Regional economy prediction system and method based on space-ground integration technology

Abstract

The invention discloses a regional economy prediction system and method based on a space-ground integration technology, relates to the technical field of satellite remote sensing, and solves the technical problem of limited prediction precision of regional economy caused by lack of fusion of multi-source heterogeneous data in the prior art. A satellite image is obtained through a remote sensing API, and monitoring data is obtained according to the satellite image; preprocessing the monitoring data; calling the preprocessed monitoring data, and judging whether a correlation exists between the monitoring data and the regional economic index or not; if yes, monitoring data enhancement is carried out; if not, corresponding monitoring data are removed, and judgment is continued; calling the enhanced monitoring data, constructing a prediction model to obtain predicted economic values of agriculture, industry and service industry, and performing secondary verification on the predicted economic values through the model to obtain a predicted regional economic total value; according to the invention, more accurate and timely regional economy prediction is provided.

Description

Technical Field

[0001] The present invention belongs to the field of satellite remote sensing, involves machine learning technology, and specifically relates to a regional economic prediction system and method based on space-earth integration technology. Background Art

[0002] As a new heterogeneous data fusion architecture, space-earth integration technology provides a new way of data acquisition and fusion analysis for regional economic analysis by integrating various satellite space-based data resources and ground-based data related to regional economy; introducing remote sensing data into the model is a cutting-edge and potential method. High-resolution, multi-temporal, and wide-coverage remote sensing data can reflect the changes in economic activities in the near-earth space in real time, providing objective and continuous information for monitoring regional economic activities; by combining geographic information system and big data analysis technology, the complex relationship between economic activities and geographical space can be further revealed, providing more abundant and detailed input variables for the modeling of the regional economic system.

[0003] The prior art (a patent application with the publication number of CN112668784A) discloses a regional macroeconomic prediction model and method based on big data, which is used to analyze the initial economic data provided by all economic entities in a region to predict the regional macroeconomy of the region, and the regional macroeconomic prediction model includes a big database, previous generation models, a prediction model, and a training model;

[0004] When analyzing regional economy, the prior art uses traditional economic analysis methods, often relying on historical data and linear models; at the same time, the types of regional economic data itself have obvious heterogeneity, and there are differences in data structure, time span, and data meaning, resulting in multi-source heterogeneous data in regional economy not being effectively analyzed and interpreted, lacking the fusion of multi-source heterogeneous data, and thus the prediction accuracy of regional economy is limited.

[0005] The present invention provides a regional economic prediction system and method based on space-earth integration technology to solve the above technical problems. Summary of the Invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art; for this purpose, the present invention proposes a regional economic prediction system and method based on space-earth integration technology, which is used to solve the technical problem that the prior art lacks the fusion of multi-source heterogeneous data, resulting in limited prediction accuracy of regional economy.

[0007] To achieve the above object, the first aspect of the present invention provides a regional economic prediction system based on space-earth integration technology, including: a data acquisition module, a correlation verification module, and a model construction module;

[0008] Data acquisition module: Obtain satellite images through remote sensing APIs, and acquire monitoring data based on the satellite images; preprocess the monitoring data; among them, the monitoring data includes: industrial data, agricultural data, service industry data, etc.;

[0009] Relevant verification module: Retrieve the preprocessed monitoring data, and determine whether there is a correlation between the monitoring data and the regional economic index; if yes, perform monitoring data enhancement; if no, eliminate the corresponding monitoring data and continue to judge;

[0010] Model construction module: Retrieve the enhanced monitoring data, construct a prediction model to obtain the predicted economic values of agriculture, industry, and service industry, and obtain the predicted total regional economic value through secondary verification of the predicted economic values by the model.

[0011] Preferably, the acquisition of monitoring data based on the satellite images includes:

[0012] Retrieve the satellite images to determine the planting area of the crops in the set agricultural area, and count the corresponding agricultural data of the set agricultural area; among them, the agricultural data includes: daily average temperature, humidity, and precipitation; Obtain the factory building roof clusters in the visible light satellite images within the set industrial area and locate the industrial park through GIS software; acquire industrial data through sensors; among them, the industrial data includes: SO2, CO, heat radiation value, NO, PM, and infrared; Obtain the service industry data within the set area; among them, the service industry data includes: the area of the science and technology innovation park and the number of settled enterprises, the output quantity of applied patents; the radar reflection intensity of the parking lots of large shopping malls and transportation hubs, the throughput of airports; the radar reflection intensity of the cargo yards, the optical images of the yards; the radar reflection intensity of the port container area, the number of ships in the waterway, the port throughput, etc.

