Island tourism sea development activity suitability evaluation method based on meta analysis and integrated learning

Abstract

The invention relates to the technical field of ocean space resource and ecological environment analysis and evaluation, in particular to an island tourism sea development activity suitability evaluation method based on meta analysis and ensemble learning, which comprises the following steps of: extracting evaluation indexes by adopting a generative large language model, and constructing a hierarchical evaluation system; the hierarchical evaluation system comprises a first-level index and a second-level index, and the first-level index is divided into a non-restrictive index and a restrictive index; according to activity requirements, a research area is delimited in combination with sea use characteristics; based on a hierarchical evaluation system, performing index quantification and hierarchical score assignment on the collected data; integrating the processed data, and constructing an index data set; and evaluating the index data set based on meta analysis and ensemble learning to realize suitability evaluation of island tourism sea development activities. The index system and the machine learning framework in the invention are suitable for the suitability evaluation of island tourism sea use development activities in different areas, and support the optimal configuration and sustainable development decision of ocean tourism resources.

Description

Technical Field

[0001] This invention relates to the field of environmental analysis technology, and in particular to a method for evaluating the suitability of marine development activities for island tourism based on meta-analysis and ensemble learning. Background Technology

[0002] Island tourism is an important growth engine for the global marine economy, but indiscriminate development can cause irreversible damage to its ecosystems. Therefore, conducting suitability assessments for marine use for tourism and recreation is crucial for the scientific planning and sustainable development of marine resources.

[0003] Existing suitability assessment methods mainly follow a multi-criteria decision analysis framework, covering three key stages: indicator selection, weight determination, and comprehensive evaluation. In indicator selection, subjective methods such as expert experience or the Delphi method are often relied upon. For example, some studies use seawater transparency and coral coverage to assess diving suitability, or select coastline type and access distance to evaluate beach tourism. The comprehensiveness and universality of the indicator system are often limited by the experts' knowledge structure. When determining weights, methods such as the analytic hierarchy process (AHP), entropy method, or principal component analysis are commonly used. Although some mathematical modeling is achieved, the underlying data and judgment matrix still cannot completely avoid the subjectivity of human intervention. In the final comprehensive evaluation stage, thresholds are often set through expert experience or natural segmentation is used for grading, which also suffers from the drawback of subjective judgment.

[0004] In summary, existing evaluation systems, in practical applications, exhibit strong subjectivity throughout the entire process from initial indicator selection to final grading, easily leading to insufficient objectivity in evaluation results. In recent years, the successful application of meta-analysis in fields such as medicine and climate has demonstrated its potential to reduce subjective bias by systematically integrating existing research findings; simultaneously, machine learning methods have also shown powerful objective modeling capabilities in suitability assessments. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a suitability evaluation method for island tourism and marine development activities based on meta-analysis and ensemble learning. This invention employs meta-analysis to derive a suitability evaluation index system for island activities from existing relevant literature, and then constructs a standardized dataset based on each index. In the evaluation stage, the study area is divided into restricted and unrestricted areas according to limiting indicators. An ensemble learning method is then used to further classify the unrestricted areas based on their suitability, and the spatial analysis results of the unrestricted areas are combined to form the final suitability evaluation result for island activities.

[0006] The technical means employed in this invention are as follows: A suitability evaluation method for island tourism marine development activities based on meta-analysis and ensemble learning includes: extracting evaluation indicators using a generative large language model to construct a hierarchical evaluation system; the hierarchical evaluation system includes primary and secondary indicators, with the primary indicators further divided into unrestricted and restricted indicators; delineating the research area based on activity needs and marine use characteristics; quantifying and assigning scores to the collected data based on the hierarchical evaluation system; integrating the processed data to construct an indicator dataset; and evaluating the indicator dataset based on meta-analysis and ensemble learning to achieve a suitability evaluation of island activities.

[0007] Furthermore, the non-restrictive indicators include: natural conditions, marine environment, tourism resources, infrastructure, and development plans; the restrictive indicators are limiting conditions. The secondary indicators are refinements of the primary indicators. The natural conditions include: slope, elevation, and vegetation cover; the marine environment includes seawater quality, seawater temperature, and seawater depth; the tourism resources include the accessibility of reef landscapes, scenic spots, bathing beaches, and nearshore recreation; the infrastructure includes population density, catering facilities, accommodation facilities, medical facilities, recreational clusters, public safety, public restrooms, and transportation conditions; the development plan includes coastline type, marine functional zoning, and sea area use rights; and the restrictive conditions include marine ecological red line areas, marine navigation channel areas, marine engineering dumping areas, marine anchorage areas, and marine port administrative areas.

