Low-altitude weather early warning method and early warning system based on binary Logistic regression

Abstract

The invention discloses a low-altitude weather early warning method and early warning system based on binary Logistic regression, and the method comprises the steps: obtaining training group data and test group data through the processing of meteorological data in previous years, obtaining a regression coefficient through building a training model, employing the training group data, employing a likelihood function estimation algorithm and a Newton iteration method in sequence, and removing a constant term. Obtaining a weather prediction preliminary model formula; and detecting the weather prediction preliminary model formula through test group data, judging whether the prediction success rate reaches a set standard or not, and outputting a weather prediction model formula after the prediction success rate reaches the set standard. The early training group data and the detection group data are a large amount of data, and the prediction accuracy is further improved through the likelihood function estimation algorithm and the Newton iteration method, so that the accuracy of the result obtained by the weather prediction model formula is high.

Description

Technical Field

[0001] The present invention belongs to the technical field of low-altitude weather warning, and in particular relates to a low-altitude weather warning method and warning system based on binary logistic regression. Background Art

[0002] Currently, low-altitude meteorological factors vary dramatically. Low clouds, thunderstorms, low-level wind shear, cloud-covered mountains, low visibility, turbulent terrain, and ice accumulation pose a serious threat to low-altitude flight safety. Currently, my country's low-altitude meteorological services, both in terms of system and technical equipment, cannot compare with those of developed countries with advanced general aviation capabilities. Low-altitude flight meteorological services are limited in scope and lack specialized meteorological products. Significant meteorological factors affecting low-altitude flight include wind speed, precipitation, temperature, and visibility. Excessive wind speed and precipitation can pose safety risks to eVTOL aircraft. Low visibility in low-altitude airspace significantly degrades line-of-sight between aircraft. Although future air taxi development will increasingly adopt unmanned modes, relying on collaborative obstacle avoidance algorithms to plan flight paths, low visibility during takeoff and landing, particularly at takeoff and landing sites, still poses significant safety risks, posing a potential threat to flight safety. Air taxis all use eVTOL (electric vertical take-off and landing aircraft). Too low temperature will cause the durability of the aircraft battery to decrease and still pose a safety hazard.

[0003] Currently, research on meteorological warnings, both domestically and internationally, focuses primarily on full-airspace weather forecasts and manual analysis-based warnings. In China, Gao Xue, in "Research on a GIS-Based Agricultural Meteorological Disaster Monitoring and Warning System," proposed a meteorological disaster warning method that analyzes acquired digital images and meteorological elements, employing support vector machines to invert weather hazards and phenomena, thereby verifying and revising the warning results displayed as disaster levels. Xie Fei, in "Design and Application of a Multi-Scenario Urban Rainfall and Waterlogging Forecasting and Warning Model," developed a real-time short-term rainfall forecasting method. This method dynamically adjusts regional danger levels based on short-term rainfall predictions. Fu Yaping, Li Fang, et al., in "Design of a Severe Convective Weather Warning System Based on Phased Array Doppler Radar Technology," proposed a severe convective weather warning method based on phased array Doppler radar technology. This method combines particle motion frequency shifts with specific coordinate system conditions to construct a corresponding matrix and transform the formula. After preliminary data processing, it is applied to the system, thereby improving the accuracy of warnings. In foreign countries, Mohapatra M, Sharma M et al. mentioned in "Cyclone warning services in India during recent years: A review" that by combining statistics, meteorology, satellite monitoring, radar technology, numerical weather prediction (NWP) methods and dynamic statistical methods, and using technologies such as multi-model ensemble (MME) and single model ensemble prediction system (EPS), India's cyclone warning capabilities have been significantly improved. Bechini R, Baldini L, Chandrasekar V et al. proposed using polarimetric radar to observe ice crystals in precipitation clouds in "Polarimetric radar observations in the ice region of precipitating clouds at C-band and X-band radar frequencies" and developed a quantitative precipitation estimation method for flash flood disaster warning. However, the adaptability of existing meteorological warnings to low-altitude flights is still insufficient: (1) Traditional meteorological warning methods are mainly aimed at warnings in the entire airspace, while the meteorological characteristics of low-altitude airspace show drastic changes. The applicability and effectiveness of traditional meteorological warning methods in low-altitude scenarios are both insufficient. In addition, low-altitude weather is unpredictable, and real-time warning is also a factor that needs to be considered. Currently, there are relatively few weather warning methods for low-altitude airspace, and weather warning methods for air taxis are even scarcer. (2) Traditional weather warnings usually rely on sensors to obtain weather information and issue warnings through analysis. However, this analysis process often relies on human operation, which is often subjective. Due to the complexity of low-altitude airspace, any slight error can lead to serious consequences, so the influence of human subjective judgment must be avoided as much as possible.

