Geographical information system-based model parameter optimization method

Abstract

The application discloses a model parameter optimization method based on a geographic information system, and relates to the technical field of geographic information systems.The application solves the problem that the optimization direction of a single operation item is reasonable, but the directions conflict when multiple operation items are cooperatively optimized.The application takes error proportion as a core evaluation index, and minimizes error proportion as an objective from the screening of a standard training set, the optimization of a single parameter, to the cooperative debugging of multiple operation items, finally locking the optimization parameter with the minimum error proportion.Through multiple iterations and accurate debugging, the application effectively reduces the deviation between the model output value and the actual observation value, and significantly improves the operation accuracy and stability of GIS related models, compared with the problem that error control is fuzzy and the optimization effect is unstable in the traditional optimization method.

Description

Technical Field

[0001] This invention relates to the field of geographic information system technology, specifically to a method for optimizing model parameters based on geographic information systems. Background Technology

[0002] In practical applications of Geographic Information Systems (GIS), the performance of various system models (such as hydrological simulation models and spatial interpolation models) is closely related to the model parameter settings. Parameter optimization is a core element in improving the accuracy of model simulation. With the deepening application of GIS technology in various fields, the requirements for the scientific rigor and accuracy of model parameter optimization are constantly increasing. Among these, the rational determination of the optimization direction has become a key factor affecting the efficiency and effectiveness of parameter optimization, and it is also a technical pain point that urgently needs to be addressed in the current GIS model parameter optimization process.

[0003] Existing GIS model parameter optimization methods lack scientific and systematic support for determining the optimization direction, exhibiting significant irrationality and arbitrariness. Traditional optimization methods often rely on the professional experience of operators, randomly selecting debugging directions without considering model operation patterns, parameter characteristics, and error change trends to accurately judge and lock onto the optimization direction. This leads to frequent directional deviations during parameter debugging, significantly increasing computational resource consumption, prolonging the optimization cycle, and making it difficult to quickly approach the optimal parameter range with minimal error, severely impacting the efficiency and accuracy of parameter optimization.

[0004] Specifically, in scenarios involving the collaborative optimization of multiple operational parameters, existing methods lack an effective mechanism for confirming the collaborative optimization direction. Without proper parameter combination debugging, the optimal optimization direction for each operational item is not clearly defined. This often results in situations where the optimization direction for a single operational item is reasonable, but conflicts arise during the collaborative optimization of multiple operational items. This prevents the achievement of synergistic optimization between individual parameters and the overall model, further reducing the model's optimization effectiveness. Furthermore, existing methods lack a standardized confirmation process for optimization directions. The optimization directions determined by different operators lack uniformity, which not only raises the practical threshold but also leads to poor consistency and repeatability of optimization results, making it unsuitable for the personalized optimization needs of different GIS models and different operational items (such as CN, confluence threshold, IDW idempotency, etc.).

[0005] Furthermore, existing optimization methods fail to deeply integrate the determination of the optimization direction with core indicators such as error ratio and compliant training set. This prevents real-time adjustments to the optimization direction based on error changes, leading to local optima and difficulty in locking onto truly optimal parameters. This blindness and irrationality in determining the optimization direction has become a significant bottleneck restricting the improvement of GIS model parameter optimization efficiency, failing to meet the demands of high-precision and high-efficiency model applications.

[0006] Therefore, in addressing the problems of blindly determining the optimization direction, lack of scientific support, absence of standardized processes, and difficulty in achieving collaborative adaptation of multiple operational items in the optimization of existing GIS model parameters, developing a model parameter optimization method based on geographic information systems that can accurately and scientifically determine the optimization direction has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a model parameter optimization method based on geographic information systems, which solves the problem that while the optimization direction of a single operational item is reasonable, there are often conflicting directions when multiple operational items are optimized collaboratively.

