CN114626586A - Large-scale building energy consumption prediction method based on prophet-LightGBM hybrid model - Google Patents
Large-scale building energy consumption prediction method based on prophet-LightGBM hybrid model Download PDFInfo
- Publication number
- CN114626586A CN114626586A CN202210190686.XA CN202210190686A CN114626586A CN 114626586 A CN114626586 A CN 114626586A CN 202210190686 A CN202210190686 A CN 202210190686A CN 114626586 A CN114626586 A CN 114626586A
- Authority
- CN
- China
- Prior art keywords
- energy consumption
- model
- data
- lightgbm
- prophet
- 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
Links
Images
Classifications
-
- 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
- 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
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
The invention belongs to the technical field of building energy consumption prediction, and particularly relates to a large-scale building energy consumption prediction method based on a prophet-LightGBM hybrid model. The method comprises the following steps: acquiring energy consumption data of a large-scale building group and hourly meter reading data of various characteristic variables to form a data set, and preprocessing the data set; and (3): 1, dividing the data into a test set and a training set in proportion; respectively analyzing the energy consumption data and predicting the energy consumption value in a future period of time by adopting a prediction method of prophet and LightGBM, and then solving the corresponding coefficient of the hybrid model by adopting a Particle Swarm Optimization (PSO) algorithm so as to optimize the prediction model and obtain the prophet-LightGBM hybrid model; and the prediction accuracy is evaluated through RMSE and MAPE, and then compared with the result predicted by using a single model, the method has higher prediction accuracy.
Description
Technical Field
The invention belongs to the technical field of building energy consumption prediction, and particularly relates to a large-scale building energy consumption prediction method based on a prophet-LightGBM hybrid model.
Background
In recent years, urban building groups tend to be dense, and the scale-to-grade ratio of super high-rise buildings is high. The changes have higher requirements on building service and power supply stability, and large-scale high-integration-degree building groups have higher and higher requirements on electric quantity. The reasonable planning and scientific scheduling of the power system need to predict the energy consumption data at the next moment in advance, so the importance of the prediction of the building energy consumption data is increasingly highlighted. In order to ensure the healthy development of the ultra-large city and the scientific scheduling of the smart grid, the prediction of the power consumption and the improvement of the prediction accuracy are very important.
The energy consumption prediction method is essentially to predict the corresponding numerical value of a future time sequence, and find the period of time sequence change and the correlation between characteristic variables by mining the evolution characteristics of known time sequence data, so as to predict the future trend of the time sequence, such as the change peak value of energy consumption, daily redeeming purchase amount of fund, the traffic flow of a city main road and the like. The method has the advantages that the method finds and predicts the future change trend by analyzing and mining the characteristic data of the past time sequence, has great significance in the field of building energy conservation and emission reduction, and particularly has more and more importance in the field of building energy conservation after the proportion of building energy consumption in China is higher and higher, and the strategic target tasks of carbon peak reaching and carbon neutralization are provided.
Due to the fact that building energy consumption has complex randomness, an accurate prediction result cannot be easily obtained, and in recent years, methods for predicting energy consumption include: the method comprises three aspects of an engineering method, statistical regression and artificial intelligence, wherein the engineering method uses a thermophysical equation to accurately input conditions of the environment, operation, air conditioning equipment and the like of a building, a prediction result obtained by the method through a complex simulation tool is quite effective and accurate, but a plurality of difficulties exist in the aspect of practical application, a plurality of accurate parameters are difficult to obtain, a huge amount of labor cost and time cost are required to be consumed for complex modeling, and the use convenience is greatly improved. The statistical regression method is characterized in that energy consumption data is associated with variables influencing the energy consumption data, energy consumption is predicted by simplifying the variables and estimating important parameters used by the energy, a statistical regression model is easier to design, but the statistical regression model lacks robustness, so that the prediction precision is not accurate enough, and the engineering requirements can not be met frequently. As for the artificial intelligence method, the method is easy to use, can perform real-time prediction according to relevant variables such as environmental climate, building characteristics and the like, has better prediction performance compared with a statistical regression method, has lower requirements on physical parameters for reducing prediction cost compared with an engineering method, and is more and more favored by researchers in recent years, but a single neural network model is easily affected by model defects to cause generalization capability or the prediction accuracy cannot be further improved, for example, a gradient dissipation phenomenon exists in a long-time prediction process of an RNN algorithm; algorithms such as a support vector machine, KNN and the like cannot obtain a good data result under mass data. The hybrid prediction model is particularly important at this moment, and the advantages of a plurality of algorithms can be combined to improve the prediction accuracy.
