigmod activation function; wdAnd bdIs a parameter to be learned;
and step 3: training the prediction model constructed in the step 2, constructing pollutant space-time data into a topological graph structure aiming at the constructed model, constructing a training set, a testing set and a verification set, initializing and training the model by using the training set data, adjusting parameters of the model by using the verification set and evaluation indexes, and testing the effect of the model by using the testing set;
in this embodiment, the percentage of the three data sets of the training data set, the verification data set and the test data set is 70%, 10% and 20% in sequence.
Firstly, self-supervision training is carried out on an ST-GAE part, so that the model fully extracts the characteristics of multi-order neighbors, the spatial dependence relation in the PM2.5 propagation process is excavated, and a loss function is used in the training process
Figure BDA0003092223030000058
Training is carried out;
and after a pre-training model is obtained, inputting the reconstructed feature matrix into an LSTM network, extracting a time sequence dependency relationship, training by using a loss function MSE, finely adjusting the model, completing the training of the model, and finally outputting PM2.5 concentration values of all stations in the whole topological graph in a period of time in the future.
Loss function
Figure BDA0003092223030000061
Adopting cross entropy, specifically:
Figure BDA0003092223030000062
wherein y represents a certain element value in the adjacency matrix a and is 0 or 1;
Figure BDA0003092223030000063
representing reconstructed adjacency matrices
Figure BDA0003092223030000064
The value of the corresponding element in (1) is 0 or 1; n is the number of nodes;
the loss function MSE is a mean square error, and specifically includes:
Figure BDA0003092223030000065
wherein, yiActual values representing PM2.5 concentrations;
Figure BDA0003092223030000066
a predicted value representing the model PM2.5 concentration;
and 4, step 4: calculating the accuracy of model prediction, if the accuracy exceeds a preset threshold, executing the step 5, otherwise, returning to the step 3;
and 5: inputting the multi-site data into a trained prediction model, and obtaining the PM2.5 predicted concentration value of all sites in a period of time in the future.
The embodiment also relates to a storage medium, wherein any PM2.5 prediction method is stored in the storage medium.
While the invention has been described with reference to specific embodiments, the invention is not limited thereto, and various equivalent modifications and substitutions can be easily made by those skilled in the art within the technical scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (10)