[0004] A new weather warning method is needed to solve the above technical problems. Summary of the Invention

[0005] The purpose of the present invention is to provide a low-altitude weather warning method and warning system based on binary logistic regression, which is used to solve the technical problems of low accuracy and high human participation in the existing low-altitude warning technology.

[0006] The present invention also aims to provide an early warning system that adopts a low-altitude weather early warning method based on binary logistic regression.

[0007] The technical solution of the present invention to solve the technical problem is:

[0008] A low-altitude weather warning method based on binary logistic regression includes the following steps:

[0009] S1: The processing system obtains meteorological data from the previous year, preprocesses the data, and obtains training data; obtains meteorological data from last year, preprocesses the data, and obtains test data;

[0010] S2: Establish a training model. Using the training set data, sequentially apply the likelihood function estimation algorithm and the Newton iteration method to obtain the regression coefficient. Then, determine whether the significance level of the independent variable and the constant term of the meteorological data meets the set standard. If so, obtain the preliminary weather warning model formula and execute step S3. If not, remove the constant term, retain only the independent variable, re-establish the training model, and execute step S2.

[0011] S3: The test system uses the test group data to test the weather forecast preliminary model formula to determine whether the prediction success rate meets the set standard. If so, the weather forecast model formula is output; if not, the process returns to step S2. Preferably, the process further includes:

[0012] S4: The processing system obtains real-time weather data, substitutes the real-time weather data into the weather forecast model formula, obtains the weather warning index, and issues a weather warning based on the weather warning index.

[0013] Preferably, the preprocessing in step S1 includes: screening data and improving data.

[0014] Preferably, the preprocessing in step S1 is specifically as follows: screening the data to obtain numerical values of visibility, wind speed, rainfall and temperature variables, and processing problematic data and missing data by replacing or filling in the average values of corresponding variables of adjacent dates.

[0015] Preferably, in step S1, the ratio of the number of groups of training group data to the number of groups of test group data is 8:2.

[0016] The step S2 is specifically as follows:

[0017] S2.1: Establish a training model. The specific formula is:

[0018]

[0019] z=β0+β1X1+β2X2+...+β n X n ;

[0020] Where P(Y=1|X) represents the probability that the dependent variable Y=1 given the independent variable X, where Y=1 means the weather conditions are good and takeoff is possible; Y=0 means the weather conditions are bad and takeoff is not possible; z represents a linear combination; e is a natural constant; β0 represents a constant term, β1, β2, ..., β n represents the regression coefficients of different independent variables X;

[0021] S2.2: Substitute the training set data into the likelihood function. The specific formula of the likelihood function is:

[0022]

[0023] Transformed into: Taking the logarithm of the likelihood function gives the log-likelihood function. The specific formula of the log-likelihood function is:

[0024]

[0025] at this time,

[0026]

[0027] Where n represents the number of training data sets; i represents the specific group of training data sets; L(β) represents the likelihood function; l(β) represents the log-likelihood function; Y i Represents the state value of this event, Y i Taking 0 means this event occurs, Y i Taking 1 means that this event does not occur; P i Represents the probability of the current training set data; that is, if Y i =1, then P(Y i ∣X i )=P i If Y i =0, then P(Y i ∣X i )=1-P i ; j represents the type of independent variable, that is, when j is 1, it represents temperature; when j is 2, it represents wind speed; when j is 3, it represents rainfall; when j is 4, it represents visibility; represents the gradient of the j-th independent variable; K represents the number of independent variables;

[0028] The curvature of the log-likelihood function is represented by the Hessian matrix, specifically:

[0029]

[0030] Expand to get: Among them, H(β) represents the curvature of the log-likelihood function; H xy Represents the element in the xth row and yth column of the matrix; X ix represents the value of the independent variable x in the i-th group of data used to train the model; X iy represents the value of the yth independent variable in the i-th group of data;

[0031] S2.3: Calculate the regression coefficient using the Newton iteration method, specifically:

[0032] β (t+1) =β (t) -H -1 ▽l(β (t) );