Disclosure of Invention
In view of the above problems, the present invention aims to provide a large-scale building energy consumption prediction method based on a prophet-LightGBM hybrid model with high prediction accuracy.
The invention carries out fitting on two models of LightGBM (Ke G, Meng Q, Finley T, et al. Lightgbm: A high-level effective gradient boosting tree [ C ]// advanced in Neural Information Processing systems [ 2017: 3146;, distributed gradient boosting frame based on decision tree algorithm) and propheter (S.J.Taylor and B.Letham, "detecting at scale," The American State optimization, vol.72, No.1, pp. 37-45,2018. [ Online ]. Available: htps:// doi.org/10.1080/00031305.2017.1380080, previously known model), and The combination coefficient of The two models is calculated by utilizing a particle swarm optimization algorithm to improve The prediction accuracy, thereby fully playing The advantages of The different models in The aspect of time sequence prediction.
The invention provides a large-scale building energy consumption prediction method based on a prophet-LightGBM hybrid model, which comprises the following specific steps:
(1) acquiring energy consumption data of a large-scale building group and hourly meter reading data of various characteristic variables to form a data set, and preprocessing the data set;
(2) and (3): 1, dividing data in proportion to serve as a test set and a training set;
(3) respectively analyzing the energy consumption data and predicting the energy consumption value in a future period of time by adopting a priori knowledge model (prophet) and a LightGBM prediction method; then, solving corresponding coefficients of the mixed model by adopting a Particle Swarm Optimization (PSO) algorithm, thereby optimizing the prediction model and obtaining a prophet-LightGBM mixed model;
(4) by root mean square error RMSE:
mean absolute percent error MAPE:
and evaluating the prediction accuracy, and comparing with the result predicted by using a single model.
In the step (1), the characteristic variables include 14 characteristic variables, namely, a region number, a building number, a region purpose, a building area, a building age, a building floor number, an energy consumption meter reading, time and date, outdoor temperature, cloud cover degree, dew point temperature, atmospheric pressure, wind direction and wind speed.
The data preprocessing comprises analyzing the data characteristic variable relation, filling the missing historical data numerical values in the data set, and adopting a filling mode to copy the data at the previous moment or supplement the data by using a linear interpolation method according to the missing time.
In the step (3), the prophet model is used for fitting the energy consumption trend and the numerical value of a period of time in the future, inputting the energy consumption numerical value of the historical meter reading and a time sequence corresponding to the energy consumption numerical value, inputting the time length to be predicted, and obtaining energy consumption data predicted in the future through model training; the model is formed as follows:
y(t)＝g(t)+s(t)+h(t)+∈t；
wherein y (t) is a predicted energy consumption value; g (t) is a trend term in the model, s (t) is a period term, h (t) is a holiday factor, epsilontIs an error term.
Trend term g (t):
in the formula, C (t) is a value which changes along with time and indicates the maximum allowable value of the environment, namely the maximum allowable electric quantity of the power grid;
because the energy consumption of a building is influenced by the climate temperature to a great extent, the energy consumption value can change along with the change of seasonal or crowd circadian rhythms such as the date, and the period term s (t) is in the form of:
in the formula: p is the period length of the time sequence, and N is the period number; a isn，bnIs the parameter to be estimated;
the building energy consumption obviously belongs to an abnormal value in the holiday period, the building power consumption of market residents can be increased rapidly, office buildings can be reduced, and therefore analysis on holidays is also particularly important, and the holiday factor h (t) is in the basic form:
in the formula: l is the number of holidays, kiThe influence range is festival and holiday.
The LightGBM model firstly analyzes the characteristic relation, and utilizes a mutual exclusion characteristic binding algorithm to fuse and bind to reduce the quantity of characteristic vectors so as to achieve the purpose of dimension reduction; the characteristic variable is conveniently discretized by adopting a histogram algorithmOperation is carried out, so that the model is more stable and the robustness is improved; adopting a strategy of inhibiting the growth depth of the decision tree to prevent overfitting on the basis of histogram processing; then, a unilateral gradient sampling algorithm is used for eliminating small gradient samples, specifically, the absolute value of the split characteristic values is taken and then is ranked from large to small, and the maximum value x is selectedmaxAnd randomly taking y small data, and then processing the data as follows:
the algorithm is more concerned with the samples with insufficient training, and finally, y + x is usedmaxThe information gain is calculated for each data.
