CN116256468A - Air quality prediction optimization method based on dynamic self-adaptive particle swarm - Google Patents

Air quality prediction optimization method based on dynamic self-adaptive particle swarm Download PDF

Info

Publication number
CN116256468A
CN116256468A CN202211262759.8A CN202211262759A CN116256468A CN 116256468 A CN116256468 A CN 116256468A CN 202211262759 A CN202211262759 A CN 202211262759A CN 116256468 A CN116256468 A CN 116256468A
Authority
CN
China
Prior art keywords
bigru
algorithm
data
prediction model
particle
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
CN202211262759.8A
Other languages
Chinese (zh)
Inventor
蒋哲峰
张秀娟
高嘉良
王珂
管鑫
杨伟
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.)
China Telecom Wanwei Information Technology Co Ltd
Original Assignee
China Telecom Wanwei Information Technology Co Ltd
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 China Telecom Wanwei Information Technology Co Ltd filed Critical China Telecom Wanwei Information Technology Co Ltd
Priority to CN202211262759.8A priority Critical patent/CN116256468A/en
Publication of CN116256468A publication Critical patent/CN116256468A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/0004Gaseous mixtures, e.g. polluted air
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/004Artificial life, i.e. computing arrangements simulating life
    • G06N3/006Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q50/00Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10Services
    • G06Q50/26Government or public services

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Development Economics (AREA)
  • Quality & Reliability (AREA)
  • General Engineering & Computer Science (AREA)
  • Operations Research (AREA)
  • Primary Health Care (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Educational Administration (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Game Theory and Decision Science (AREA)
  • Combustion & Propulsion (AREA)
  • Food Science & Technology (AREA)
  • Medicinal Chemistry (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)

Abstract

The invention belongs to the technical field of data mining processing, and relates to an air quality prediction optimization method based on a dynamic self-adaptive particle swarm. The method introduces a dynamic space search strategy into a particle swarm algorithm, so that the particle position is oscillated to improve the optimizing capability of particles, and the global searching capability and the local searching capability of the algorithm are balanced by adaptively adjusting a learning factor. And optimizing parameters of the BiGRU neural network by using a DAPSO algorithm, and predicting the concentration of the particulate matters by using an optimized neural network model (DAPSO-BiGRU). And combining the improved intelligent algorithm with a neural network, and optimizing a neural network model through the intelligent algorithm to predict the concentration of each particulate matter.