[0013] In the present invention, because agricultural production highly depends on natural environmental conditions, temperature, humidity, and precipitation directly affect the growth cycle, development speed, and yield formation of crops, so the temperature, humidity, and precipitation data in the urban area are used to predict the future yields of various crops in the urban area.

[0014] Preferably, the preprocessing of the monitoring data includes:

[0015] Retrieve the monitoring data, and calculate the mean value of the monitoring data; determine whether there is no missing data in the set cycle for the monitoring data; if yes, then determine whether there is duplicate data in the set cycle for the monitoring data; if yes, retain one set of duplicate data and eliminate the remaining corresponding duplicate data; if no, obtain the maximum and minimum values of each monitoring data index;

[0016] If no, use the mean value of the monitoring data as the monitoring data to fill in the missing data;

[0017] Set the abnormal index range, and determine whether the monitoring data exceeds the abnormal index range; if so, eliminate the corresponding monitoring data; if not, continue to judge;

[0018] The setting of the abnormal index range includes: multiplying the maximum value and the minimum value by a set coefficient respectively to obtain the upper limit and the lower limit of the abnormal index; taking the upper limit and the lower limit of the abnormal index as the abnormal index range; wherein, the value range of the set coefficient is greater than or equal to 1.

[0019] It should be noted that the specific value of the set coefficient depends on the accuracy requirements of the staff in this field for the monitoring data.

[0020] In the present invention, since a large number of mechanical equipment and office equipment will be used during factory operation, a large amount of heat will be dissipated and gases will be emitted. At the same time, the decomposition rate of nitrogen oxides is relatively fast, and the generation to decomposition is basically concentrated in the same area, which can exclude the influence of chemical concentration changes caused by chemicals drifting with the wind from other places, and can be used as a sensitivity index for factory operation. Therefore, the heat, aerosol, nitrogen dioxide, and sulfur dioxide concentration changes can be used to statistically measure the operation activity, so as to more accurately predict the industrial GDP.

[0021] Preferably, the judgment of whether there is a correlation between the monitoring data and the regional economic index includes:

[0022] Obtain the preprocessed monitoring data and set the correlation coefficient threshold; mark the i-th data in several types of monitoring data as X i , and the i-th data Y i of the regional economic index corresponding to the corresponding monitoring data, and mark the average value of the corresponding type of monitoring data as The average value of the corresponding type of regional economic index is marked as where i = 1, 2, 3,..., n, and n is a positive integer;

[0023] Calculate the correlation coefficient through the formula ;

[0024] Judge whether the correlation coefficient is greater than the correlation coefficient threshold; if so, perform the Granger causality test; if not, mark that there is no correlation between the corresponding type of monitoring data and the regional economic index.

[0025] Preferably, the performance of the Granger causality test includes:

[0026] Mark the regional economic index lagged by j periods as y t-j , and mark the i-th period of monitoring data after time t as x t-i , and obtain the regression coefficients α i , δ j and λ i; where the regression coefficient of the regional economic index on the monitoring data is denoted as α i , the regression coefficient of the regional economic index on its own lag value is denoted as β i , the regression coefficient of the monitoring data on the regional economic index is denoted as δ j , the regression coefficient of the monitoring data on its own lag value is denoted as λ i ;

[0027] The regional economic index at time t is calculated through the formula ;

[0028] The monitoring data corresponding to time t is calculated through the formula ;

[0029] where, u 1t and u 2t are preset white noises, and it is set that the white noises u 1t and u 2t are uncorrelated;

[0030] When and only when the regional economic index of an industry is the dependent variable, it is judged whether the significance level of the estimated coefficient α i is greater than 0.05;

[0031] If so, the corresponding type of monitoring data and the regional economic index are marked as having no causal relationship;

[0032] If not, when the monitoring data is the dependent variable, it is judged whether the significance level of the estimated coefficient δ j is greater than 0.05; if so, the corresponding type of monitoring data and the regional economic index are marked as having no one-way causal relationship; if not, the corresponding type of monitoring data and the regional economic index are marked as having a one-way causal relationship through Granger causality test.