[0008] Furthermore, the study area is delineated with a 3-kilometer buffer zone on the seaward side, based on the remotely sensed island coastline; other land areas within the buffer zone, including mainland and small islands, are excluded; and grid cells are constructed within the study area.

[0009] Furthermore, the process of quantifying and grading the collected data includes: The index quantification includes classifying the collected data types into land raster data, land vector data, sea surface raster data, and sea surface vector data; and using nearest neighbor analysis to assign values ​​to the grid cells of the study area in combination with the collected data. The tiered scoring system assigns scores to both non-restrictive and restrictive indicators in the indicator system using a 0, 1, 3, and 5 scale. The non-restrictive indicators reflect the differentiation of suitability, using a 1, 3, and 5 scoring scale, where 1 represents unsuitable, 3 represents generally suitable, and 5 represents suitable. The restrictive indicators reflect the mandatory constraints of objective limitations on the suitability of activities, using a 0 and 1 scoring scale, where 0 represents not feasible and 1 represents feasible. Based on the scoring standards, the indicators are tiered using distance-based tiering, compatibility-based tiering, and natural breakpoint tiering standards. The distance-based tiering standard classifies indicators as follows: within 1 kilometer is suitable; between 1 and 3 kilometers is generally suitable; and more than 3 kilometers away is unsuitable.

[0010] Furthermore, each data point in the index dataset contains various evaluation indicators for the corresponding grid cell. On-site surveys are conducted in some research areas to determine the category labels of the training samples, and the corresponding grid cells are labeled as unsuitable, generally suitable, and suitable.

[0011] Furthermore, the evaluation of the indicator dataset based on meta-analysis and ensemble learning specifically includes: The study area i The suitability type determination formula for each grid is expressed as:

[0012] in, This is the overall suitability result. A comprehensive suitability result of 0 indicates unsuitability, a comprehensive suitability result of 1 indicates moderate suitability, and a comprehensive suitability result of 2 indicates suitability. This is the result of the restrictive indicator. A result of 0 indicates inappropriateness, and a result of 1 indicates appropriateness. The result is the judgment result of the unrestricted indicator ensemble learning. When the judgment result of the unrestricted indicator ensemble learning is 0, it indicates that it is inappropriate; when the judgment result of the unrestricted indicator ensemble learning is 1, it indicates that it is generally appropriate; and when the judgment result of the unrestricted indicator ensemble learning is 2, it indicates that it is appropriate. The result of the judgment of the limiting indicators The calculation formula is:

[0013] in, Indicates the first i The first grid j One limiting indicator, m The five restrictions are: marine ecological red line area, marine waterway area, marine engineering dumping area, marine anchorage area and marine port administrative area. The discrimination result of the non-restrictive indicator The calculation formula is:

[0014] in, It is the first k The first basic model for the second i The prediction results for each grid cell, It is a meta-learner; During training, the Z-score method is used to transform variables of different scales into a standard normal distribution, eliminating the interference of attribute scale differences on model training. Five-fold cross-validation is performed on each base model on the training set to obtain its prediction results on each fold of data, and these prediction results are used as new features to input into the meta-learner.

[0015] Compared with the prior art, the present invention has the following advantages: This invention systematically sorts out suitability evaluation indicators through meta-analysis and uses Large Language Model (LLM) for automated extraction and fusion, constructing a scientific and comprehensive suitability evaluation index system for island tourism and recreation, which has strong systematicity and representativeness.

[0016] This invention proposes a suitability evaluation framework based on restrictive discrimination and ensemble learning. First, a "one-vote veto" screening process is used through restrictive indicators to ensure policy compliance. Then, a stacking ensemble learning method is employed to finely classify unrestricted regions, significantly improving the model's accuracy and stability. Experimental results show that the model achieves comprehensive evaluation metrics (Accuracy, Precision, Recall, and F1-Score) exceeding 0.91 across the three study regions, averaging around 0.95, outperforming single-base models.