Description

Air quality prediction optimization method based on dynamic self-adaptive particle swarm
Technical Field
The invention belongs to the technical field of data mining processing, and relates to an air quality prediction optimization method based on a dynamic self-adaptive particle swarm.
Background
The particle swarm PSO algorithm is a global optimization technology based on swarm intelligence proposed by Eberhart and Kennedy in 1995, and inspiration of the particle swarm PSO algorithm is derived from social behaviors of animals. In the conventional particle swarm algorithm, the position of each particle represents a candidate solution to the optimization problem, and the degree of the solution is determined by the fitness value found by the objective function. The particles may be returned to the previous optimal position based on the memory, which is determined by the previous optimal performance pid of the individual particles and the previous optimal performance gid of all particles. The particle swarm algorithm is used as a global optimization algorithm, has the characteristics of simple adjustment, easy realization and high efficiency, but meanwhile, the PSO algorithm cannot control the speed of particles, and has the defects of large calculated amount and long calculation period.
Disclosure of Invention
In order to overcome the problems in the prior art, the invention aims to provide a dynamic self-adaptive particle swarm algorithm which solves the problems in the prior art.
An air quality prediction optimization method based on dynamic self-adaptive particle swarm comprises the following steps:
s1, data analysis:
the used data are air quality data of the China environmental monitoring total station; the data includes time, PM2.5, PM10, SO 2 、NO 2 、O 3 And daily average of carbon monoxide. And for PM2.5, O3, PM10, SO 2 、NO 2 Filling data missing values and carrying out data normalization on particulate matter data such as CO;
using a normalization method of dispersion normalization, namely, linearly scaling the original sample data to enable the original sample data to be distributed between 0 and 1, wherein the formula is as follows:
Figure DEST_PATH_IMAGE001
(1)
s2, evaluating concentration of particulate matters
Each influencing factor PM2.5, PM10, SO 2 、NO 2 、O 3 And carbon monoxide, etc., use gray association degree to analyze the method of the degree of correlation among every factor, if two factors have consistent change trend, namely have higher synchronous change degree, the degree of correlation of two is higher, otherwise, the degree of correlation of two is lower;
s3, biGRU algorithm
Outputting the state of the unidirectional neural network from front to back, wherein at any moment, the input of the BiGRU corresponds to 2 GRUs with opposite directions, and the output is also influenced by the GRUs;
the present hidden state of the bidirectional GRU is mainly dependent on the output of the corresponding hidden state when the present input xt, (t-1) is input
Figure 369588DEST_PATH_IMAGE002
And the output of hidden layer states in different directions +.>
Figure DEST_PATH_IMAGE003
. The bidirectional GRU is two GRUs with opposite directions, and hidden layer at a certain moment is formed by the hidden layer and the forward hidden layer +.>
Figure 160870DEST_PATH_IMAGE004
Backward hidden layer state->
Figure DEST_PATH_IMAGE005
Is prepared from (I)>
Figure 99876DEST_PATH_IMAGE006
(2)
Figure DEST_PATH_IMAGE007
(3)
Figure 46972DEST_PATH_IMAGE008
(4)
W t 、v t Forward hidden layer h for biglu t Backward hiddenLayer state h t Weight, b t Then it is the bias of the hidden state;
s4, DAPSO particle swarm optimization algorithm
Figure DEST_PATH_IMAGE009
(5)
Figure 438639DEST_PATH_IMAGE010
(6)
Wherein k represents the current iteration number of the example, kmax is the maximum allowed iteration number, and when t=1, as can be known from formula (5), xi (k+1) =xi (k), as k increases, logkmax (k) also increases and does not exceed 1 all the time, and a logarithmic function is introduced to enable the position of each particle to obtain oscillation, and the search space of the particle is dynamically adjusted, so that the particles are effectively guided to find the optimal solution;
the random inertia weight is set by the following formula:
Figure DEST_PATH_IMAGE011
(7)
the randomness of the algorithm is enhanced by adopting a learning factor which is adaptively adjusted along with the inertia weight, the utilization degree of particle population information is effectively adjusted, and the adjustment formula is as follows:
Figure 90069DEST_PATH_IMAGE012
(8)
Figure DEST_PATH_IMAGE013
(9)
the learning factor dynamically changes as the inertia weight decreases;
s5, using air pollution prediction model
And respectively predicting the concentration of each particulate matter by using the newly constructed DAPSO-BiGRU prediction model, PSO-BiGRU prediction model, GRU prediction model and BiGRU prediction model, comparing the prediction result with the DAPSO-BiGRU prediction model, and comprehensively evaluating by using three degrees of average absolute error MAE, root mean square error RMSE and average absolute percentage error MAPE to obtain a prediction model.
The beneficial effects of the invention are as follows: an improved dynamic self-adaptive particle swarm optimization algorithm (DAPSO) is provided for the defect that a particle swarm algorithm is high in convergence speed and easy to fall into local optimum, and introduces a dynamic space search strategy into the particle swarm algorithm, so that particle positions are oscillated to improve the optimizing capability of particles, and meanwhile, the global searching capability and the local searching capability of the algorithm are balanced by adaptively adjusting learning factors. And optimizing parameters of the BiGRU neural network by using a DAPSO algorithm, and predicting the concentration of the particulate matters by using an optimized neural network model (DAPSO-BiGRU). And combining the improved intelligent algorithm with a neural network, and optimizing a neural network model through the intelligent algorithm to predict the concentration of each particulate matter. The research results are helpful for better understanding of weather causes of serious air pollution events in related areas, and developing effective regional air quality management strategies to reduce air pollution and minimize adverse effects on human beings and ecosystems.
Drawings
FIG. 1 is a predictive flow diagram;
FIG. 2 is a block diagram of a bi-directional GRU model;
FIG. 3 is a graph showing the prediction of the concentration of each particulate matter.
Detailed Description
The invention will be described in detail below with reference to the drawings and the detailed description.
The first part is shown in fig. 1, and the method comprises the following complete steps:
s1, data analysis
The data used are daily air quality data from 2017, 1 st to 2021, 9 nd 30 th day in the Chengdu city of the chinese environmental monitoring total station. The data includes time, daily average of the cities where PM2.5, PM10, SO2, NO2, O3 and carbon monoxide are located. And performing data missing value filling and data normalization processing on particulate matter data such as PM2.5, O3, PM10, SO2, NO2, CO and the like.
The invention adopts a mean value method with simpler calculation process to process the missing value. The data average value of a certain time range before and after the missing value is calculated to fill up, so that a better reduction effect can be achieved.
The method is characterized in that a normalization method of dispersion standardization is used, namely the original sample data is linearly scaled to be distributed between 0 and 1, the method is suitable for data of which the data are distributed in a limited range, and the formula is as follows:
Figure DEST_PATH_IMAGE015
(1)
s2, evaluating concentration of particulate matters
Considering the mutual influence among air pollutants, the influence factors PM2.5, PM10, SO2, NO2, O3, carbon monoxide and the like have complex nonlinear relations, and gray correlation analysis is used herein, namely a method for evaluating the correlation degree among the factors by using gray correlation degree, wherein the gray correlation degree refers to the similarity or different degrees of development trends among the factors. If the two factors have consistent change trend, namely higher synchronous change degree, the association degree of the two factors is higher. Otherwise, the association degree of the two is lower.
PM2.5, PM10, SO are calculated herein as data from 1 in capital 2017 to 10 in 2020 2 、NO 2 、O 3 And carbon monoxide grey correlation between the daily averages of the six data, the results are shown in table 1:
table 1 gray correlation of influence factors
Figure DEST_PATH_IMAGE017
S3, biGRU algorithm
For unidirectional neural networks, the state of the network is mostly output from front to back. At any instant, the input of a BiGRU will always correspond to 2 GRUs in opposite directions, while the output is also affected by this GRU. The specific structure of the biglu is shown in fig. 2.
As can be seen from fig. 2, the present hidden state of the bidirectional GRU is mainly dependent onOutput of corresponding hidden layer state when xt (t-1) is input now
Figure DEST_PATH_IMAGE019
And the output of hidden layer states in different directions +.>
Figure DEST_PATH_IMAGE021
Because a bi-directional GRU can be divided into two GRUs with opposite directions, its hidden state at a certain moment can be aided by the forward hidden state +.>
Figure DEST_PATH_IMAGE023
Backward hidden layer
Figure DEST_PATH_IMAGE025
The method comprises the following steps:
Figure DEST_PATH_IMAGE027
(2)
Figure DEST_PATH_IMAGE029
(3)
Figure DEST_PATH_IMAGE031
(4)
W t 、v t forward hidden layer h for biglu t Backward hidden layer state h t Weight, b t Then it is the bias of the hidden state.
S4, DAPSO particle swarm optimization algorithm
The current position of the particle in the PSO algorithm is obtained by directly adding the previous position and the speed of the particle, so that the searching efficiency of the particle swarm algorithm is low, and even a premature convergence phenomenon occurs. The formula vi can only indicate the searching speed, and does not mean that the particle is approaching to the correct position gradually, and possibly the particle is moving away from the optimal position, and the particle searching process is a dynamic nonlinear process and has no definite track. DAPSO is a dynamic particle swarm optimization algorithm that can adaptively adjust parameters. The algorithm combines the traditional particle swarm optimization algorithm with the dynamic space searching strategy, so that the position oscillation process of particles in the iterative searching process is more active, and the problem that the particle swarm algorithm is easy to trap into a local extremum is solved. The following adjustments are made:
Figure DEST_PATH_IMAGE033
(5)/>
Figure DEST_PATH_IMAGE035
(6)
where k represents the current iteration number of the example and kmax is the maximum allowed iteration number. As can be seen from equation (5), when t=1, xi (k+1) =xi (k), as k increases, logkmax (k) also increases, but does not exceed 1 all the time, and the logarithmic function is introduced to enable the position of each particle to obtain oscillation, and dynamically adjust the search space of the particle, so that the particle is effectively guided to find the optimal solution.
The random inertia weight is set by the following formula:
Figure DEST_PATH_IMAGE037
(7)
the randomness of the algorithm is enhanced by adopting a learning factor which is adaptively adjusted along with the inertia weight, the utilization degree of particle population information is more effectively adjusted, and the adjustment strategy is as follows:
Figure DEST_PATH_IMAGE039
(8)
Figure DEST_PATH_IMAGE041
(9)
the learning factor dynamically changes as the inertia weight decreases.
S5, application of air pollution prediction model
The DAPSO-BiGRU prediction model, the PSO-BiGRU prediction model, the GRU prediction model and the BiGRU prediction model which are newly constructed in the research are used for respectively predicting the concentration of each particulate matter, the prediction result is compared with the DAPSO-BiGRU prediction model, and three measurement strategies are utilized: and (3) comprehensively evaluating the Mean Absolute Error (MAE), the Root Mean Square Error (RMSE) and the Mean Absolute Percentage Error (MAPE), and exploring that the prediction model has high prediction precision, so as to obtain the optimal prediction model.
The second part of the invention provides an improved dynamic self-adaptive particle swarm optimization (DAPSO) algorithm, which is improved on the basis of a PSO algorithm, and the position of particles can be adjusted by dynamically adjusting parameters of the algorithm, so that the particles can find a globally optimal solution. Parameters of the BiGRU neural network are optimized by using a DAPSO algorithm, and the concentration of the particulate matters is predicted by using an optimized neural network model (DAPSO-BiGRU). To verify the predictive effectiveness of the DAPSO-BiLSTM model presented herein, experiments were performed and compared to the GRU, biGRU, PSO-BiGRU neural network model.
The innovation of the patent is that: (1) A new air pollution prediction network structure of the self-adaptive dynamic particle swarm optimization algorithm optimized BiGRU neural network is researched. (2) An effective parameter optimization technology is explored for realizing a deep learning model in air quality research.
Third section:
FIG. 3 is a graph showing the concentration prediction of each particulate matter, and it is clear from the graph showing the prediction of each particulate matter and from the three error metric values that the DAPSO-LSTM prediction model is better than the PSO-LSTM and LSTM prediction models. We use bi-directional GRUs as predictive models in this patent.
Table 2 evaluation of the results of the prediction models
Figure DEST_PATH_IMAGE043
Firstly, the GRU combines the forget gate and the input gate on the basis of the LSTM, and changes the forget gate and the input gate into an update gate, and the information at the current moment can be freely selected by depending on the gate, and the information depends on the past and the current. A reset gate is then added to control how much of the update of the information at the current time depends on the hidden state at the previous time, adding a degree of flexibility, and also combining cell and hidden into one output. This structure makes the network more easy to train with fewer parameters. Secondly, the purpose of the bidirectional GRU is to input a sequence, and obtain the characteristic representation of the bidirectional GRU at each moment, namely, the context semantic information from each moment of output to the moment is represented by a fixed-length vector. Specifically, the bi-directional recurrent neural network processes the input sequence sequentially in the time dimension in order and in reverse order, i.e., forward (forward) and backward (backward), respectively, and concatenates the outputs of each time step RNN into a final output layer. The output node of each time step thus contains complete past and future context information for the current time instant in the input sequence. This will make the prediction result more accurate, and thus a bi-directional gated recurrent neural network based on dynamically adaptive particle swarms will benefit from an improvement in air quality prediction accuracy.