[0033] Among them, β (t) represents the parameter estimate of the current iteration; β (t+1) Indicates the new value of the parameter after updating in the current iteration step; ▽l(β (t) ) represents the gradient of the log-likelihood function, H -1 represents the inverse matrix of the Hessian matrix; t represents the number of iterations;

[0034] S2.4: Determine whether the significance level of the independent variables and constant terms of the meteorological data meets the set standards. If so, obtain the preliminary weather warning model formula and execute step S3. If not, remove the constant term, retain only the independent variables, re-establish the training model, and execute step S2.1. The preliminary weather warning model formula is specifically: or Among them, Y represents the weather warning index; X1 represents the temperature value, X2 represents the wind speed value, X3 represents the precipitation value, and X4 represents the visibility value.

[0035] Preferably, the weather warning index is specifically: Y∈[0.8,1): the regional weather conditions are good;

[0036] Y∈[0.5,0.8): The regional meteorological conditions are good; Y∈[0.2,0.5): The regional meteorological conditions are poor;

[0037] Y∈(0,0.2): Regional weather conditions are very bad.

[0038] Preferably, the setting standard in step S3 is: the prediction success rate is not less than 95%.

[0039] A low-altitude weather warning system using a low-altitude weather warning method based on binary logistic regression, comprising:

[0040] The processing system is used to obtain the meteorological data of the previous year, pre-process the data, and obtain the training set data; obtain the meteorological data of the previous year, pre-process the data, and obtain the test set data;

[0041] The training system is used to establish a training model. Using the training group data, it adopts the likelihood function estimation algorithm and the Newton iteration method in sequence to obtain the regression coefficient, removes the constant term, and obtains the preliminary weather forecast model formula; the testing system is used to use the test group data to test the preliminary weather forecast model formula to determine whether the prediction success rate meets the set standard.

[0042] Preferably, the processing system is also used to obtain real-time weather data, substitute the real-time weather data into the weather forecast model formula to obtain a weather warning index, and issue a weather warning based on the weather warning index.

[0043] The present invention has the following beneficial effects: by processing meteorological data from previous years to obtain training data and test data, a training model is established. Using the training data, a likelihood function estimation algorithm and Newton iteration method are sequentially employed to obtain regression coefficients. The constant term is then removed based on the significance levels of the independent variable and the constant term, resulting in a preliminary weather warning model formula. The preliminary weather warning model formula is then tested using the test data to determine whether the prediction success rate meets a set standard. If the standard is met, the weather forecast model formula is output. Because the initial training and test data are large in volume, and the likelihood function estimation algorithm and Newton iteration method are used to further improve prediction accuracy, the results obtained by the weather forecast model formula have a high accuracy rate. Finally, real-time data is substituted into the model to generate warnings based on the range of the weather warning index. This allows air taxis to obtain warnings of current or future weather conditions by outputting meteorological data to the weather forecast model formula obtained using this warning method before flight, providing an important reference for flight decision-making and significantly improving flight safety. In addition, since the early training group data and the test group data are large amounts of data, and the likelihood function estimation algorithm and Newton iteration method are used to further improve the prediction accuracy, the results obtained by the weather prediction model formula are highly accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 1 is a flow chart of a low-altitude weather warning method based on binary logistic regression according to the present invention;

[0045] Figure 2It is the specific logistic regression solution result of the low-altitude weather warning method based on binary logistic regression of the present invention;

[0046] Figure 3 It is the iterative calculation result of the specific warning model of the low-altitude weather warning method based on binary logistic regression of the present invention;

[0047] Figure 4 It is a statistical graph of the warning accuracy of the low-altitude weather warning method based on binary logistic regression of the present invention;

[0048] Figure 5 It is a ROC curve diagram of the low-altitude weather warning method based on binary logistic regression of the present invention;

[0049] Figure 6 It is a structural diagram of a low-altitude weather warning system that adopts a low-altitude weather warning method based on binary logistic regression according to the present invention. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present invention will be described clearly and comprehensively below with reference to the accompanying drawings in the embodiments of the present invention.

[0051] like Figure 1 As shown, the present invention discloses a low-altitude weather warning method based on binary logistic regression, comprising the following steps:

[0052] S1: The processing system obtains meteorological data from the previous year, preprocesses the data, and obtains training group data; obtains meteorological data from last year, preprocesses the data, and obtains test group data; the ratio of the number of training group data to the test group data is 8:2.