[0008] To achieve the above objectives, the present invention provides a model parameter optimization method based on a geographic information system, comprising the following steps: Step 1: From the historical data associated with the system model, confirm the error ratio between the model output value and the actual observed value, and determine the compliance ratio. Then, based on the compliance ratio, determine the compliance training set associated with different operational items. The specific method is as follows: The model output value SC associated with a single set of historical data. i and the actual observed value GZ i Where i represents different groups of data, using: DB i =|SC i -GZ i |÷GZ i Confirm the error ratio DB associated with the corresponding single set of data. i ; If DB i If the error rate is ≤2%, the error rate will be marked as the compliance rate; otherwise, no marking will be made. The single set of data associated with the compliance rate is recorded as compliance data. From the recorded sets of compliance data, the operation parameters belonging to the same operation item are extracted, and the minimum and maximum values ​​are confirmed to generate the compliance operation range associated with the corresponding operation parameters. The compliance operation range associated with each set of different operation parameters is confirmed in turn to generate a compliance training set associated with multiple operation items. After the target training set is confirmed, the minimum value is selected from the confirmed target ratios, and the operating parameters of different operating items associated with the minimum value are recorded as the overall optimal parameters. Step 2: Based on the qualified training sets confirmed for different operational items, execute a single training process for the model, keeping a certain operational item varying according to the qualified operational range within the qualified training set, and recording the error ratio in real time. From the large amount of training data associated with the single training process, find the optimal parameters associated with the corresponding operational item. The specific method is as follows: For each running item existing in the system model, determine the total number of running items and execute the corresponding number of single training processes; In a single training process, a set of running items is selected and recorded as training items. The target running range associated with the training items is confirmed. The overall optimal parameters associated with other running items are extracted and run simultaneously, so that the training items gradually change from the minimum value of the target running range to the maximum value. Each change of a unit value completes a set of model output processes, and the error ratio between the model output value and the actual observation value in the model output process is recorded. The unit value is a preset unit. For the training change value and error ratio associated with the corresponding training item in a single training process, a set of two-dimensional coordinate system is generated with the training change value as the X-axis and the error ratio as the Y-axis. The two-dimensional points associated with each stage of the unit value change are identified, and the several sets of two-dimensional points associated in a single training process are connected to generate the standard curve of the single training process. Selecting a high-density segment from the standard curve: From a large number of error proportions in a single training process, select the maximum error proportion and construct a set of standard upper lines parallel to the X-axis and perpendicular to the Y-axis. The value corresponding to the standard upper line in this current iteration is the maximum error proportion. Simultaneously, confirm the minimum error proportion and construct a standard lower line parallel to the X-axis and perpendicular to the Y-axis. The initial value corresponding to the standard lower line is the minimum error proportion. Gradually move the standard lower line upwards, and record the overall line length L of the curve segments associated with the standard upper and lower lines during each movement process. k Where k represents different movement processes, and then confirm the error ratio range F between the standard upper limit and the standard lower limit. k Using: F k ÷L k =M k Lock density value M k Extract several density values ​​M from several movement processes. k The minimum value is recorded, and the moving process associated with the minimum value is recorded and marked as a candidate process. Adjust the standard upper and lower limits by a set of error ratio units, and use the method of moving the standard lower limit to the standard upper limit to lock the candidate process associated with this standard upper limit; The upper limit of the standard is gradually lowered until the difference between the upper and lower limits of the standard is only one set of error ratio units, and the candidate processes associated with the upper limit of the standard at each different position are confirmed. From the confirmed candidate processes, the minimum value of the associated density value of the candidate processes is identified, and the candidate processes associated with the minimum value are recorded as selected processes. The curve segments associated with the selected processes are recorded as selected curve segments. The training change values ​​associated with the selected curve segments are averaged, and the resulting average is used as the individual optimal parameter of the corresponding running item. Step 3: Based on the individual optimal parameters and overall optimal parameters associated with the corresponding operation item, confirm the selected range associated with the corresponding operation item. Then, confirm each pair of selected ranges to lock the direction associated with each different operation item. The specific method is as follows: Based on the individual optimal parameters and the overall optimal parameters associated with each individual operation item, the selected range associated with the corresponding operation item is determined; Then randomly select two sets of running items A and B, execute the model output process according to the determined selected range, ensure that other running items run according to the determined overall optimal parameters, and randomly select a set of debugging directions, record whether the error ratio associated with the model output process decreases. If it decreases, record this debugging direction as the determined direction. Otherwise, continue to select debugging directions until the determined direction is locked. Then randomly select two sets of running items B and C. Keep the locked direction of running item B unchanged, confirm the direction of C, and run other running items according to the determined overall optimal parameters until the error ratio becomes smaller, then lock the determined direction of C. Similarly, select a group of running items from the previous group of determined processes, and then select other running items that have not been determined to combine them, thereby locking the determination direction associated with other subsequent running items in turn; Step 4: Based on the determined direction and selected range associated with different running items, perform proportional quantization on the selected range, and execute the model output process according to the determined direction until the error ratio is at its minimum. Then, lock the optimization parameters associated with each running item and execute the process. Preferably, in step four, the specific method for performing proportional quantization on the selected range is as follows: Based on the set quantization value H, where H is a preset value, the minimum value of the selected range associated with the corresponding running item is quantized to 1, and the maximum value of the selected range is quantized to H; Identify the range values ​​associated with different quantization values ​​for 1-H within each selected range.