After prediction results of a prophet model and a LightGBM model are obtained, the Particle Swarm Optimization (PSO) algorithm is used for fitting and optimizing the two models, so that the prediction accuracy is improved, and the advantages of different models are brought into play. Firstly, the maximum operation times of the algorithm and the position change speed v during the iterative operation are seti(k) Position xi(k) And a fitness function:
the iteration formula and the position formula are as follows:
wherein w is an inertia factor, c1，c2To accelerate constant, piIs the minimum distance of the monomers, pgIs the global minimum distance, k is the number of iteration rounds, i is the number of particles, Q (t) is the prophet model coefficient in alpha, the LightGBM model coefficient in beta,
is a predicted value of the prophet model,
LightGBM model predictor, ytAre true values.
In the step (4), the evaluated error evaluation criterion is as follows:
root mean square error RMSE:
mean absolute percent error MAPE:
in the formula: y isiThe number of the energy consumption is recorded;
predicting an energy consumption value for the model over a future period of time; and m is the number of predicted values under different time resolutions.
And comparing the estimated accuracy with the prediction accuracy of a single model.
Drawings
FIG. 1 is a flowchart of a prophet model prediction method of the present invention.
Fig. 2 is a flowchart of a LightGBM model prediction method according to the present invention.
FIG. 3 is a flowchart of a prophet-LightGBM hybrid model prediction method of the present invention.
FIG. 4 is a diagram illustrating energy consumption prediction results of different models at 24h time resolution a day according to an embodiment of the present invention. Wherein, (a) is prophet model, (b) is LightGBM model, (c) is prophet-LightGBM mixed model, and (d) the comparison of each model and actual value is integrated.
Detailed Description
The large-scale building energy consumption prediction method based on the prophet-LightGBM hybrid model, as shown in FIGS. 1-3, includes:
the prophet model prediction method comprises the following steps:
s1, acquiring a data set:
the data used in the case can be publicly accessed through https:// www.kaggle.com/submit 261124/ashrae-great-energy-predictor-iii-dataset, the building type of the dataset covers the values of three-year (2016-2018) meter reading per hour of thousands of buildings from various different places around the world, such as education, shopping malls, restaurants, medical treatment, civil houses, factories, office buildings, public services, worship halls, warehouses, and the like; the data set collects energy consumption data of 1449 buildings of different countries of different continents in the world in three years and meteorological data from airports or meteorological stations of different buildings which are nearest, wherein the meteorological data and the energy consumption data record 76159440 groups of data once every hour; the dataset variables are shown in table 1;
table 1 data set variable information
| Number of features | Variables of | Description of the |
| 1 | Site_id | Region numbering |
| 2 | Building_id | Building number |
| 3 | Primary_use | Regional use |
| 4 | square_feet | Building area m2 |
| 5 | year_built | Construction age limit |
| 6 | floor_count | Number of stories in building |
| 7 | meter_reading | Energy consumption meter reading (kw h) |
| 8 | timestamp | Time and date Year/month/day Hour/ |
| 9 | air_temperature | Outdoor temperature (weather station) ° |
| 10 | cloud_coverage | Coverage of cloud layer |
| 11 | dew_temperature | Dew point temperature (weather station) °c |
| 12 | sea_level_pressure | Atmospheric pressure (weather station) mmHg |
| 13 | wind_direction | Wind direction (weather station) (0-360 degree) |
| 14 | wind_speed | Wind speed (weather station) (m/s) |
S2, data processing: analyzing the internal relation among the characteristic variables, filling up the missing historical data values in the data set, eliminating abnormal values and cleaning the original data;
s3, dividing the data set: and (3): 1, dividing the data;
s4, adopting a prophet model to fit data for prediction;
and S5, judging the prediction accuracy through error analysis, and further adjusting the parameter optimization model.
The LightGBM model prediction method comprises the following steps:
s1, acquiring a data set, as above;
s2, data processing: analyzing the internal relation among the characteristic variables, filling up the missing historical data values in the data set, eliminating abnormal values and cleaning the original data;
s3, dividing the data set, the same as above;
s4, training and modeling the training set by adopting a LightGBM model to generate prediction data;
and S5, judging the prediction accuracy through error analysis, and further adjusting the parameter optimization model.
The prophet-LightGBM mixed model prediction method comprises the following steps:
s1, acquiring the data set, as above;
s2, data processing: analyzing the internal relation among the characteristic variables, filling up the missing historical data values in the data set, eliminating abnormal values and cleaning the original data;
s3, dividing the data set, the same as above;
s4, training and modeling the training set by adopting a LightGBM model to generate prediction data;
s5, adopting a prophet model to fit data for prediction;
s6, constructing a corresponding fitness function for the prediction results obtained in the steps 4 and 5:
optimally solving the coefficients of the hybrid model by adopting a PSO algorithm to obtain alpha and beta coefficients 0.1303 and 0.8433;
s7, calculating the prediction result of the hybrid model;
s8, judging the prediction accuracy through error analysis, using RMSE and MAPE, and comparing the prophet-LightGBM mixed model with the real values of the prophet model, the LightGBM model and the power consumption, and the comparison results of the models without the same indexes are listed in Table 2.