[0053] S2: Establish a training model. Using the training set data, sequentially apply the likelihood function estimation algorithm and the Newton iteration method to obtain the regression coefficient. Then, determine whether the significance level of the independent variable and the constant term of the meteorological data meets the set standard. If so, obtain the preliminary weather warning model formula and execute step S3. If not, remove the constant term, retain only the independent variable, re-establish the training model, and execute step S2.

[0054] S3: The test system uses the test group data to test the preliminary weather forecast model formula to determine whether the prediction success rate meets the set standard. If so, the weather forecast model formula is output; if not, the system returns to step S2; the set standard is: the prediction success rate is not less than 95%.

[0055] S4: The processing system obtains real-time weather data, substitutes the real-time weather data into the weather forecast model formula, obtains the weather warning index, and issues a weather warning based on the weather warning index.

[0056] The preprocessing in step S1 includes data screening and data improvement. Specifically, the preprocessing involves filtering the data to obtain values for temperature, wind speed, precipitation, and visibility variables. Problematic or missing data is then replaced or filled in by using the average values of the corresponding variables from adjacent dates.

[0057] Step S2 is specifically as follows:

[0058] S2.1: Establish a training model. The specific formula is:

[0059]

[0060] z=β0+β1X1+β2X2+...+β n X n ;

[0061] Where P(Y=1|X) represents the probability that the dependent variable Y=1 given the independent variable X, where Y=1 means the weather conditions are good and takeoff is possible; Y=0 means the weather conditions are bad and takeoff is not possible; z represents a linear combination; e is a natural constant; β0 represents a constant term, β1, β2, ..., β n represents the regression coefficients of different independent variables X;

[0062] S2.2: Substitute the training set data into the likelihood function. The specific formula of the likelihood function is:

[0063]

[0064] Transformed into: Taking the logarithm of the likelihood function gives the log-likelihood function. The specific formula of the log-likelihood function is:

[0065]

[0066] at this time,

[0067]

[0068] Where n represents the number of training data sets; i represents the specific group of training data sets; L(β) represents the likelihood function; l(β) represents the log-likelihood function; Y i Represents the state value of this event, Y i Taking 0 means this event occurs, Y i Taking 1 means that this event does not occur; P i Represents the probability of the current training set data; that is, if Y i =1, then P(Y i ∣X i )=P i If Yi =0, then P(Y i ∣X i )=1-P i ; j represents the type of independent variable, that is, when j is 1, it represents temperature; when j is 2, it represents wind speed; when j is 3, it represents rainfall; when j is 4, it represents visibility; represents the gradient of the j-th independent variable; K represents the number of independent variables;

[0069] The curvature of the log-likelihood function is represented by the Hessian matrix, specifically:

[0070]

[0071] Expand to get: Among them, H(β) represents the curvature of the log-likelihood function; H xy Represents the element in the xth row and yth column of the matrix; X ix represents the value of the independent variable x in the i-th group of data used to train the model; X iy represents the value of the yth independent variable in the i-th group of data;

[0072] S2.3: Calculate the regression coefficient using the Newton iteration method, specifically:

[0073] β (t+1) =β (t) -H -1 ▽l(β (t) );

[0074] Among them, β (t) represents the parameter estimate of the current iteration; β (t+1) Indicates the new value of the parameter after updating in the current iteration step; ▽l(β (t) ) represents the gradient of the log-likelihood function, H -1 represents the inverse matrix of the Hessian matrix; t represents the number of iterations;

[0075] S2.4: Determine whether the significance level of the independent variables and constant terms of the meteorological data meets the set standards. If so, obtain the preliminary weather warning model formula and execute step S3. If not, remove the constant term, retain only the independent variables, re-establish the training model, and execute step S2.1. The preliminary weather warning model formula is specifically: or Among them, Y represents the weather warning index; X1 represents the temperature value, X2 represents the wind speed value, X3 represents the precipitation value, and X4 represents the visibility value.

[0076] Removing the constant term in the step can make the independent variable more significant, as shown in Tables 2 and 3. When the logistic regression solution result is Table 2, after removing the constant term, the iterative calculation result of the prediction model will be as shown in Table 3. At this time, it can be clearly seen that after removing the constant term, the P value is smaller, which means that the result in the statistical test is statistically significant, so the constant term is removed.