[0009] Preferably, in step four, the specific method for locking the optimization parameters is as follows: Based on the defined directions associated with different operating items, initial values ​​are selected from different selected ranges. Based on the set quantization values, the range values ​​associated with the corresponding operating items are gradually adjusted. Each subsequent adjustment process differs from the previous process by a set of quantization values. Based on the gradual adjustment process, the model output process is executed. In the executed model output process, the error ratio associated with different adjustment stages is determined. From the determined error ratios, the minimum value is selected. The adjustment stage associated with the minimum value is recorded as the optimization stage. The range values ​​of different operating items within the optimization stage are recorded as optimization parameters and executed.

[0010] This invention provides a method for optimizing model parameters based on geographic information systems. Compared with existing technologies, it has the following advantages: By calculating the error ratio between the model output value and the actual observed value using historical data, the qualified training set is selected based on a clear qualified ratio. At the same time, by combining the characteristics of different operation items, the qualified operation range of each operation item is locked, eliminating invalid data interference from the source, and ensuring that the subsequent optimization process is based on valid and compliant training data. By using a single training process, fixing the parameters of other operational items as the overall optimal parameters, and allowing only the target operational item to gradually change within the acceptable range, the error ratio is recorded synchronously and a standard curve is generated. Then, the optimal curve segment is locked by density value calculation, and the average value is taken to determine the optimal parameters for each individual item. This process abandons the subjectivity and randomness of traditional trial-and-error methods. Through visualization with a two-dimensional coordinate system and quantitative analysis of density values, it accurately captures the parameter range with the lowest error ratio. It takes into account both the geophysical significance of the operational parameters and achieves refined optimization of individual parameters, solving the problems of difficult accurate positioning of individual parameters and large error fluctuations in traditional methods, and ensuring that the parameters of each operational item are in their own optimal operating state. By combining operational items in pairs and gradually locking in the direction of each operational item, this approach breaks through the limitations of traditional optimization methods that focus on independent optimization of a single parameter and ignore the coupling effects of multiple parameters. By first confirming the selected range of the optimal parameters for each individual item and the overall optimal parameters, and then using a combination and debugging approach to gradually expand the scope, the optimization direction of each operational item is locked in. This ensures that when multiple operational item parameters are optimized collaboratively, the optimal characteristics of each individual parameter are preserved while also taking into account the overall model's operational efficiency. Using the error ratio as the core evaluation metric, this method aims to minimize the error ratio throughout the entire process, from selecting the qualified training set and optimizing individual parameters to coordinating the debugging of multiple operational items. Through multiple rounds of iteration and precise debugging, the optimal parameters with the minimum error ratio are ultimately identified. Compared to the problems of fuzzy error control and unstable optimization results in traditional optimization methods, this method effectively reduces the deviation between the model output value and the actual observation value through a progressive error control strategy, significantly improving the operational accuracy and stability of GIS-related models. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] First Embodiment

