A construction method of a port geographic information service platform based on GIS

Abstract

The present application discloses a method for constructing a port geographic information service platform based on GIS, belonging to the technical field of information service platform construction. The method includes: constructing a GIS data model; performing multi-source collection on port geographic information data, extracting features related to berth demand prediction from the multi-source data and then fusing them; storing the fused data into the GIS model, analyzing the fused data, selecting the first principal component column according to the magnitude of the feature values, and constructing a berth demand model based on the first principal component column; adjusting the berth demand model based on the SHAP value analysis method; deploying the final berth demand model to the port environment for real-time prediction, and visualizing the prediction results on the GIS platform for decision support. This method comprehensively considers multiple aspects such as data processing, model construction, security factors, real-time prediction, and personalized recommendation, significantly improving the overall efficiency and service quality of port management.

Description

Technical Field

[0001] The present invention relates to the technical field of information service platform construction, and particularly to a method for constructing a port geographic information service platform based on GIS. Background Art

[0002] With the rapid development of port business and the continuous improvement of the requirements for logistics transportation efficiency, port management is facing increasing challenges. Especially in berth allocation, how to efficiently and reasonably arrange berth resources to meet the ship docking needs and improve port operation efficiency has become an urgent problem to be solved. Most of the existing berth allocation methods rely on manual experience or simple rules and are difficult to adapt to the complex and changeable port operation environment. In order to improve port operation efficiency and service quality, many ports have begun to adopt advanced information technology and data analysis technology to optimize berth management and resource allocation. Among them, the Geographic Information System (GIS), as a powerful spatial data management and analysis tool, plays an important role in port geographic information management.

[0003] In the prior art, the sources of port geographic information data are diverse, including but not limited to ship position information, cargo information, meteorological data, port facility data, etc. These data usually come from different data sources and have different formats. There are challenges in data management in the prior art, and it is difficult to effectively integrate these multi-source data and capture the influencing factors of berth demand, resulting in inaccurate prediction results and thus affecting the effective utilization of berth resources. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for constructing a port geographic information service platform based on GIS to solve the problems raised in the above background art.

[0005] One aspect of the present application provides a method for constructing a port geographic information service platform based on GIS, including:

[0006] S5000: Construct a GIS data model;

[0007] S200: Receive multi-source collection of port geographic information data from multiple collectors, respectively extract m features related to berth demand prediction from the multi-source data, fuse the data extracted from multiple data sources, and store the fused data in the GIS data model;

[0008] S300: Analyze the fused data, and successively select the eigenvectors corresponding to k largest eigenvalues as the first principal component columns according to the eigenvalue magnitudes, and construct a berth demand model based on the first principal component columns;

[0009] S400: Based on the SHAP value analysis method, identify the key factors affecting the port berth demand and their interactions, judge the impacts of the key factors and the interactions on the berth demand model, and give the second principal component column based on the key factors and their interactions;

[0010] S500: Judge whether the corresponding eigenvectors of the second principal component column and the first principal component column are consistent. If so, still use the original berth demand model; if not, adjust the berth demand model;

[0011] S600: Deploy the final berth demand model to the port environment for real-time prediction, and visualize the prediction results on the GIS platform for decision support;

[0012] S700: Give the optimal berth allocation suggestion based on the prediction results and the user's historical behavior and preferences.

[0013] As an alternative solution to the technical solution of this application, the analysis of the data in the time series database includes:

[0014] S310: Standardize the data in the time series database;

[0015] S320: Calculate the covariance matrix, and calculate the corresponding eigenvalues and eigenvectors based on the covariance matrix;

[0016] S330: Sort the eigenvalues in descending order based on their magnitudes, and select the eigenvectors corresponding to the k largest eigenvalues that can explain at least 80% of the variance as the principal components to form the first principal component column;

[0017] S340: Select the first principal component column as the new feature axis, and project the original data into the low-dimensional space corresponding to the k largest eigenvalues.

[0018] As an alternative solution to the technical solution of this application, in step S200, after storing the fused data into the GIS data model, establish an indexing mechanism for the corresponding data based on the GIS data model.

[0019] As an alternative solution to the technical solution of this application, the construction of the berth demand model includes:

[0020] S350: Create input-output pairs with the data projected into the low-dimensional space corresponding to the k largest eigenvalues, where the input corresponds to the values of the past p time steps and the output is the value of the next time step; and divide the input-output pairs into a training data set Y1 and a test data set Y2;

[0021] S360: Construct an LSTM model;

[0022] S370: Train the LSTM model based on the training data set.