[0077] The specific weather warning index is: Y∈[0.8,1): the regional weather conditions are good, that is, air taxi operations can be carried out; Y∈[0.5,0.8): the regional weather conditions are relatively good, that is, air taxi operations can be carried out, but it is necessary to pay attention to weather changes at any time; Y∈[0.2,0.5): the regional weather conditions are poor, that is, it is necessary to pay attention to weather changes, and air taxi operations may need to be reduced; Y∈(0,0.2): the regional weather conditions are very poor, and air taxi operations need to be stopped.

[0078] like Figure 4 As shown, using the above-mentioned early warning method, when the 365 groups of data from the previous year are used as the training group data and the 91 groups of data from the previous year are used as the test group data, after obtaining the weather forecast model formula, the test group data is used for testing. Among the 91 days of data, the number of days with successful predictions is 89 days, and the number of days with failed predictions is only 2 days, with a prediction success rate of 97.8%. It can be seen that the test group test results show that the model prediction accuracy is high.

[0079] like Figure 5 As shown in the figure, the ROC curve of the test group data was made using SPSS software, and the area under the curve (AUC) was calculated to be 0.996706. It can be seen that the constructed low-altitude weather warning model almost perfectly distinguished between positive and negative samples, indicating that the warning model is effective.

[0080] In actual application, for example, the meteorological data of Shenzhen Baoan Meteorological Observation Point in 2023 and 2024 were released. The 2024 annual meteorological data (a total of 365 groups) were selected as the test group data, and part of the 2023 meteorological data were selected as the training group data (a total of 91 groups). The ratio of the training group and the test group data volume was 8:2. Through the above warning method, the weather forecast model formula was finally obtained:

[0081]

[0082] In the subsequent weather forecast process, only the four independent variables of temperature, wind speed, precipitation, and visibility need to be passed to the weather forecast model formula, where X1 represents the temperature value, X2 represents the wind speed value, X3 represents the precipitation value, and X4 represents the visibility value. For example, if X1 = 24, X2 = 8, X3 = 1, and X4 = 28, then: The model calculates a weather warning index of Y = 0.9999985915227596. Based on the pre-set interval, the weather warning index is within the range [0.8, 1). Therefore, the warning model issues the following warning: "Regional weather conditions are favorable, allowing air taxi operations." This warning can be displayed on a computer, mobile phone, or watch. There are no restrictions here; users can freely choose the appropriate display based on their needs.

[0083] like Figure 6 As shown, a low-altitude weather warning system using a low-altitude weather warning method based on binary logistic regression includes:

[0084] The processing system is used to obtain meteorological data from the previous year, pre-process the data, and obtain training data; obtain meteorological data from last year, pre-process the data, and obtain test data; obtain real-time weather data, substitute the real-time weather data into the weather forecast model formula, obtain the weather warning index, and issue weather warnings based on the weather warning index;

[0085] The training system is used to establish a training model. Using the training group data, it adopts the likelihood function estimation algorithm and the Newton iteration method in sequence to obtain the regression coefficient, removes the constant term, and obtains the preliminary weather forecast model formula; the testing system is used to use the test group data to test the preliminary weather forecast model formula to determine whether the prediction success rate meets the set standard.

[0086] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

Claims

1. A low-altitude weather warning method based on binary logistic regression, characterized in that: The following steps are involved: S1: Obtain meteorological data from the previous year, preprocess the data, and obtain training set data; Obtain last year's meteorological data, preprocess the data, and obtain the test group data; S2: Establish a training model. Using the training set data, sequentially apply the likelihood function estimation algorithm and the Newton iteration method to obtain the regression coefficient. Then, determine whether the significance level of the independent variable and the constant term of the meteorological data meets the set standard. If so, obtain the preliminary weather warning model formula and execute step S3. If not, remove the constant term, retain only the independent variable, re-establish the training model, and execute step S2. S3: Use the test group data to test the weather forecast preliminary model formula to determine whether the prediction success rate meets the set standard. If so, output the weather forecast model formula; if not, return to step S2.

2. The low-altitude weather warning method based on binary logistic regression according to claim 1 is characterized in that: Also includes: S4: Obtain real-time weather data, substitute the real-time weather data into the weather forecast model formula to obtain the weather warning index, and issue a weather warning based on the weather warning index.

3. The low-altitude weather warning method based on binary logistic regression according to claim 1 is characterized in that ,The preprocessing in step S1 includes: screening data and improving data.