TABLE 2 comparison of evaluation results of the respective models
| Model (model) | RMSE | MAPE |
| Prophet | 13.57 | 11.58 |
| LightGBM | 12.22 | 10.20 |
| prophet-LightGBM | 10.45 | 8.76 |
According to the large-scale building energy consumption prediction method based on the prophet-LightGBM hybrid model, the prediction precision can be improved, and the intelligent scheduling of the power grid can be better assisted.
The above-mentioned technical details and the flow of steps implemented by the algorithm are only used as an illustrative example to better illustrate the method of the present invention, and should not be understood as a limitation of the present invention, and other researchers in the field can make various modifications and combinations within the scope of the present invention, and the modifications and combinations are still within the scope of the present invention.
Claims (7)
1. A large-scale building energy consumption prediction method based on a prophet-LightGBM hybrid model is characterized by comprising the following specific steps:
(1) acquiring energy consumption data of a large-scale building group and hourly meter reading data of various characteristic variables to form a data set, and preprocessing the data set;
(2) and (3): 1, dividing data in proportion to serve as a test set and a training set;
(3) respectively analyzing the energy consumption data and predicting the energy consumption value in a future period of time by adopting a prophet and LightGBM prediction method, and then solving a corresponding coefficient of the hybrid model by adopting a Particle Swarm Optimization (PSO) algorithm so as to optimize the prediction model and obtain a prophet-LightGBM hybrid model;
(4) the prediction accuracy is evaluated by RMSE and MAPE, and the mixed model prediction results are compared with the results predicted using a single model.
2. The method for predicting building energy consumption according to claim 1, wherein the characteristic variables in the step (1) include 14 characteristic variables of a region number, a building number, a region use, a building area, a building age, a building number of floors, an energy meter, time and date, outdoor temperature, cloud coverage, dew point temperature, atmospheric pressure, wind direction and wind speed.
3. The building energy consumption prediction method according to claim 1, characterized in that the preprocessing of the data in step (1) comprises: analyzing the relation of data characteristic variables, filling the missing historical data values in the data set, and copying the data at the last moment or supplementing the data by using a linear interpolation method according to the missing time in a filling mode.
4. The building energy consumption prediction method according to claim 1, wherein the prophet model in the step (3) is obtained by fitting an energy consumption trend and a value in a period of time in the future, inputting an energy consumption value of a historical meter reading and a time sequence corresponding to the energy consumption value, inputting a time length to be predicted, and obtaining energy consumption data predicted in the future through model training; the model is formed as follows:
y(t)＝g(t)+s(t)+h(t)+∈t，
wherein y (t) is a predicted energy consumption value; g (t) is a trend term in the model, s (t) is a period term, h (t) is a holiday factor, etIs an error term;
the trend term g (t) is:
in the formula, C (t) is a value which changes along with time and indicates the maximum allowable value of the environment, namely the maximum allowable electric quantity of the power grid;
the period term s (t) is:
in the formula: p is the period length of the time sequence, and N is the period number; a isn，bnIs the parameter to be estimated;
holiday factors h (t) are:
in the formula: l is the number of holidays, kiThe influence range is festival and holiday.
5. The building energy consumption prediction method according to claim 4, wherein in the step (3), the LightGBM model firstly analyzes the feature relationship, and uses a mutual exclusion feature binding algorithm to fuse and bind the feature vectors to reduce the number of the feature vectors, thereby achieving the purpose of dimension reduction; the histogram algorithm is adopted to discretize the characteristic variables, so that the operation is convenient, the model is more stable, and the robustness is increased; adopting a strategy of inhibiting the growth depth of the decision tree to prevent overfitting on the basis of histogram processing; then, a unilateral gradient sampling algorithm is used for eliminating small gradient samples, specifically, the absolute value of the split characteristic values is taken and then is ranked from large to small, and the maximum value x is selectedmaxAnd randomly taking y smaller data, and then processing the smaller data as follows:
the algorithm is more concerned with the samples with insufficient training, and finally, y + x is usedmaxThe information gain is calculated for each data.