[0014] Please see Figure 1 This application provides a method for optimizing model parameters based on geographic information systems, including the following steps: Step 1: From the historical data associated with the system model, confirm the error ratio between the model output value and the actual observed value, and lock in the compliance ratio. Then, based on the compliance ratio, confirm the compliance training set associated with different operating items. Specifically, during the training process, the system model itself involves a variety of different operating parameters, such as CN, confluence threshold, and IDW idempotency. The operating parameters associated with different operating items will result in different compliance training sets. In order to facilitate the subsequent optimization process of model parameters, it is necessary to effectively and quickly lock in the different operating parameters associated with the compliance state based on the different characteristics involved in different operating items, and confirm the interval, thereby locking in the specific training set in the compliance state. Step 2: Based on the qualified training sets confirmed for different running items, execute a single training process for the model, keeping a certain running item changing according to the qualified running range within the qualified training set, and recording the error ratio in real time. From the large amount of training data associated with the single training process, find the single optimal parameter associated with the corresponding running item. Specifically, different running items are associated with different qualified running ranges. In order to select the single optimal parameter from multiple different running items, it is necessary to change the value from the corresponding qualified running range, starting from the minimum value and moving towards the maximum value. During the change process, record the numerical changes in the corresponding model training process. When performing single training, a single running item changes according to the qualified running range, while other running items run according to the determined overall optimal parameter and remain unchanged. Step 3: Based on the individual optimal parameters and the overall optimal parameters associated with the corresponding operation item, confirm the selected range associated with the corresponding operation item, and then confirm each pair of selected ranges to lock the direction associated with each different operation item. Step 4: Based on the determined direction and selected range associated with different running items, perform proportional quantization on the selected range, and execute the model output process according to the determined direction until the error ratio is at its minimum. Then, lock the optimization parameters associated with each running item and execute the process. The specific method for performing geometric quantization on the selected range is as follows: Based on the set quantization value H, where H is a preset value, generally 10, and its specific value is determined in advance by the operator based on experience, the minimum value of the selected range associated with the corresponding running item is quantized to 1, and the maximum value of the selected range is quantized to H. Confirm the range values ​​associated with different quantization values ​​from 1 to H within each selected range; The specific method for locking the optimization parameters is as follows: Based on the defined direction associated with different running items, initial values ​​are selected from different selected ranges. Based on the set quantization values, the range values ​​associated with the corresponding running items are gradually adjusted. Each subsequent adjustment process differs from the previous process by a set of quantization values. Based on the gradual adjustment process, the model output process is executed. In the executed model output process, the error ratio associated with different adjustment stages is determined. From the determined error ratios, the minimum value is selected. The adjustment stage associated with the minimum value is recorded as the optimization stage. The range values ​​of different running items within the optimization stage are recorded as optimization parameters and executed. Specifically, when making numerical changes, the changes are made according to the determined direction. The initial stage is based on the initial range value associated with the determined direction. In each subsequent stage, a set of quantized values ​​is changed, which corresponds to the associated range value. Then, the corresponding model output process is executed to lock the optimal parameters from the specific directions that are confirmed in turn. Here, the individual running parameters of a single running item are the optimal parameters for the corresponding running item itself, while the overall running parameters are the optimal parameters for a single running item under the overall running state. If the two optimal parameters are the same, they can be directly recorded as the optimization parameters. If they are different, it is necessary to select the optimal parameter from the range of different optimal parameters to ensure that the system model is in the best running state and achieves the best optimization effect.