[0033] Preferably, the enhancement of the monitoring data includes:

[0034] S110: Retrieve the relevant monitoring data, randomly select a real value within the range of the original data ± noise and copy it to the original corresponding row data, and traverse all the monitoring data;

[0035] S120: Process the data with added noise using wavelet transform, and perform multi-level decomposition to obtain the approximation coefficients and detail coefficients at different scales;

[0036] S130: Mark the original sample size as ysy, mark the hidden dimension of the Transformer model as L, and mark the number of layers of the model as h; mark the output dimension of the LSTM model as dh, and mark the input dimension of the LSTM model as dx;

[0037] The sample interpolation number is calculated through the formula YBC = Max(10×4[dh(dh + dx)+dh], 10×12×L×h 2 ) - ysy;

[0038] S140: Determine whether the interpolation sample number is greater than 0; if yes, merge the interpolation samples; if no, go to step S120.

[0039] In the present invention, data enhancement is performed through wavelet transform, which can increase the amount of data with correlation; the approximate coefficient reflects the general change trend of the monitoring data in a relatively long time range, and the detail coefficient reflects the fluctuation of the temperature of the monitoring data in a short time.

[0040] Preferably, constructing the prediction model to obtain the predicted economic value of agriculture includes:

[0041] Retrieve the planting area of crops in the set area and label it as A k , and agricultural data, and use the agricultural data as input to obtain the temperature regression coefficient α through the random forest model kx , the humidity regression coefficient β kx and the precipitation regression coefficient θ kx ; label the type of crop as k, and label the growth stage of the crop as x; T kx is the temperature data of the k-th crop at the x-th growth stage; H kx is the humidity data of the k-th crop at the x-th growth stage; P kx is the precipitation data of the k-th crop at the x-th growth stage; label the market price as P k ; where, k = 1, 2, 3,..., m, and m is a positive integer; x = 1, 2, 3,..., h, and h is a positive integer;

[0042] The predicted agricultural economic value is calculated through the formula .

[0043] Preferably, constructing the prediction model to obtain the predicted economic values of industry and service industry includes:

[0044] Retrieve industrial data, label the heat as RL, label the aerosol content as QR, label the nitrogen dioxide content as ED, and label the sulfur dioxide content as EL; count the area of the set industrial area and label it as GMJ; input the industrial data into the LSTM model to obtain the industrial coefficient;

[0045] The predicted industrial economic value is calculated through the formula ; where, p is the set number of statistical days; p = 1, 2, 3,..., g; and g is a positive integer;

[0046] Among them, y1 is industrial coefficient one, y2 is industrial coefficient two, y3 is industrial coefficient three, and y4 is industrial coefficient four;

[0047] The service industry data is used as input and the model predicted service industry economic value is obtained through random forest model, LSTM model and Transformer model respectively. The average value of the model predicted service industry economic value is taken as the predicted service industry economic value YGDP.

[0048] Preferably, the method of performing secondary verification on the predicted economic value through the model to obtain the predicted regional economic value includes:

[0049] The enhanced monitoring data are input into the random forest model, LSTM model and Transformer model respectively, and the qth model is fitted in turn to obtain the predicted economic value corresponding to the fth type of monitoring data. The sum of the economic values ​​corresponding to the monitoring data is used as the predicted regional economic value; the predicted agricultural economic value, the predicted industrial economic value and the predicted service industry economic value are used as the calibration value of the predicted regional economic value. f ; Wherein, f = 1, 2, 3, ..., l, and l is a positive integer; q = 1, 2, 3;

[0050] By formula Calculate and obtain the verification value of the predicted regional economic total value;

[0051] The model corresponding to the minimum verification value is determined as the optimal model, and the sum of the output results of the optimal model is used as the predicted regional economic total value.