[0017] This invention has strong scalability and practical value. The proposed indicator system and machine learning framework can be applied to the suitability evaluation of marine tourism and recreation in different regions, and support the optimal allocation of marine tourism resources and sustainable development decision-making. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a framework diagram of the suitability evaluation method for island tourism marine development activities in this invention.

[0020] Figure 2This is a schematic diagram of the suitability hierarchical evaluation index system in this invention.

[0021] Figure 3 This is a schematic diagram of model performance evaluation in an embodiment of the present invention.

[0022] Figure 4 This is the activity suitability evaluation result for island 1 in the study area of ​​this invention.

[0023] Figure 5 This is the result of the activity suitability evaluation of island 2 in the study area of ​​this invention.

[0024] Figure 6 This is the result of the activity suitability evaluation of island 3 in the study area of ​​this invention. Detailed Implementation

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. 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.

[0027] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0028] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0029] like Figure 1 As shown, this invention provides a method for evaluating the suitability of marine development activities for island tourism based on meta-analysis and ensemble learning. The method includes systematically retrieving literature related to the themes of "marine suitability," "tourism suitability," and "island suitability" from SCI-indexed databases such as Web of Science and CNKI. Traditional manual statistical methods are inefficient for the large number of retrieved documents, while conventional bibliometric methods struggle to accurately extract indicator information. Therefore, this invention employs a generative large language model (LLM) to automatically and batch-process the extraction and statistical analysis of evaluation indicators from the selected effective literature. After obtaining initial indicator sets for the three dimensions of "marine use," "tourism," and "island," indicators with similar or identical meanings are first uniformly merged, and some macro-level indicators are refined based on the characteristics of marine tourism and recreation on islands. Then, indicators with low frequency, insufficient relevance, or unsuitability for small-scale evaluation are eliminated, ultimately constructing a hierarchical evaluation system containing 6 primary indicators and 26 secondary indicators.

[0030] A generative large language model is used to extract evaluation indicators and construct a hierarchical evaluation system. The hierarchical evaluation system includes primary indicators and secondary indicators. The primary indicators are divided into non-restrictive indicators and restrictive indicators. In specific implementation, as a preferred embodiment of the present invention, the primary indicators are Natural Conditions (NC), Marine Environment (ME), Tourism Resources (TR), Supporting Infrastructure (SI), Development Planning (DP), and Regional Constraints (RC). The first five are non-restrictive indicators, and the last one is a restrictive indicator.The secondary indicators NC include slope (SL), elevation (EL), and vegetation cover (VC); ME includes seawater quality (SWQ), seawater temperature (SWT), and seawater depth (SWD); TR includes the convenience of reef landscape (CRL), convenience of scenic spot (CSS), convenience of the beach (CB), and convenience of offshore recreation (COR); SI includes population density (PD), dining facilities (DF), accommodation facilities (AF), medical facilities (MF), amusement group (AP), public security (PS), public restrooms (PR), and transportation conditions (TC). DP includes Coastline Type (CT), Marine Functional Zoning (MFZ), and Sea Use Rights (SUR); RC includes Marine Ecological Redline (MER), Marine Channel Zone (MCZ), Marine Dumping Zone (MDZ), Marine Anchorage Zone (MAZ), and Marine Port Zone (MPZ). These indicators comprehensively cover key aspects such as natural conditions, marine environment, tourism resources, infrastructure, development planning, and regional constraints.

[0031] Based on the needs of the activity and the characteristics of sea use, the research area is delineated. In specific implementation, as a preferred embodiment of the present invention, the research area is delineated with the remotely interpreted island coastline as the baseline and a 3-kilometer buffer zone is made on the seaward side. Other land areas within the buffer zone are excluded, including the mainland and small islands. A 0.1×0.1-kilometer grid cell is constructed within the research area.

[0032] Based on a hierarchical evaluation system, the collected data is quantified and graded. Since the collected data is multi-source and heterogeneous and cannot be used directly, it needs to be quantified to assign values ​​to each indicator within the study area grid. Specifically, in a preferred embodiment of this invention, the quantification and grading of the collected data includes: The index quantification includes classifying the collected data types into land raster data, land vector data, sea surface raster data, and sea surface vector data; and using nearest neighbor analysis to assign values ​​to the grid cells of the study area in combination with the collected data.