[0015] Second Embodiment

[0016] In this embodiment, the main focus is on the confirmation process of the qualified training set: The specific method for confirming the qualified training set is as follows: The model output value SC associated with a single set of historical data. i and the actual observed value GZ i Where i represents different groups of data, using: DB i =|SCi -GZ i |÷GZ i Confirm the error ratio DB associated with the corresponding single set of data. i ; If DB i If the error rate is ≤2%, the error rate will be marked as the compliance rate; otherwise, no marking will be made. The single set of data associated with the compliance rate is recorded as compliance data. From the recorded sets of compliance data, the operation parameters belonging to the same operation item are extracted, and the minimum and maximum values ​​are confirmed to generate the compliance operation range associated with the corresponding operation parameters. The compliance operation range associated with each set of different operation parameters is confirmed in turn to generate a compliance training set associated with multiple operation items. After confirming the target training set, select the minimum value from the confirmed target ratios, and record the running parameters of different running items associated with the minimum value as the overall optimal parameters. (The overall optimal parameters and the individual optimal parameters are different. Further steps are needed to determine the optimal parameters of the model.) Specifically, in historical data, there are corresponding model output values ​​and actual observation values ​​for the corresponding system model. Based on the numerical difference between the output value and the observation value, the error ratio between the output value and the observation value can be effectively confirmed. Thus, the standard operating range of the corresponding operating parameters can be confirmed from the historical data. Within the corresponding numerical range, the system model can be effectively in normal operating condition, which facilitates subsequent optimization of the model parameters and finding the optimal value within the determined numerical range.

[0017] Third Embodiment

[0018] In this embodiment, the specific implementation process mainly focuses on the confirmation process of the optimal parameters for a single unit: The specific method for confirming the optimal parameters of a single set of operating items is as follows: For each running item existing in the system model, determine the total number of running items and execute the corresponding number of single training processes; In a single training process, a set of running items is selected and recorded as training items. The target running range associated with the training items is confirmed. The overall optimal parameters associated with other running items are extracted and run simultaneously (that is, other running items run according to the determined overall optimal parameters). The training items gradually change from the minimum value of the target running range to the maximum value. Each change of a unit value completes a set of model output processes. The error ratio associated with the model output value and the actual observation value in the model output process is recorded. The unit value is a preset unit, which is determined in advance by the operator based on experience. The unit values ​​associated with different running items may be the same or different. For the training change value (i.e. the range of the standard operating range) and error ratio associated with the corresponding training item in a single training process, a set of two-dimensional coordinate system is generated with the training change value as the X-axis and the error ratio as the Y-axis. The two-dimensional points associated with each stage of the unit value change are identified, and the several sets of two-dimensional points associated in a single training process are connected to generate the standard curve of the single training process. Selecting a high-density segment from the standard curve: From a large number of error proportions in a single training process, select the maximum error proportion and construct a set of standard upper lines parallel to the X-axis and perpendicular to the Y-axis. The value corresponding to the standard upper line in this current iteration is the maximum error proportion. Simultaneously, confirm the minimum error proportion and construct a standard lower line parallel to the X-axis and perpendicular to the Y-axis. The initial value corresponding to the standard lower line is the minimum error proportion. Gradually move the standard lower line upwards, and record the overall line length L of the curve segments associated with the standard upper and lower lines during each movement process. k Where k represents different movement processes, and then confirm the error ratio range F between the standard upper limit and the standard lower limit. k Using: F k ÷L k =M k Lock density value M k (The smaller the range and the longer the line length, the lower the resulting density will be.) Extract several density values ​​M from several movement processes. k The minimum value is recorded, and the moving process associated with the minimum value is recorded and marked as a candidate process. Then, adjust the standard upper and lower limits by a set of error ratio units, and use the method of moving from the standard lower limit to the standard upper limit to lock the candidate process associated with this standard upper limit. The upper limit of the standard is gradually lowered until the difference between the upper and lower limits of the standard is only one set of error ratio units, and the candidate processes associated with the upper limit of the standard at each different position are confirmed. From the confirmed candidate processes, the minimum value of the density value associated with the candidate processes is determined (that is, the minimum value is selected again from the minimum values). The candidate processes associated with the minimum value are recorded as the selected processes, and the curve segments associated with the selected processes are recorded as the selected curve segments. The training change values ​​associated with the selected curve segments are averaged, and the resulting average is used as the individual optimal parameter of the corresponding running item. Specifically, during the training process, for a single running item, the optimal parameters associated with other running items are first identified and put into operation. Then, a single test is performed on the single running item, and the corresponding training process is executed. Based on the determined standard operating range, the corresponding range of values ​​is identified. Then, based on the different error ratios associated with different range values, corresponding error ratio change curves are generated. Subsequently, based on the maximum and minimum values ​​associated with the corresponding curves, the upper and lower density confirmation method is used to extract and confirm the part of the curve between the two upper and lower lines. From this, the specific part of the curve segment with the highest density is locked, and the standard state is confirmed. Thus, the single optimal parameter associated with the corresponding running item is selected.