[0052] To achieve the above object, the second aspect of the present invention provides a regional economic forecasting method based on the space-ground integration technology, comprising:

[0053] Obtain satellite images through remote sensing API, and obtain monitoring data based on satellite images; pre-process the monitoring data; the monitoring data includes: industrial data, agricultural data, and service industry data;

[0054] Retrieve the pre-processed monitoring data and determine whether there is a correlation between the monitoring data and the regional economic index; if yes, enhance the monitoring data; if no, remove the corresponding monitoring data and continue to determine;

[0055] Retrieve the enhanced monitoring data and build a prediction model to obtain the predicted economic values ​​of agriculture, industry and services. Use the model to conduct a secondary verification of the predicted economic values ​​to obtain the predicted regional economic total value.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] 1. The present invention combines satellite remote sensing data and meteorological monitoring data, enhancing the depth and breadth of traditional economic analysis methods; integrating space-air information data with ground economic data and combining satellite data and meteorological data can provide more accurate climate and environmental variable inputs for regional economic risk analysis. By establishing a comprehensive prediction model, this integrated modeling method can observe and analyze regional economic activities from multiple dimensions, and then conduct effective risk assessment and prediction; at the same time, the data provided by space-air information technology has the characteristics of high update frequency, wide coverage, and high real-time nature, which are conducive to directly reflecting the immediate impact of economic activities on the environment.

[0058] 2. The present invention uses a machine learning model to continuously learn and optimize model parameters from historical data and real-time data, thereby ensuring the accuracy of prediction results and reflecting the changing trends of economic activities; this adaptive ability of machine learning shows certain advantages, especially when dealing with emergencies; in addition, another key advantage of machine learning technology in regional economic risk prediction is its ability to process multi-source heterogeneous data, and these multi-source data can be integrated into a unified prediction framework, and through algorithm optimization, it helps to predict the fluctuations and potential risks of economic activities; when the economic development status of a certain region fluctuates, the machine learning model can comprehensively integrate space-air data, meteorological data, and economic data to quickly reflect the changes in the regional economy. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0060] Figure 1 It is a schematic diagram of the module relationship included in the present invention;

[0061] Figure 2 It is a schematic diagram of the specific steps for processing the monitoring data of the present invention;

[0062] Figure 3 It is a schematic diagram of the specific steps for constructing the model of the present invention;

[0063] Figure 4 It is a schematic diagram of the process of regional economic prediction of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] See also Figure 1 , the first aspect of the present invention provides a regional economic forecasting system based on the space-ground integration technology, including: a data acquisition module, a related verification module and a model building module;

[0066] Data acquisition module: obtain satellite images through remote sensing API, obtain monitoring data based on satellite images; pre-process the monitoring data; the monitoring data includes: industrial data, agricultural data and service industry data;

[0067] Related verification module: retrieve the pre-processed monitoring data and determine whether there is a correlation between the monitoring data and the regional economic index; if yes, enhance the monitoring data; if no, remove the corresponding monitoring data and continue to determine;

[0068] Model building module: retrieve enhanced monitoring data, build a prediction model to obtain the predicted economic values ​​of agriculture, industry and services, and use the model to conduct secondary verification of the predicted economic values ​​to obtain the predicted regional economic total value.

[0069] See also Figure 2 , the specific steps of monitoring data processing, retrieve satellite images to determine the planting area of ​​crops in the set agricultural area, and count the agricultural data corresponding to the set agricultural area; among them, agricultural data include: daily average temperature, humidity and precipitation; obtain the factory roof cluster in the visible light satellite image in the set industrial area and GIS software to locate the industrial park; obtain industrial data through sensors; among them, industrial data include: SO2, CO, thermal radiation value, NO, PM and infrared; obtain the corresponding service industry data of specific sections in the set area; among them, service industry data include: the area of ​​science and technology park and the number of settled enterprises, the number of patent applications; radar reflection intensity of large shopping mall parking lots and transportation hub parking lots, airport throughput; radar reflection intensity of cargo yards, yard optical images; radar reflection intensity of port container areas, number of ships in the channel and port throughput, etc.;

[0070] Retrieve monitoring data and calculate the mean of monitoring data; determine whether the monitoring data is missing in the set period; if yes, determine whether there is duplicate data in the set period; if yes, retain a set of duplicate data and remove the rest of the corresponding duplicate data; if no, obtain the maximum and minimum values ​​of each monitoring data indicator;