Claims (1)

1. The air quality prediction optimization method based on the dynamic self-adaptive particle swarm is characterized by comprising the following steps of:
s1, data analysis
The used data are air quality data of the China environmental monitoring total station; the data includes time, PM2.5, PM10, SO 2 、NO 2 、O 3 And daily average of carbon monoxide. And for PM2.5, O3, PM10, SO 2 、NO 2 Filling data missing values and carrying out data normalization on particulate matter data such as CO;
using a normalization method of dispersion normalization, namely, linearly scaling the original sample data to enable the original sample data to be distributed between 0 and 1, wherein the formula is as follows:
Figure DEST_PATH_IMAGE002
(1)
s2, evaluating concentration of particulate matters
Each influencing factor PM2.5, PM10, SO 2 、NO 2 、O 3 And carbon monoxide, etc., use gray association degree to analyze the method of the degree of correlation among every factor, if two factors have consistent change trend, namely have higher synchronous change degree, the degree of correlation of two is higher, otherwise, the degree of correlation of two is lower;
s3, biGRU algorithm
Outputting the state of the unidirectional neural network from front to back, wherein at any moment, the input of the BiGRU corresponds to 2 GRUs with opposite directions, and the output is also influenced by the GRUs;
the two-way GRU is two GRUs with opposite directions, and the hidden state at a certain moment is obtained by virtue of the hidden state of the two GRUs with the forward hidden state and the backward hidden state:
Figure DEST_PATH_IMAGE004
(2)
Figure DEST_PATH_IMAGE006
(3)
Figure DEST_PATH_IMAGE008
(4)
W t 、v t forward hidden layer h for biglu t Backward hidden layer state h t Weight, b t Then it is the bias of the hidden state;
s4, DAPSO particle swarm optimization algorithm
Figure DEST_PATH_IMAGE010
(5)
Figure DEST_PATH_IMAGE012
(6)
Wherein k represents the current iteration number of the example, kmax is the maximum allowed iteration number, and when t=1, as can be known from the formula (5), xi (k+1) =xi (k), as k increases, logkmax (k) also increases and does not exceed 1 all the time, and a logarithmic function is introduced to enable the position of each particle to obtain oscillation, and the search space of the particle is dynamically adjusted, so that the particles are effectively guided to find an optimal solution;
the random inertia weight is set by the following formula:
Figure DEST_PATH_IMAGE014
(7)
the randomness of the algorithm is enhanced by adopting a learning factor which is adaptively adjusted along with the inertia weight, the utilization degree of particle population information is effectively adjusted, and the adjustment formula is as follows:
Figure DEST_PATH_IMAGE016
(8)
Figure DEST_PATH_IMAGE018
(9)
the learning factor dynamically changes as the inertia weight decreases;
s5, using air pollution prediction model
And respectively predicting the concentration of each particulate matter by using the newly constructed DAPSO-BiGRU prediction model, PSO-BiGRU prediction model, GRU prediction model and BiGRU prediction model, comparing the prediction result with the DAPSO-BiGRU prediction model, and comprehensively evaluating by using three degrees of average absolute error MAE, root mean square error RMSE and average absolute percentage error MAPE to obtain a prediction model.
CN202211262759.8A 2022-10-15 2022-10-15 Air quality prediction optimization method based on dynamic self-adaptive particle swarm Pending CN116256468A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211262759.8A CN116256468A (en) 2022-10-15 2022-10-15 Air quality prediction optimization method based on dynamic self-adaptive particle swarm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211262759.8A CN116256468A (en) 2022-10-15 2022-10-15 Air quality prediction optimization method based on dynamic self-adaptive particle swarm