1. A PM2.5 prediction method based on graph self-supervision learning is characterized in that the PM2.5 prediction method comprises the following steps:
step 1: carrying out graph construction;
step 2: constructing a PM2.5 prediction model;
and step 3: training the prediction model constructed in the step 2;
and 4, step 4: calculating the accuracy of model prediction, if the accuracy exceeds a preset threshold, executing the step 5, otherwise, returning to the step 3;
and 5: inputting the multi-site data into the trained prediction model to obtain the PM2.5 predicted concentration values of all sites.
2. The PM2.5 prediction method based on graph self-supervision learning according to claim 1, characterized in that the step 1 specifically comprises:
the method comprises the steps of collecting air pollutants and environment space-time monitoring data of a plurality of sites in an area, preprocessing the data, regarding the sites in the area as nodes in a graph, regarding the environment space-time data collected in the sites as node characteristic vectors, and regarding the distance between two nodes as the connecting edge when the distance between the two nodes is smaller than a preset threshold value L.
3. The graph self-supervised learning based PM2.5 prediction method as claimed in claim 2, wherein the environmental spatiotemporal monitoring data comprise pollutant concentration values, meteorological monitoring values and spatiotemporal condition values.
4. The PM2.5 prediction method based on graph self-supervision learning according to claim 1, characterized in that the step 2 specifically comprises:
the method comprises the steps of constructing a graph self-coding network ST-GAE and a long and short term memory network LSTM based on a deep learning principle to establish a spatio-temporal joint prediction model, wherein the graph self-coding network ST-GAE is used for extracting a spatial dependency relationship, the long and short term memory network LSTM is used for extracting a temporal dependency relationship, and the graph self-coding network ST-GAE is connected with the long and short term memory network LSTM.
5. The method of claim 4, wherein the graph self-supervised learning based PM2.5 prediction method is characterized in that the graph self-coding network ST-GAE uses an Encoder-Decoder architecture, an Encoder part of the ST-GAE uses GCN to fuse multi-level neighbor information to obtain a new feature matrix, and the Encoder part can be expressed by the following formula:
Z＝GCN(X,A)
wherein the content of the first and second substances,
Figure FDA0003092223020000021
is a feature matrix of the node;
Figure FDA0003092223020000022
is a adjacency matrix of the graph;
Figure FDA0003092223020000023
representing the graph fused with the neighbor information;
the GCN calculation method comprises the following steps:
Figure FDA0003092223020000024
wherein the content of the first and second substances,
Figure FDA0003092223020000025
W0and W1Is a parameter to be learned; d is the degree matrix of the graph.
6. The method according to claim 4, wherein the graph self-encoding network ST-GAE adopts a feedforward neural network as a Decoder to reconstruct an original graph structure, and the method comprises:
Figure FDA0003092223020000026
wherein the content of the first and second substances,
Figure FDA0003092223020000027
is the reconstructed adjacency matrix; g is a sigmod activation function; wdAnd bdAre parameters to be learned.
7. The graph self-supervision learning-based PM2.5 prediction method according to claim 4, wherein the step 3 specifically comprises:
the ST-GAE part is first self-supervised trained, and loss function is used in the training process
Figure FDA0003092223020000028
Training is carried out;
and after the pre-training model is obtained, inputting the reconstructed feature matrix into an LSTM network, extracting a timing sequence dependency relationship, training by using a loss function MSE, and finely tuning the model to finish the training of the model.
8. The graph self-supervised learning-based PM2.5 prediction method as claimed in claim 7, wherein the loss function
Figure FDA0003092223020000029
The cross entropy is specifically:
Figure FDA00030922230200000210
wherein y represents a certain element value in the adjacency matrix a and is 0 or 1;
Figure FDA00030922230200000211
representing reconstructed adjacency matrices
Figure FDA00030922230200000212
The value of the corresponding element in (1) is 0 or 1; and N is the number of nodes.
9. The PM2.5 prediction method based on graph self-supervised learning according to claim 7, wherein the loss function MSE is a mean square error, specifically:
Figure FDA00030922230200000213
wherein, yiActual values representing PM2.5 concentrations;
Figure FDA00030922230200000214
representing the predicted value of model PM2.5 concentration.
10. A storage medium storing a PM2.5 prediction method according to any one of claims 1 to 9.
CN202110598978.2A 2021-05-31 2021-05-31 PM2.5 prediction method based on graph self-supervision learning and storage medium Pending CN113408191A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110598978.2A CN113408191A (en) 2021-05-31 2021-05-31 PM2.5 prediction method based on graph self-supervision learning and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110598978.2A CN113408191A (en) 2021-05-31 2021-05-31 PM2.5 prediction method based on graph self-supervision learning and storage medium

Publications (1)

Publication Number Publication Date
CN113408191A true CN113408191A (en) 2021-09-17

Family

ID=77675296

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110598978.2A Pending CN113408191A (en) 2021-05-31 2021-05-31 PM2.5 prediction method based on graph self-supervision learning and storage medium

Country Status (1)

Country Link
CN (1) CN113408191A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117633661A (en) * 2024-01-26 2024-03-01 西南交通大学 Slag car high-risk pollution source classification method based on evolution diagram self-supervised learning

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109142171A (en) * 2018-06-15 2019-01-04 上海师范大学 The city PM10 concentration prediction method of fused neural network based on feature expansion
CN110021165A (en) * 2019-03-18 2019-07-16 浙江工业大学 A kind of traffic flow forecasting method based on Autoencoder-LSTM Fusion Model
US20200033866A1 (en) * 2018-07-24 2020-01-30 Pony.ai, Inc. Generative adversarial network enriched driving simulation
US20200160154A1 (en) * 2018-11-15 2020-05-21 Element Ai Inc. Systems and methods for assessing item compatibility
CN112149887A (en) * 2020-09-08 2020-12-29 北京工业大学 PM2.5 concentration prediction method based on data space-time characteristics
CN112330951A (en) * 2020-09-11 2021-02-05 浙江工业大学 Method for realizing road network traffic data restoration based on generation of countermeasure network
CN112784121A (en) * 2021-01-28 2021-05-11 浙江工业大学 Traffic accident prediction method based on space-time diagram representation learning