[0071] If not, use the average value of the monitoring data to fill in the missing monitoring data;

[0072] Set the abnormal index range and determine whether the monitoring data exceeds the abnormal index range; if yes, exclude the corresponding monitoring data; if not, continue to judge;

[0073] The setting of the abnormal index range includes: multiplying the maximum value and the minimum value by a set coefficient respectively to obtain the upper limit and the lower limit of the abnormal index; using the upper limit and the lower limit of the abnormal index as the abnormal index range; where the value range of the set coefficient is greater than or equal to 1;

[0074] For example, in an existing Area A, the distribution of green vegetation in a specific area is determined by high-resolution series satellites, Changguang series satellites, and European meteorological satellites; industrial parks are determined based on website materials and on-site investigations, and the industrial parks are located through the blue or red factory roof clusters in the visible light satellite images within the city and GIS software; the transportation industry mainly measures the data of the ring expressway (2 - 3 kilometers) and specific service area parking lots in specific sections, and uses the change in nitrogen dioxide emission concentration to count the data of the ring expressway;

[0075] Check whether there are missing values in the satellite data. If there are, fill in the missing certain type of satellite data in the mean value method; check whether there are duplicate satellite data in the satellite data. If there are, exclude the row of data; obtain the maximum and minimum values of each satellite data index, increase the maximum value by 10% as the upper limit of the abnormal index, and decrease the minimum value by 10% as the lower limit of the abnormal index, and exclude the data beyond the range;

[0076] To ensure the time dimension consistency of satellite data and regional economic index, aggregate the satellite data by month, and at the same time, to avoid the inconsistency of the dimension of different types of data, perform linear normalization processing;

[0077] Through the formula Calculate the correlation coefficient;

[0078] Set the correlation coefficient threshold to 0.6. When the correlation coefficient r exceeds 0.6, it indicates that there is a strong correlation between this type of satellite remote sensing data and the regional economic index.

[0079] In the present invention, since nitrogen dioxide mainly comes from vehicle exhaust and is positively correlated with the traffic flow, and at the same time, the decomposition rate of nitrogen oxides is relatively fast, by detecting the content of nitrogen dioxide, it can be used as a sensitivity index of traffic activity.

[0080] Please refer to Figure 3, The specific steps of model construction are as follows: retrieve the enhanced monitoring data, construct a prediction model to obtain the predicted economic values of agriculture, industry, and service industries. Input the enhanced monitoring data into the random forest model, LSTM model, and Transformer model respectively, and fit the q-th model in turn to obtain the predicted economic value corresponding to the f-th type of monitoring data. Take the sum of the economic values corresponding to the monitoring data as the predicted total regional economic value; take the predicted agricultural economic value, predicted industrial economic value, and predicted service industry economic value as the calibration value z of the predicted total regional economic value. f ; where f = 1, 2, 3, …, l, and l is a positive integer; q = 1, 2, 3.

[0081] Through the formula Calculate the verification value of the predicted total regional economic value.

[0082] Determine the model corresponding to the minimum verification value as the optimal model, and take the sum of the output results of the optimal model as the predicted total regional economic value.

[0083] For example, in the random forest model, randomly draw n samples with replacement from the enhanced monitoring data to form a new training subset; through bootstrap sampling, approximately times of the samples will be selected, and the remaining about 0.368 times of the samples are used as out-of-bag data for model evaluation; in the process of constructing each decision tree, for the split of each node, randomly select m' features (m' < m) in the satellite data as candidate features to increase the diversity between decision trees; for each training subset, use the randomly selected features above to construct a decision tree; the construction process of the decision tree is a recursive process, by continuously selecting the optimal features and split points, dividing the data set into smaller subsets until the stopping condition is met; where the stopping condition is that the depth of the tree reaches the maximum; the splitting criterion is the mean squared error.

[0084] Repeat the above steps K times to construct K decision trees to form a random forest; for predicting a new sample x, input it into each decision tree to obtain K prediction results.

[0085] Through the formula Calculate the predicted economic value of the monitoring data; where h b (x) is the prediction of the b-th decision tree for the sample x; where b = 1, 2, 3, …, B, and B is a positive integer.