[0023] As an alternative embodiment of the technical solution of this application, step S400 includes:

[0024] S410: Calculate SHAP values based on the test data set;

[0025] S420: Determine k key features according to the absolute value of the SHAP values; and calculate the SHAP interaction values between the k key features. If the SHAP interaction value between any two key features is not greater than a preset value, then use the k key features determined according to the absolute value of the SHAP values as the second principal component column;

[0026] S430: If the SHAP interaction value between two key features A and B is greater than a preset threshold, establish a new key feature AB based on the key feature A and the key feature B; after establishing the new key feature, select key features again according to the magnitude of the feature values, so that the number of key features remains k, and the k key features form the second principal component column.

[0027] As an alternative embodiment of the technical solution of this application, determining whether the corresponding feature vectors of the second principal component column and the first principal component column in step S500 includes:

[0028] Whether the feature vectors in the second principal component column are all of the feature vectors in the first principal component column;

[0029] Whether the order of the feature vectors in the second principal component column is the same as the order of the feature vectors in the first principal component column;

[0030] If the feature vectors in the second principal component column are not all of the feature vectors in the first principal component column, and / or the order of the feature vectors in the second principal component column is inconsistent with the order of the feature vectors in the first principal component column, it is determined that the corresponding feature vectors of the second principal component column and the first principal component column are inconsistent.

[0031] As an alternative embodiment of the technical solution of this application, step S700 includes:

[0032] S710: Collect the user's berth usage history and preference information;

[0033] S720: Determine the available status of berths within a future period based on the prediction results of the berth demand model; formulate a berth allocation strategy for specific users in combination with the collected user berth usage history and preference information;

[0034] S730: Solve the berth allocation optimization based on linear programming:

[0035]

[0036] subject to ,

[0037] ,

[0038] ;

[0039] wherein is the cost vector, is the decision variable vector, and are the coefficient matrices of the equality constraint and the inequality constraint respectively, and are the right - hand - side vectors of the equality constraint and the inequality constraint respectively, and are the lower and upper bounds of the variables.

[0040] As an alternative solution of the technical solution of this application, it further includes:

[0041] S800: Generate a recommendation system based on the preferences of historical users;

[0042] Determine the available state of berths within a future period based on the prediction result of the berth demand model, and combine with the recommendation system to specify a berth allocation strategy for new users.

[0043] Another aspect of this application provides a device for constructing a port geographic information service platform, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the foregoing method are implemented.

[0044] A third aspect of this application provides a computer - readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the foregoing method are implemented.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] The technical solution described in this application collects and integrates multi-source data, establishes a time-series database and an indexing mechanism, greatly improving the integrity, accuracy, and availability of data, which provides a high-quality data foundation for subsequent analysis. Using the principal component analysis method to construct a berth demand model can effectively capture the main factors affecting demand and improve the accuracy of prediction. By introducing the SHAP value analysis method to identify the key factors and their interactions affecting port berth demand, the model comprehensively considers safety factors and improves the reliability of prediction. By comparing two principal component columns, the self-testing and dynamic adjustment of the model are realized, enhancing the adaptability and stability of the model. Deploying the model to the actual environment for real-time prediction and visualizing it on the GIS platform provides an intuitive and timely decision-making support tool for managers. Based on the prediction results and the user's historical behavior, the optimal berth allocation suggestions are given, realizing the efficient utilization of resources and the improvement of user satisfaction.

[0047] This method comprehensively considers multiple aspects such as data processing, model construction, safety factors, real-time prediction, and personalized recommendation, significantly improving the overall efficiency and service quality of port management. Through the application of data-driven and intelligent algorithms, it lays a foundation for the intelligentization and automation of port operation. Brief Description of the Drawings

[0048] Figure 1 It is a flowchart of the construction method of the GIS-based port geographic information service platform described in the embodiments of this application. Detailed Embodiments

[0049] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0050] The embodiments of this application record a construction method of a GIS-based port geographic information service platform. The construction method mainly includes the following steps:

[0051] S100: Construct a GIS data model;

[0052] In this step, it is necessary to define the entities, attributes in the GIS data model, and the relationships between them. In a specific embodiment of the present application, for the convenience of constructing the subsequent berth demand model and giving optimal berth allocation suggestions, the entities in the GIS data model are defined as berths, users, ships, and weather conditions, and their attributes are correspondingly defined as the location coordinates of the berth, berth type, berth capacity, user ID, user preferences, etc.; the relationships include the relationship between the berth and the user, the relationship between the berth and the berth demand, etc.

[0053] S200: Perform multi-source collection on the port geographical information data, respectively extract m features related to berth demand prediction from the multi-source data, and fuse the data extracted from multiple data sources; connect the fused data to the GIS data model and establish a data indexing mechanism.