6. The building energy consumption prediction method according to claim 1, wherein the PSO algorithm is optimized by the particle swarm in the step (3), and after prediction results of the prophet model and the LightGBM model are obtained, the PS is usedPerforming fitting optimization on two models by using an O algorithm; firstly, the maximum operation times of the algorithm and the position change speed v during the iterative operation are seti(k) Position xi(k) And a fitness function:
the iteration formula and the position formula are as follows:
vi(k+1)＝wvi(k)+c1r1(pi-xi(k))+c2r2(pg-xi(k))
xi(k+1)＝xi(k)+vi(k+1)
wherein w is an inertia factor, c1，c2To accelerate constant, piIs the minimum distance of the monomers, pgIs the global minimum distance, k is the number of iteration rounds, i is the number of particles, Q (t) is the prophet model coefficient in alpha, the LightGBM model coefficient in beta,
in order to predict the value of the prophet model,
LightGBM model predictor, ytAre true values.
7. The building energy consumption prediction method according to claim 1, characterized in that the error evaluation criterion evaluated in step (4) is:
in the formula: y isiThe number of the energy consumption is recorded;
predicting an energy consumption value for the model over a future period of time; and m is the number of predicted values at different time resolutions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210190686.XA CN114626586A (en) | 2022-02-28 | 2022-02-28 | Large-scale building energy consumption prediction method based on prophet-LightGBM hybrid model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210190686.XA CN114626586A (en) | 2022-02-28 | 2022-02-28 | Large-scale building energy consumption prediction method based on prophet-LightGBM hybrid model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114626586A true CN114626586A (en) | 2022-06-14 |
Family
ID=81900706
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210190686.XA Pending CN114626586A (en) | 2022-02-28 | 2022-02-28 | Large-scale building energy consumption prediction method based on prophet-LightGBM hybrid model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114626586A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115471013A (en) * | 2022-10-19 | 2022-12-13 | 海南港航控股有限公司 | Passenger flow prediction system based on Prophet-PSO model |
| CN117236703A (en) * | 2023-11-15 | 2023-12-15 | 青岛民航凯亚系统集成有限公司 | Airport service system resource safety margin analysis method and system |
-
2022
- 2022-02-28 CN CN202210190686.XA patent/CN114626586A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115471013A (en) * | 2022-10-19 | 2022-12-13 | 海南港航控股有限公司 | Passenger flow prediction system based on Prophet-PSO model |
| CN117236703A (en) * | 2023-11-15 | 2023-12-15 | 青岛民航凯亚系统集成有限公司 | Airport service system resource safety margin analysis method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Johari et al. | Urban building energy modeling: State of the art and future prospects | |
| CN108701274B (en) | Urban small-scale air quality index prediction method and system | |
| CN107045673B (en) | Public bicycle flow variation prediction method based on stack model fusion | |
| CN114626586A (en) | Large-scale building energy consumption prediction method based on prophet-LightGBM hybrid model | |
| CN112288164B (en) | Wind power combined prediction method considering spatial correlation and correcting numerical weather forecast | |
| CN104484715A (en) | Neural network and particle swarm optimization algorithm-based building energy consumption predicting method | |
| CN105184455A (en) | High dimension visualized analysis method facing urban electric power data analysis | |
| CN109272364A (en) | Automatic Valuation Modelling modeling method | |
| CN106910144B (en) | Large building time-by-time energy consumption online prediction method based on time-division actual energy consumption coefficient | |
| CN102495937A (en) | Prediction method based on time sequence | |
| CN110619432A (en) | Feature extraction hydrological forecasting method based on deep learning | |
| CN114429053B (en) | Basin scale WEFE system adaptability simulation optimization method | |
| CN108846501A (en) | A kind of low influence development facility construction scale of rainwater determines method | |
| CN105184388A (en) | Non-linear regression method for urban power load short-period prediction | |
| Ali et al. | GIS-based residential building energy modeling at district scale | |
| CN115507822A (en) | Flood risk prediction method under hydrographic cyclic variation drive | |
| CN113128055B (en) | Distributed hydrological model space calibration method based on runoff yield coefficient | |
| CN111310257A (en) | Regional building energy consumption prediction method under BIM environment | |
| CN116796904A (en) | Method, system, electronic equipment and medium for predicting new line passenger flow of rail transit | |
| CN113033110A (en) | Important area personnel emergency evacuation system and method based on traffic flow model | |
| CN106600055A (en) | Wind speed prediction method the basis of self excitation threshold autoregression model | |
| CN113743659B (en) | Urban layout prediction method based on component method and Markov cellular automaton and application | |
| CN116562583A (en) | Multidimensional water resource supply and demand prediction method and system | |
| CN111310979A (en) | House rent price prediction method based on machine learning and multi-source information | |
| CN113537596A (en) | Short-time passenger flow prediction method for new line station of urban rail transit |
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 |