[0033] For land vector data, which covers the landmass of islands, the Near Analysis operation in the Geographic Information System (GIS) software is used to assign values ​​to each grid cell in the study area to measure the impact of such indicators on sea use on the islands. Land raster data is similar to land vector data, but due to the varying resolutions of land raster data and the inability to perform direct spatial analysis, the Extract Values ​​to Points operation in the GIS software is first used to convert the land raster data into land vector data, and then the Near Analysis operation is used to assign values ​​to each grid cell in the study area. For sea surface vector data, the Near Analysis operation is directly used to assign values ​​to each grid cell in the covered study area. For sea surface raster data, such as seawater temperature, the Extract Values ​​to Points operation is used to assign values ​​to each grid cell in the covered study area.

[0034] Although the data after indicator quantification has numerical or typological measurement, the different dimensions make it impossible to evenly reflect the impact of each indicator on the evaluation results. Therefore, a tiered scoring system is still needed. The tiered scoring system assigns scores to both non-restrictive and restrictive indicators in the indicator system using a 0, 1, 3, and 5 scale. Non-restrictive indicators reflect the differentiation of suitability, using a 1, 3, and 5 scoring scale, where 1 represents unsuitable, 3 represents generally suitable, and 5 represents suitable. Restrictive indicators reflect the mandatory constraints of objective limitations on the suitability of activities, using a 0 and 1 scoring scale, where 0 represents not feasible and 1 represents feasible. Based on the scoring standards, indicators are graded using distance grading standards, compatibility grading standards, and natural breakpoint grading standards. For distance grading, indicators within 1 kilometer are suitable, those between 1 and 3 kilometers are generally suitable, and those more than 3 kilometers away are unsuitable. The compatibility grading standard grades indicators of the category measurement class (such as coastline type, marine functional zoning, and sea area use rights) based on the actual use of the category and its compatibility with island tourism and recreational sea use. For other indicators, the natural breakpoint method is used for classification.

[0035] The processed data are integrated to construct an indicator dataset. In a preferred embodiment of this invention, each data point in the indicator dataset contains various evaluation indicators for the corresponding grid cell. To ensure the reliability of model training, on-site surveys are conducted in some research areas to determine the category labels of the training samples, and the corresponding grid cells are labeled as unsuitable, moderately suitable, and suitable to form the training dataset. To ensure training effectiveness, the training data is divided into a training set and a validation set in a 7:3 ratio.

[0036] This paper evaluates the suitability of island activities by using meta-analysis and ensemble learning to assess the indicator dataset. In a preferred implementation, considering the complex nature of each indicator's contribution to the suitability of island tourism and recreation, the present invention proposes a method for evaluating island tourism and recreation based on restrictive and ensemble learning. The suitability of island tourism and recreation is determined by a combination of restrictive and non-restrictive indicator categories.

[0037] The study area i The suitability type determination formula for each grid is expressed as:

[0038] in, This is the overall suitability result. A comprehensive suitability result of 0 indicates unsuitability, a comprehensive suitability result of 1 indicates moderate suitability, and a comprehensive suitability result of 2 indicates suitability. This is the result of the restrictive indicator. A result of 0 indicates inappropriateness, and a result of 1 indicates appropriateness. The result is the judgment result of the unrestricted indicator ensemble learning. When the judgment result of the unrestricted indicator ensemble learning is 0, it indicates that it is inappropriate; when the judgment result of the unrestricted indicator ensemble learning is 1, it indicates that it is generally appropriate; and when the judgment result of the unrestricted indicator ensemble learning is 2, it indicates that it is appropriate. In the restrictive discrimination, if at least one of the five restrictive conditions is 0, meaning the current grid is within the restrictive index area, it is considered unsuitable. The result of the restrictive index discrimination is... The calculation formula is:

[0039] in, Indicates the first i The first grid j One limiting indicator, m The five restrictions are: marine ecological red line area, marine waterway area, marine engineering dumping area, marine anchorage area, and marine port administrative area.