[0086] Among them, the predicted industrial economic value of Region A by the random forest model is hundred million yuan, the agricultural economic value is hundred million yuan, and the service industry economic value is hundred million yuan.

[0087] The predicted regional economic total value of the random forest model is 163.195 billion yuan calculated by the formula 5673.78 + 3773.51 + 6872.21;

[0088] Retrieve the LSTM model. The predicted industrial economic value of Region A by the LSTM model is billion yuan, the predicted agricultural economic value is billion yuan, and the predicted service economic value is billion yuan;

[0089] The predicted regional economic total value of the LSTM model is 131.16 billion yuan calculated by the formula 5579.28 + 3923.41 + 3613.31;

[0090] Retrieve the Transformer model. For a numerical data matrix X ∈ R with n samples and m features per sample n×m , map it to a high-dimensional vector space through a linear transformation; let the embedding matrix be where d model is the dimension of the model;

[0091] To enable the model to capture the sequential information of features, positional encoding needs to be added;

[0092] The positional encoding formula is as follows:

[0093] For even dimensions:

[0094] For odd dimensions:

[0095] where pos represents the position of the feature (ranging from 1 to m), and d represents the dimension of the vector (ranging from 0 to d model - 1);

[0096] The final input representation: Z pos = E pos + PE pos ;

[0097] Multi-head attention layer: x attn = MultiHead(Z, Z, Z);

[0098] Feed-forward neural network layer: x ffn = FFN(x attn_out );

[0099] Map the output of the decoder to the target dimension through a linear layer; let and b out ∈ R be the weight matrix and bias vector of the output layer, then the predicted value is: The industrial economic value of region A is predicted by the Transformer model. billion yuan, and the agricultural economy is billion yuan, and the economic value of the service industry is 100 million yuan;

[0100] The Transformer model predicts the regional economic value of 1,413.528 billion yuan through the formula 6526.22+4315.15+3293.91;

[0101] The predicted industrial economic value is obtained as 5963.55, the predicted agricultural economic value is 4936.21 and the predicted service industry economic value is 3299.56;

[0102] but The calculated validation value of the predicted regional economic total value corresponding to the random forest model is 412.41;

[0103] The calculated verification value of the predicted regional economic total value corresponding to the LSTM model is 375.92;

[0104] The calculated validation value of the predicted regional economic total value corresponding to the Transformer model is 279.35;

[0105] Since 517.52>375.92>279.35, the Transformer model is taken as the optimal model, and the sum of the output results of the optimal model, i.e., 1,413.528 billion yuan, is taken as the predicted regional economic value of region A.

[0106] See also Figure 4 The second aspect of the present invention provides a method for regional economic forecasting based on the space-ground integration technology, including:

[0107] Obtain satellite images through remote sensing API, and obtain monitoring data based on satellite images; pre-process the monitoring data; the monitoring data includes: industrial data, agricultural data, and service industry data;

[0108] Retrieve the pre-processed monitoring data and determine whether there is a correlation between the monitoring data and the regional economic index; if yes, enhance the monitoring data; if no, remove the corresponding monitoring data and continue to determine;

[0109] Retrieve the enhanced monitoring data and build a prediction model to obtain the predicted economic values ​​of agriculture, industry and services. Use the model to conduct a secondary verification of the predicted economic values ​​to obtain the predicted regional economic total value.

[0110] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0111] The working principle of the present invention is as follows: the present invention obtains satellite images through a remote sensing API, and obtains monitoring data based on the satellite images; pre-processes the monitoring data; retrieves the pre-processed monitoring data, and determines whether there is a correlation between the monitoring data and the regional economic index; if so, enhances the monitoring data; if not, removes the corresponding monitoring data, and continues to judge; retrieves the enhanced monitoring data, constructs a prediction model to obtain the predicted economic values ​​of agriculture, industry and service industries, and performs secondary verification of the predicted economic values ​​through the model to obtain the predicted regional economic total value.