[0019] Fourth embodiment

[0020] In this embodiment, compared to the above embodiments, the main focus is on the specific confirmation process of determining the direction: Lock the specific direction associated with each different run item: Based on the individual optimal parameters and the overall optimal parameters associated with each individual operation item, the selected range associated with the corresponding operation item is determined; Then randomly select two sets of running items A and B, execute the model output process according to the determined selected range, ensure that other running items run according to the determined overall optimal parameters, and randomly select a set of debugging directions, record whether the error ratio associated with the model output process decreases. If it decreases, record this debugging direction as the determined direction. Otherwise, continue to select debugging directions until the determined direction is locked. Then randomly select two sets of running items B and C. Keep the locked direction of running item B unchanged, confirm the direction of C, and run other running items according to the determined overall optimal parameters (including A) until the error ratio becomes smaller, and lock the determined direction of C. Similarly, select a group of running items from the previous defined process, then select other running items that have not yet been defined and combine them, and then lock the defined direction associated with other running items in turn.

[0021] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0022] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for optimizing model parameters based on geographic information systems, characterized in that, Includes the following steps: Step 1: From the historical data associated with the system model, confirm the error ratio between the model output value and the actual observed value, and determine the compliance ratio. Then, based on the compliance ratio, determine the compliance training set associated with different operational items. The specific method is as follows: The model output value SC associated with a single set of historical data. i and the actual observed value GZ i Where i represents different groups of data, using: DB i =|SC i -GZ i |÷GZ i Confirm the error ratio DB associated with the corresponding single set of data. i ; If DB i If the error rate is ≤2%, the error rate will be marked as the compliance rate; otherwise, no marking will be made. The single set of data associated with the compliance rate is recorded as compliance data. From the recorded sets of compliance data, the operation parameters belonging to the same operation item are extracted, and the minimum and maximum values ​​are confirmed to generate the compliance operation range associated with the corresponding operation parameters. The compliance operation range associated with each set of different operation parameters is confirmed in turn to generate a compliance training set associated with multiple operation items. After the target training set is confirmed, the minimum value is selected from the confirmed target ratios, and the operating parameters of different operating items associated with the minimum value are recorded as the overall optimal parameters. Step 2: Based on the qualified training sets confirmed for different operational items, execute a single training process for the model, keeping a certain operational item varying according to the qualified operational range within the qualified training set, and recording the error ratio in real time. From the large amount of training data associated with the single training process, find the optimal parameters associated with the corresponding operational item. The specific method is as follows: For each running item existing in the system model, determine the total number of running items and execute the corresponding number of single training processes; In a single training process, a set of running items is selected and recorded as training items. The target running range associated with the training items is confirmed. The overall optimal parameters associated with other running items are extracted and run simultaneously, so that the training items gradually change from the minimum value of the target running range to the maximum value. Each change of a unit value completes a set of model output processes, and the error ratio between the model output value and the actual observation value in the model output process is recorded. The unit value is a preset unit. For the training change value and error ratio associated with the corresponding training item in a single training process, a set of two-dimensional coordinate system is generated with the training change value as the X-axis and the error ratio as the Y-axis. The two-dimensional points associated with each stage of the unit value change are identified, and the several sets of two-dimensional points associated in a single training process are connected to generate the standard curve of the single training process. Selecting a high-density segment from the standard curve: From a large number of error proportions in a single training process, select the maximum error proportion and construct a set of standard upper lines parallel to the X-axis and perpendicular to the Y-axis. The value corresponding to the standard upper line in this current iteration is the maximum error proportion. Simultaneously, confirm the minimum error proportion and construct a standard lower line parallel to the X-axis and perpendicular to the Y-axis. The initial value corresponding to the standard lower line is the minimum error proportion. Gradually move the standard lower line upwards, and record the overall line length L of the curve segments associated with the standard upper and lower lines during each movement process. k Where k represents different movement processes, and then confirm the error ratio range F between the standard upper limit and the standard lower limit. k Using: F k ÷L k =M k Lock density value M k Extract several density values ​​M from several movement processes. k The minimum value is recorded, and the moving process associated with the minimum value is recorded and marked as a candidate process. Step 3: Based on the individual optimal parameters and the overall optimal parameters associated with the corresponding operation item, confirm the selected range associated with the corresponding operation item, and then confirm each pair of selected ranges to lock the direction associated with each different operation item. Step 4: Based on the defined direction and selected range associated with different running items, perform proportional quantization on the selected range, and execute the model output process according to the defined direction until the error ratio is at its minimum. Then, lock the optimization parameters associated with each running item and execute the process.