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109142171A (en) * 2018-06-15 2019-01-04 上海师范大学 The city PM10 concentration prediction method of fused neural network based on feature expansion
US20200033866A1 (en) * 2018-07-24 2020-01-30 Pony.ai, Inc. Generative adversarial network enriched driving simulation
US20200160154A1 (en) * 2018-11-15 2020-05-21 Element Ai Inc. Systems and methods for assessing item compatibility
CN110021165A (en) * 2019-03-18 2019-07-16 浙江工业大学 A kind of traffic flow forecasting method based on Autoencoder-LSTM Fusion Model
CN112149887A (en) * 2020-09-08 2020-12-29 北京工业大学 PM2.5 concentration prediction method based on data space-time characteristics
CN112330951A (en) * 2020-09-11 2021-02-05 浙江工业大学 Method for realizing road network traffic data restoration based on generation of countermeasure network
CN112784121A (en) * 2021-01-28 2021-05-11 浙江工业大学 Traffic accident prediction method based on space-time diagram representation learning

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZHIYUAN WU ET AL: "MSSTN: Multi-Scale Spatial Temporal Network for Air Pollution Prediction", 2019 IEEE INTERNATIONAL CONFERENCE ON BIG DATA (BIG DATA), pages 1 - 10 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117633661A (en) * 2024-01-26 2024-03-01 西南交通大学 Slag car high-risk pollution source classification method based on evolution diagram self-supervised learning
CN117633661B (en) * 2024-01-26 2024-04-02 西南交通大学 Slag car high-risk pollution source classification method based on evolution diagram self-supervised learning

Similar Documents

Publication Publication Date Title
CN111798051B (en) Air quality space-time prediction method based on long-term and short-term memory neural network
CN111832814B (en) Air pollutant concentration prediction method based on graph attention mechanism
CN109492822B (en) Air pollutant concentration time-space domain correlation prediction method
CN109142171B (en) Urban PM10 concentration prediction method based on feature expansion and fusing with neural network
CN106650767B (en) Flood forecasting method based on cluster analysis and real-time correction
CN108009674A (en) Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks
CN109146161A (en) Merge PM2.5 concentration prediction method of the stack from coding and support vector regression
CN112085163A (en) Air quality prediction method based on attention enhancement graph convolutional neural network AGC and gated cyclic unit GRU
CN111458471B (en) Water area detection early warning method based on graph neural network
CN114240000A (en) Air quality prediction method based on space-time graph convolution network
CN113487066A (en) Long-time-sequence freight volume prediction method based on multi-attribute enhanced graph convolution-Informer model
CN110766222B (en) PM2.5 concentration prediction method based on particle swarm parameter optimization and random forest
CN114707688A (en) Photovoltaic power ultra-short-term prediction method based on satellite cloud chart and space-time neural network
CN113516304B (en) Regional pollutant space-time joint prediction method and device based on space-time diagram network
CN111369057A (en) Air quality prediction optimization method and system based on deep learning
CN114510513A (en) Short-term meteorological forecast data processing method for ultra-short-term photovoltaic power prediction
CN116187210B (en) Time-space multisource offshore area water quality time sequence prediction method of LSTM coupling mechanism model
CN114841400A (en) Air quality prediction method based on multi-task space-time diagram convolution
CN112183625A (en) PM based on deep learning2.5High-precision time-space prediction method
CN112232543A (en) Multi-site prediction method based on graph convolution network
CN115860286A (en) Air quality prediction method and system based on time sequence door mechanism
CN114882373A (en) Multi-feature fusion sandstorm prediction method based on deep neural network
CN113408191A (en) PM2.5 prediction method based on graph self-supervision learning and storage medium
CN117494034A (en) Air quality prediction method based on traffic congestion index and multi-source data fusion
CN117057471A (en) District electricity prediction method based on graph multi-head attention mechanism

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