[0040] For unrestricted metrics, the Stacking ensemble learning method is employed for discrimination. Its core advantage lies in its ability to integrate the strengths of multiple different base models, resulting in more powerful and stable predictive performance than any single model. Stacking, as a hierarchical ensemble strategy, trains multiple base models in the first layer to generate predictive outputs, and then uses a meta-learner in the second layer to synthesize these predictions, achieving efficient fusion and weighted optimization of information between base models. Considering both model complexity and generalization performance, four common and complementary classification models were selected: Decision Tree (DT), K-Nearest Neighbors (KNN), Random Forest (RF), and XGBoost (XGB) as base models, with Logistic Regression (LR) as the meta-learner. The discrimination results for the unrestricted metrics are presented below. The calculation formula is:

[0041] in, It is the first k The first basic model for the second i The prediction results for each grid cell, It is a meta-learner.

[0042] During training, the Z-score method is used to transform variables of different scales into a standard normal distribution, eliminating the interference of attribute scale differences on model training. Five-fold cross-validation is performed on each base model on the training set to obtain its prediction results on each fold of data, and these prediction results are used as new features to input into the meta-learner.

[0043] Example This embodiment describes the implementation steps of a proposed method for evaluating the suitability of islands for tourism and recreational use based on meta-analysis and machine learning: First, a systematic search of literature databases was conducted to retrieve literature related to "sea use," "tourism," and the suitability of "islands." Then, a large language model was used to automatically extract evaluation indicators from the selected literature. Next, the extracted indicators were merged and refined using synonyms, and irrelevant or inapplicable indicators were eliminated, ultimately forming a comprehensive evaluation system comprising 6 primary indicators and 26 secondary indicators. This system clearly distinguishes between non-restrictive indicators used to measure the quality of conditions and restrictive indicators that have a "one-vote veto" effect.

[0044] For the aforementioned indicators, data corresponding to each indicator were collected from publicly available data sources (such as electronic maps, electronic nautical charts, and remote sensing imagery). A fine-grained grid was established over the study area, and spatial analysis techniques from the geographic information system software ArcGIS were used to quantify and assign values ​​to each indicator within each grid cell. Finally, based on the standardized scoring of all indicator values, unrestricted indicators were divided into three levels: 1, 3, and 5, while restricted indicators were divided into two levels: 0 and 1.

[0045] Restricted zoning: First, a judgment is made based on restrictive indicators. Any grid that falls within any restricted zone is directly judged as "unsuitable".

[0046] Ensemble Learning Evaluation: For unrestricted regions, a Stacking ensemble learning method is used for fine-grained evaluation. This method first uses four basic models—decision tree, K-nearest neighbors, random forest, and gradient boosting tree—for preliminary prediction. Then, a logistic regression model is used to synthesize the results of each basic model, and finally, the suitability level of each grid is output ("suitable", "generally suitable", or "unsuitable").

[0047] By overlaying the restrictive zoning results with the machine learning evaluation results, a final spatial distribution map of the suitability of the island for marine tourism and recreation is generated, which intuitively displays the "unsuitable", "generally suitable" and "suitable" areas.

[0048] Figure 1 This is a technical roadmap for the method of evaluating the suitability of marine areas for island tourism and recreation proposed in this invention; Figure 2 This invention proposes an evaluation index system for the suitability of marine use for island tourism and recreation. Figure 3It is a performance evaluation of the model; Figure 4 These are the results of the suitability evaluation for marine use for tourism and recreation in study area 1; Figure 5 These are the results of the suitability evaluation for marine use for tourism and recreation in study area 2; Figure 6 This is the result of the suitability evaluation for marine use for tourism and recreation in study area 3.

[0049] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A suitability evaluation method for island tourism marine development activities based on meta-analysis and ensemble learning, characterized in that, include: A generative large language model is used to extract evaluation indicators and construct a hierarchical evaluation system. The hierarchical evaluation system includes primary indicators and secondary indicators, and the primary indicators are divided into non-restrictive indicators and restrictive indicators; Based on the needs of the activity and the characteristics of sea use, the research area was delineated. Based on the hierarchical evaluation system, the collected data will be quantified and graded. The processed data are integrated to construct an indicator dataset; The suitability assessment of island tourism marine development activities is achieved by evaluating the index dataset based on meta-analysis and ensemble learning.