2. The model parameter optimization method based on geographic information systems according to claim 1, characterized in that, In step two, the specific method for confirming the optimal parameters of a single set of running items also includes: Adjust the standard upper and lower limits by a set of error ratio units, and use the method of moving the standard lower limit to the standard upper limit to lock the candidate process associated with this standard upper limit; The upper limit of the standard is gradually lowered until the difference between the upper and lower limits of the standard is only one set of error ratio units, and the candidate processes associated with the upper limit of the standard at each different position are confirmed. From the confirmed candidate processes, identify the minimum value of the associated density value of the candidate processes associated with the minimum value, and record the candidate processes associated with the minimum value as selected processes. Record the curve segments associated with the selected processes as selected curve segments. Perform mean processing on the training change values ​​associated with the selected curve segments, and use the obtained mean as the individual optimal parameter of the corresponding running item.

3. The model parameter optimization method based on geographic information systems according to claim 1, characterized in that, In step three, the specific method for determining the direction of each different running item is as follows: Based on the individual optimal parameters and the overall optimal parameters associated with each individual operation item, the selected range associated with the corresponding operation item is determined; Then randomly select two sets of running items A and B, execute the model output process according to the determined selected range, ensure that other running items run according to the determined overall optimal parameters, and randomly select a set of debugging directions, record whether the error ratio associated with the model output process decreases. If it decreases, record this debugging direction as the determined direction. Otherwise, continue to select debugging directions until the determined direction is locked. Then randomly select two sets of running items B and C. Keep the locked direction of running item B unchanged, confirm the direction of C, and run other running items according to the determined overall optimal parameters until the error ratio becomes smaller, then lock the determined direction of C. Similarly, select a group of running items from the previous defined process, then select other running items that have not yet been defined and combine them, and then lock the defined direction associated with other running items in turn.

4. The model parameter optimization method based on geographic information systems according to claim 1, characterized in that, In step four, the specific method for performing proportional quantization on the selected range is as follows: Based on the set quantization value H, where H is a preset value, the minimum value of the selected range associated with the corresponding running item is quantized to 1, and the maximum value of the selected range is quantized to H; Identify the range values ​​associated with different quantization values ​​for 1-H within each selected range.

5. The model parameter optimization method based on geographic information systems according to claim 4, characterized in that, In step four, the specific method for locking the optimization parameters is as follows: Based on the defined directions associated with different operating items, initial values ​​are selected from different selected ranges. Based on the set quantization values, the range values ​​associated with the corresponding operating items are gradually adjusted. Each subsequent adjustment process differs from the previous process by a set of quantization values. Based on the gradual adjustment process, the model output process is executed. In the executed model output process, the error ratio associated with different adjustment stages is determined. From the determined error ratios, the minimum value is selected. The adjustment stage associated with the minimum value is recorded as the optimization stage. The range values ​​of different operating items within the optimization stage are recorded as optimization parameters and executed.

Images