[0054] Among them, the multi-source collection of port geographical information includes sensor data, satellite images, aerial photos, meteorological data, etc. Due to the diversity of the above data sources, it is necessary to preprocess the data before data fusion, including removing error data and duplicate data, time series alignment of data, etc. Subsequently, the multi-source data are integrated together to form a unified data set, and the above unified data set is converted into a format that can be processed by GIS, such as converting the coordinates into a coordinate system supported by GIS, which is convenient for connecting the fused data to the GIS data model.

[0055] In addition, in the above step, the establishment of the data indexing mechanism is mainly to accelerate the query speed of the data and facilitate the subsequent connection of the port geographical information collected to the corresponding position of the GIS data model; the data indexing mechanism therein can be a spatial index of the R-tree or Quadtree type, or a time index that can quickly locate a specific time point or time period. In a specific embodiment of the present application, for the convenience of creating input-output pairs based on the time sequence subsequently, its indexing mechanism is a time index, for example, it can be a sparse index that only records the index when the data changes.

[0056] S300: Analyze the fused data, and successively select the eigenvectors corresponding to the k largest eigenvalues as the first principal component columns according to the eigenvalue magnitudes, and construct a berth demand model based on the first principal component columns. It can be understood that in this process, k ≤ m.

[0057] Among them, the construction process of the first principal component column is as follows:

[0058] S310: Perform standardization processing on the fused data:

[0059] ;

[0060] Among them, x is the original data point, μ is, and σ is the standard deviation of this feature;

[0061] S320: Calculate the covariance matrix, and calculate the corresponding eigenvalues and eigenvectors based on the covariance matrix;

[0062] Among them, the covariance matrix reflects the relationship between each feature. For n samples of m features, the calculation method of the covariance matrix C is:

[0063] ;

[0064] Among them, X is the standardized data matrix, and each column corresponds to a feature.

[0065] Subsequently, calculate the eigenvalues λ1, λ2,..., λ m and the corresponding eigenvectors v1, v2,..., v m . Among them, the eigenvalue represents the variance magnitude along the corresponding feature direction;

[0066] S330: Arrange in descending order according to the magnitude of the eigenvalues. The eigenvector corresponding to the larger eigenvalue represents a more important principal component. Based on the magnitude of the eigenvalues, arrange in descending order, and select the eigenvectors corresponding to the k largest eigenvalues that can explain at least 80% of the variance as the first principal component column.

[0067] In a specific implementation, select the k eigenvectors with the largest eigenvalues from the above m eigenvectors such as v1, v2,..., v m to form the first principal component column.

[0068] Subsequently, perform step S340, that is, use the selected first principal component column as the new feature axis, and project the original data into the low-dimensional space corresponding to the k largest eigenvalues:

[0069] For each sample x, the new feature vector y can be expressed as:

[0070] ;

[0071] Among them, V is the matrix composed of the selected k eigenvectors;

[0072] The projected data is:

[0073] ;

[0074] X is the standardized data matrix.

[0075] When the projected data Y is obtained, it is necessary to construct a berth demand model based on this projected data, which specifically includes the following steps:

[0076] S350: Create input-output pairs from the data projected onto the low-dimensional space corresponding to the k largest eigenvalues, where the input corresponds to the values of the past p time steps and the output is the value of the next time step:

[0077] In a specific solution, for the time series Y = [y1, y2, y3, …, y t , if you want to predict the value from time t + 1 to time t, the input is the value from time t - p to time t, and the output is the value at time t + 1; that is, when predicting the value of the next time step in the future, the input-output pairs are created as follows:

[0078] Input: [y1, y2, y3, …, y t-p ;

[0079] Output: [y t-p+1 .

[0080] Subsequently, the input-output pairs need to be divided into a training data set Y1 and a test data set Y2. Among them, the training data set Y1 contains earlier data, and the test data set Y2 contains later data. In a specific example of this application, according to the time corresponding to the data, 80% of the data at an earlier time is selected for Y1, and 20% of the data at a later time is selected for Y2.

[0081] S360: Build an LSTM model. The LSTM model is a special type of RNN (Recurrent Neural Network). Because it can remember long-term dependence information, it is more suitable for processing time series and is more applicable to berth demand prediction.

[0082] In the specific process of building the LSTM model, deep learning frameworks such as Keras or TensorFlow can be used. Those skilled in the art can implement the construction of the LSTM model with languages such as Python, so it will not be elaborated in the specific embodiments of this application.

[0083] S370: Train the LSTM model; the LSTM model will learn how to predict the value of the next time step based on the given input sequence. For example, when the input sequence is [y1, y2, y3], the goal of the LSTM model is to predict the value of y4. In the embodiments of this application, the training of the LSTM model is based on the training data set.