2. The suitability evaluation method for island tourism marine development activities based on meta-analysis and ensemble learning according to claim 1, characterized in that, The non-restrictive indicators include: natural conditions, marine environment, tourism resources, infrastructure and development planning; the restrictive indicators are limiting conditions. The secondary indicators are refinements of the primary indicators. The natural conditions include: slope, elevation, and vegetation cover; the marine environment includes seawater quality, seawater temperature, and seawater depth; the tourism resources include the accessibility of reef landscapes, scenic spots, bathing beaches, and nearshore recreation; the infrastructure includes population density, catering facilities, accommodation facilities, medical facilities, recreational clusters, public safety, public restrooms, and transportation conditions; the development plan includes coastline type, marine functional zoning, and sea area use rights; and the restrictive conditions include marine ecological red line areas, marine navigation channel areas, marine engineering dumping areas, marine anchorage areas, and marine port administrative areas.

3. The suitability evaluation method for island tourism marine development activities based on meta-analysis and ensemble learning according to claim 1, characterized in that, The study area was delineated with a 3-kilometer buffer zone on the seaward side, based on the remotely sensed island coastline. Other land areas within the buffer zone, including mainland and small islands, were excluded. Grid cells were then constructed within the study area.

4. The suitability evaluation method for island tourism marine development activities based on meta-analysis and ensemble learning according to claim 1, characterized in that, The process of quantifying and grading the collected data includes: The index quantification includes classifying the collected data types into land raster data, land vector data, sea surface raster data, and sea surface vector data; and using nearest neighbor analysis to assign values ​​to the grid cells of the study area in combination with the collected data. The tiered scoring system assigns scores to both non-restrictive and restrictive indicators in the indicator system using a 0, 1, 3, and 5 scale. The non-restrictive indicators reflect the differentiation of suitability, using a 1, 3, and 5 scoring scale, where 1 represents unsuitable, 3 represents generally suitable, and 5 represents suitable. The restrictive indicators reflect the mandatory constraints of objective limitations on the suitability of activities, using a 0 and 1 scoring scale, where 0 represents not feasible and 1 represents feasible. Based on the scoring standards, the indicators are tiered using distance-based tiering, compatibility-based tiering, and natural breakpoint tiering standards. The distance-based tiering standard classifies indicators as follows: within 1 kilometer is suitable; between 1 and 3 kilometers is generally suitable; and more than 3 kilometers away is unsuitable.

5. The suitability evaluation method for island tourism marine development activities based on meta-analysis and ensemble learning according to claim 1, characterized in that, Each data point in the index dataset contains various evaluation indicators for the corresponding grid cell. On-site surveys were conducted in some study areas to determine the category labels of the training samples, and the corresponding grid cells were labeled as unsuitable, moderately suitable, and suitable.

6. The suitability evaluation method for island tourism marine development activities based on meta-analysis and ensemble learning according to claim 1, characterized in that, The evaluation of the indicator dataset based on meta-analysis and ensemble learning specifically includes: The study area i The suitability type determination formula for each grid is expressed as: in, This is the overall suitability result. A comprehensive suitability result of 0 indicates unsuitability, a comprehensive suitability result of 1 indicates moderate suitability, and a comprehensive suitability result of 2 indicates suitability. This is the result of the restrictive indicator. A result of 0 indicates inappropriateness, and a result of 1 indicates appropriateness. The result is the judgment result of the unrestricted indicator ensemble learning. When the judgment result of the unrestricted indicator ensemble learning is 0, it indicates that it is inappropriate; when the judgment result of the unrestricted indicator ensemble learning is 1, it indicates that it is generally appropriate; and when the judgment result of the unrestricted indicator ensemble learning is 2, it indicates that it is appropriate. The result of the determination of the limiting indicators The calculation formula is: in, Indicates the first i The first grid j One limiting indicator, m The five restrictions are: marine ecological red line area, marine waterway area, marine engineering dumping area, marine anchorage area and marine port administrative area. The discrimination result of the non-restrictive indicator The calculation formula is: in, It is the first k The first basic model for the second i The prediction results for each grid cell, It is a meta-learner; During training, the Z-score method is used to transform variables of different scales into a standard normal distribution, eliminating the interference of attribute scale differences on model training. Five-fold cross-validation is performed on each base model on the training set to obtain its prediction results on each fold of data, and these prediction results are used as new features to input into the meta-learner.

Images