4. The low-altitude weather warning method based on binary logistic regression according to claim 3 is characterized in that The preprocessing in step S1 is specifically as follows: screening the data to obtain the values of temperature, wind speed, precipitation and visibility variables, and processing the problem data and missing data by replacing and filling them with the average values of the corresponding variables of adjacent dates.

5. The low-altitude weather warning method based on binary logistic regression according to claim 1 is characterized in that: In step S1, the ratio of the number of training group data to the number of test group data is 8:

2.

6. The low-altitude weather warning method based on binary logistic regression according to claim 2 is characterized in that: The step S2 is specifically as follows: S2.1: Establish a training model. The specific formula is: z=β0+β1X1+β2X2+...+β n X n ; Where P(Y=1|X) represents the probability that the dependent variable Y=1 given the independent variable X. Y=1 means that the weather conditions are good and low-altitude aircraft can take off; Y=0 means that the weather conditions are bad and low-altitude aircraft cannot take off. z represents a linear combination; e is a natural constant; β0 represents a constant term, β1, β2, ..., β n represents the regression coefficients of different independent variables X; S2.2: Substitute the training set data into the likelihood function. The specific formula of the likelihood function is: Transformed into: Taking the logarithm of the likelihood function gives the log-likelihood function. The specific formula of the log-likelihood function is: at this time, Where n represents the number of training data sets; i represents the specific group of training data sets; L(β) represents the likelihood function; l(β) represents the log-likelihood function; Y i Represents the state value of this event, Y i Taking 0 means this event occurs, Y i Taking 1 means that this event does not occur; P i Represents the probability of the current training set data; that is, if Y i =1, then P(Y i ∣X i )=P i If Y i =0, then P(Y i ∣X i )=1-P i ; j represents the type of independent variable, that is, when j is 1, it represents temperature; when j is 2, it represents wind speed; when j is 3, it represents rainfall; when j is 4, β4 represents visibility; represents the gradient of the jth independent variable; k represents the number of independent variables; the curvature of the log-likelihood function is represented by the Hessian matrix, specifically: Expand to get: Among them, the matrix H(β) represents the curvature of the log-likelihood function; H xy Represents the element in the xth row and yth column of the matrix; X ix represents the value of the independent variable x in the i-th group of data used to train the model; X iy represents the value of the yth independent variable in the i-th group of data; S2.3: Calculate the regression coefficient using the Newton iteration method, specifically: β (t) represents the parameter estimate of the current iteration; β (t+1) Indicates the new value of the parameter after updating in the current iteration step; represents the gradient of the log-likelihood function, H -1 represents the inverse matrix of the Hessian matrix; t represents the number of iterations; S2.4: Determine whether the significance level of the independent variables and constant terms of the meteorological data meets the set standards. If so, obtain the preliminary weather warning model formula and execute step S3. If not, remove the constant term, retain only the independent variables, re-establish the training model, and execute step S2.

1. The preliminary weather warning model formula is specifically: or Among them, Y represents the weather warning index; X1 represents the temperature value, X2 represents the wind speed value, X3 represents the precipitation value, and X4 represents the visibility value.

7. The low-altitude weather warning method based on binary logistic regression according to claim 6 is characterized in that: The specific weather warning index is: Y∈[0.8,1): the regional weather conditions are good; Y∈[0.5,0.8): the regional weather conditions are relatively good; Y∈[0.2,0.5): The regional meteorological conditions are poor; Y∈(0,0.2): The regional meteorological conditions are very bad.

8. The low-altitude weather warning method based on binary logistic regression according to claim 1 is characterized in that: The setting standard in step S3 is: the prediction success rate is not less than 95%.

9. A low-altitude weather warning system using the low-altitude weather warning method based on binary logistic regression according to any one of claims 1 to 8, characterized in that: include: The processing system is used to obtain meteorological data from the previous year, pre-process the data, and obtain training set data; Obtain last year's meteorological data, preprocess the data, and obtain the test group data; The training system is used to establish a training model. Using the training set data, the likelihood function estimation algorithm and the Newton iteration method are used to obtain the regression coefficient, and then the constant term is removed to obtain the preliminary weather forecast model formula. The test system is used to test the preliminary weather forecast model formula using the test group data to determine whether the prediction success rate meets the set standards.

10. The low-altitude weather warning system using the low-altitude weather warning method based on binary logistic regression according to claim 9, characterized in that: The processing system is also used to obtain real-time weather data, substitute the real-time weather data into the weather forecast model formula to obtain a weather warning index, and issue a weather warning based on the weather warning index.

Images