[0112] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A regional economic forecasting system based on space-ground integration technology, characterized in that: include: Data collection module, relevant verification module and model building module; Data acquisition module: obtain satellite images through remote sensing API and obtain monitoring data based on satellite images; Preprocessing the monitoring data; the monitoring data includes: industrial data, agricultural data and service industry data; Related verification module: retrieve the pre-processed monitoring data and determine whether there is a correlation between the monitoring data and the regional economic index; if yes, enhance the monitoring data; if no, remove the corresponding monitoring data and continue to determine; Model building module: retrieve enhanced monitoring data, build a prediction model to obtain the predicted economic values ​​of agriculture, industry and services, and use the model to conduct secondary verification of the predicted economic values ​​to obtain the predicted regional economic total value.

2. The regional economic forecasting system based on the space-ground integration technology according to claim 1 is characterized in that: The obtaining of monitoring data based on satellite images includes: Retrieve satellite images to determine the planting area of ​​crops in the set agricultural area, and collect statistics on the agricultural data corresponding to the set agricultural area; obtain the factory roof clusters in the visible light satellite images in the set industrial area and use GIS software to locate the industrial park; obtain industrial data through sensors; and obtain the corresponding service industry data in the set area.

3. The regional economic forecasting system based on the space-ground integration technology according to claim 1 is characterized in that: The preprocessing of the monitoring data includes: Retrieve monitoring data and calculate the mean value of the monitoring data; Determine whether the monitoring data is missing in the set period; If yes, determine whether there is duplicate data in the set period of the monitoring data; if yes, retain a set of duplicate data and remove the remaining corresponding duplicate data; if no, obtain the maximum and minimum values ​​of each monitoring data indicator; If not, the mean of the monitoring data will be used as the missing monitoring data; Set the abnormal index range and judge whether the monitoring data exceeds the abnormal index range; if yes, remove the corresponding monitoring data; if no, continue to judge; The setting of the abnormal index range includes: multiplying the maximum value and the minimum value by the set coefficient respectively to obtain the abnormal index upper limit and the abnormal index lower limit; taking the abnormal index upper limit and the abnormal index lower limit as the abnormal index range; wherein the value range of the set coefficient is greater than or equal to 1.

4. The regional economic forecasting system based on the space-ground integration technology according to claim 1 is characterized in that: The determination of whether there is a correlation between the monitoring data and the regional economic index includes: Obtain the preprocessed monitoring data and set the correlation coefficient threshold; mark the i-th data in several types of monitoring data as X i , the i-th data Y of the regional economic index corresponding to the corresponding monitoring data i , the average value of the corresponding type of monitoring data is marked as The average value of the regional economic index of the corresponding type is marked as Wherein, i=1, 2, 3, ..., n, and n is a positive integer; By formula The correlation coefficient was calculated; Determine whether the correlation coefficient is greater than the correlation coefficient threshold; if yes, perform Granger causality test; if no, mark the corresponding type of monitoring data and the regional economic index as having no correlation.

5. The regional economic forecasting system based on the space-ground integration technology according to claim 4 is characterized in that: The Granger causality test includes: The regional economic index of the j-period lag is marked as y t-j , mark the i-th monitoring data after time t as x t-i , obtain the regression coefficient α through third-party software i , δ j and i ; Among them, α i is the regression coefficient of the regional economic index on the monitoring data, β i is the regression coefficient of the regional economic index to its own lagged value, δ j is the regression coefficient of the monitoring data on the regional economic index, λ i is the regression coefficient of the monitoring data on its own lagged value; By formula Calculate the regional economic index at time t; By formula Calculate the monitoring data corresponding to time t; Among them, u 1t and u 2t is the preset white noise, and the white noise u is set 1t and u 2t Not relevant; If and only if there is an industry's regional economic index as the dependent variable, the estimated coefficient α i Whether the significance level is greater than 0.05; If yes, the corresponding type of monitoring data and the regional economic index are marked as having no causal relationship; If not, then when the monitoring data is the dependent variable, the estimated coefficient δ j Whether the significance level is greater than 0.05; if yes, the corresponding type of monitoring data and the regional economic index are marked as having no one-way causal relationship; if no, the corresponding type of monitoring data and the regional economic index are marked as having a one-way causal relationship, and the Granger causality test is passed.