Publications (1)

Publication Number Publication Date
CN116256468A true CN116256468A (en) 2023-06-13

Family

ID=86679899

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211262759.8A Pending CN116256468A (en) 2022-10-15 2022-10-15 Air quality prediction optimization method based on dynamic self-adaptive particle swarm

Country Status (1)

Country Link
CN (1) CN116256468A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117350175A (en) * 2023-12-04 2024-01-05 河北东医生物科技有限公司 Artificial intelligent ecological factor air environment quality monitoring method and system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117350175A (en) * 2023-12-04 2024-01-05 河北东医生物科技有限公司 Artificial intelligent ecological factor air environment quality monitoring method and system
CN117350175B (en) * 2023-12-04 2024-03-12 河北东医生物科技有限公司 Artificial intelligent ecological factor air environment quality monitoring method and system

Similar Documents

Publication Publication Date Title
CN108280551B (en) Photovoltaic power generation power prediction method utilizing long-term and short-term memory network
CN113538910B (en) Self-adaptive full-chain urban area network signal control optimization method
Bao et al. Spatial–temporal complex graph convolution network for traffic flow prediction
CN112085163A (en) Air quality prediction method based on attention enhancement graph convolutional neural network AGC and gated cyclic unit GRU
CN113554466B (en) Short-term electricity consumption prediction model construction method, prediction method and device
Xu et al. A novel hybrid model for six main pollutant concentrations forecasting based on improved LSTM neural networks
CN113112791A (en) Traffic flow prediction method based on sliding window long-and-short term memory network
CN111260124A (en) Chaos time sequence prediction method based on attention mechanism deep learning
CN114572229B (en) Vehicle speed prediction method, device, medium and equipment based on graph neural network
CN117494034A (en) Air quality prediction method based on traffic congestion index and multi-source data fusion
Wu et al. Traffic prediction based on GCN-LSTM model
CN112257847A (en) Method for predicting geomagnetic Kp index based on CNN and LSTM
CN114970946A (en) PM2.5 pollution concentration long-term space prediction method based on deep learning model and empirical mode decomposition coupling
Wu et al. Prediction of PM2. 5 concentration in urban agglomeration of China by hybrid network model
CN116256468A (en) Air quality prediction optimization method based on dynamic self-adaptive particle swarm
Yuan et al. SA–EMD–LSTM: A novel hybrid method for long-term prediction of classroom PM2. 5 concentration
Ao et al. Hybrid model of air quality prediction using k-means clustering and deep neural network
CN114881358A (en) Air quality prediction method based on adaptive dynamic graph neural network
Sun et al. Time series prediction based on time attention mechanism and lstm neural network
Mo et al. Design a regional and multistep air quality forecast model based on deep learning and domain knowledge
Lu et al. MTGnet: multi-task spatiotemporal graph convolutional networks for air quality prediction
Su et al. Graph ode recurrent neural networks for traffic flow forecasting
Lin et al. Collaborative Framework of Accelerating Reinforcement Learning Training with Supervised Learning Based on Edge Computing
Wang et al. Research on PM2. 5 Pollution Prediction Method in Hefei City Based on CNN-LSTM Hybrid Model
Li et al. Online Attention Enhanced Differential and Decomposed LSTM for Time Series Prediction

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