6. The regional economic forecasting system based on the space-ground integration technology according to claim 1 is characterized in that: The monitoring data enhancement includes: S110: Retrieve relevant monitoring data, randomly select a real value in the interval of original data ± noise, and copy it to the original corresponding row data, traversing all monitoring data; S120: using wavelet transform to process the data after adding noise, and performing multi-level decomposition to obtain approximate coefficients and detail coefficients at different scales; S130: mark the original sample size as ysy, the hidden dimension of the Transformer model as L, and the number of layers of the model as h; mark the output dimension of the LSTM model as dh, and the input dimension of the LSTM model as dx; By the formula YBC = Max (10 × 4 [dh (dh + dx) + dh], 10 × 12 × L × h 2 )-ysy calculates the sample interpolation number; S140: Determine whether the number of interpolation samples is greater than 0; if yes, merge the interpolation samples; if no, go to step S120.

7. The regional economic forecasting system based on the space-ground integration technology according to claim 1 is characterized in that: The method of constructing a prediction model to obtain the predicted economic value of agriculture includes: Retrieve the planting area of ​​the crop in the set area and mark it as A k , and agricultural data, using the agricultural data as input to obtain the temperature regression coefficient α through the random forest model kx , humidity regression coefficient β kx and the precipitation regression coefficient θ kx ; Mark the crop type as k and the crop growth stage as x; T kx is the temperature data of the kth crop in the xth growth period; H kx is the humidity data of the kth crop in the xth growth period; P kx is the precipitation data of the kth crop in the xth growing period; the market price is marked as P k ; Wherein, k = 1, 2, 3, ..., m, and m is a positive integer; x = 1, 2, 3, ..., h, and h is a positive integer; By formula The predicted agricultural economic value is calculated.

8. The regional economic forecasting system based on the space-ground integration technology according to claim 1 is characterized in that: The construction of the prediction model to obtain the predicted economic value of the industry and service industry includes: Retrieve industrial data, mark the heat as RL, the aerosol content as QR, the nitrogen dioxide content as ED, and the sulfur dioxide content as EL; calculate the area of ​​the set industrial area and mark it GMJ; input the industrial data into the LSTM model to obtain the industrial coefficient; By formula The predicted industrial economic value is calculated; wherein p is the set statistical day; p = 1, 2, 3, ..., g, and g is a positive integer; Among them, y1 is industrial coefficient one, y2 is industrial coefficient two, y3 is industrial coefficient three, and y4 is industrial coefficient four; The service industry data is used as input and the model predicted service industry economic value is obtained through random forest model, LSTM model and Transformer model respectively. The average of the model predicted service industry economic value is taken as the predicted transportation service industry economic value YGDP.

9. The regional economic forecasting system based on the space-ground integration technology according to claim 1 is characterized in that: The method of performing secondary verification on the predicted economic value through the model to obtain the predicted regional economic value includes: The enhanced monitoring data are input into the random forest model, LSTM model and Transformer model respectively, and the qth model is fitted in turn to obtain the predicted economic value corresponding to the fth type of monitoring data. The sum of the economic values ​​corresponding to the monitoring data is used as the predicted regional economic value; the predicted agricultural economic value, the predicted industrial economic value and the predicted service industry economic value are used as the calibration value of the predicted regional economic value. f ; Among them, f = 1, 2, 3; q = 1, 2, 3; By formula Calculate and obtain the verification value of the predicted regional economic total value; The model with the smallest verification value is determined to be the optimal model, and the sum of the output results of the optimal model is used as the predicted regional economic total value.

10. A regional economic forecasting method based on the space-ground integration technology, applied to the regional economic forecasting system based on the space-ground integration technology according to any one of claims 1 to 9, characterized in that: include: Obtain satellite images through remote sensing API and obtain monitoring data based on satellite images; Preprocessing the monitoring data; the monitoring data includes: industrial data, agricultural data and service industry data; Retrieve the pre-processed monitoring data and determine whether there is a correlation between the monitoring data and the regional economic index; if yes, enhance the monitoring data; if no, remove the corresponding monitoring data and continue to determine; Retrieve the enhanced monitoring data, build a prediction model to obtain the predicted economic values ​​of agriculture, industry and services, and use the model to conduct a secondary verification of the predicted economic values ​​to obtain the predicted regional economic total value.

Images