[0084] In the traditional model construction process, a training dataset is often used to train the model, and then a test dataset is used to test the trained model to evaluate the performance of the model. In this solution, since the training dataset and the test dataset have the same source, when they are used to train and test the same model respectively, the interaction between different features is often difficult to reflect. Therefore, in the embodiments of the present application, the interaction between features and the adjustment of the LSTM model are based on the SHAP value analysis in steps S400 to S500.

[0085] S400: Based on the SHAP value analysis method, identify the key factors affecting the port berth demand and their interactions, and judge the impacts of the key factors and the interactions on the berth demand model. Give the second principal component column based on the key factors and their interactions. Specifically, it includes the following steps:

[0086] S410: For a single prediction, the SHAP value is expressed as:

[0087] ;

[0088] where f(y) is the predicted value of the LSTM model;

[0089] is the baseline predicted value, and in this embodiment, the average value of the LSTM model predicted values is selected;

[0090] is the SHAP value of feature i;

[0091] y i is the value of feature i;

[0092] k is the total number of feature values.

[0093] According to the above calculation method, calculate the SHAP values of the factors characterized by the k feature vectors selected based on the maximum eigenvalue, and sort the k key features according to the absolute value of each SHAP value.

[0094] Among them, step S420 mainly includes the following sub-steps:

[0095] S421: Calculate the SHAP interaction values between every two of the k features:

[0096] For two features i and j, the SHAP interaction value between them is expressed as:

[0097] ;

[0098] where represents the contribution of feature i under the condition that feature j already exists;

[0099] Represents the contribution of feature j given the presence of feature i;

[0100] and represent the independent SHAP values of features i and j respectively.

[0101] In the above calculation method, the contribution of feature i given the presence of feature j is calculated as follows:

[0102] ;

[0103] where Y is the set of all features, represents the set of all features except features i and j;

[0104] is the predicted value when the feature subset S is added with features i and j;

[0105] is the predicted value when the feature subset S is added with feature j;

[0106] The feature subset S represents all subsets excluding features i and j, and .

[0107] The contribution of j given the presence of feature i is calculated using the same principle. .

[0108] S422: Determine whether the SHAP interaction value between any two key features is greater than a preset value. If so, go to step S430; if not, go to step S500;

[0109] S430: When the SHAP interaction value between two of the key features A and B is greater than a preset threshold, a new key feature AB is established based on the key feature A and the key feature B; after establishing the new key feature, the original key feature A and the original key feature B are discarded and replaced with the new key feature AB. At this time, the number of key features decreases, so key feature supplementation is required. In the embodiments of the present application, the (k + 1)-th key feature is selected from the features sorted in descending order based on the feature values in step S330, so that the number of key features remains k, and the k key features form the second principal component column; if the SHAP interaction value between another two key features is also greater than the preset threshold, a new feature is created according to the foregoing steps, and then the feature vectors sorted according to the feature values in step S330 are sequentially selected; if the number of feature vectors in step S330 is difficult to meet the reduction in the number of key features caused by the SHAP interaction value being greater than the preset threshold, then it is necessary to consider adjusting the threshold of the preset SHAP interaction value.

[0110] The creation method of the above new key feature AB is based on specific situations. For example, in some embodiments, the interaction between feature A and feature B can be captured in a multiplicative manner. In this case, the new feature AB = A × B; in some embodiments, new features can also be created through function combination. In this case, the new feature AB = g(A) + h(B).

[0111] After creating the new key feature, the adjusted features are used as the second principal component column. At this time, since the new key feature is generated, its corresponding feature vector is different from that of the original first principal component column. Subsequently, with this second principal component column as the new feature axis, the original data is projected and then the berth demand model is constructed.

[0112] S500: The k key features determined by sorting based on the absolute magnitude of the SHAP values are used as the second principal component column. At this time, the k key features in the second principal component column are the same as the k key features in the first principal component column. Therefore, it is only necessary to determine whether the order of the k key features in the second principal component column is the same as the sorting of the k key features in the first principal component column. If the two sorts are the same, the original berth demand model is continued to be used; if the two sorts are different, the second principal component column is used as the new feature axis, and the original data is projected and then the berth demand model is constructed.

[0113] For the construction of the berth demand model in steps S500 and S430, the ideas in steps S350 to S370 are adopted, and the methods are slightly different. The reconstruction of the berth demand model is mainly carried out in the following ways:

[0114] 1) Create input-output pairs with the data projected into the low-dimensional space corresponding to the k eigenvalues of the second principal component column. Here, the input still corresponds to the values of the past p time steps, and the output still corresponds to the value of the next time step. The selection of the time step is based on actual needs. For example, in some ports with a large ship flow, the time step is shorter, perhaps 0.5h; in some ports with a small ship flow, the time step may be 4h. The created input-output pairs are no longer divided into training data sets and test data sets, and the data corresponding to these input-output pairs are all training data.

[0115] 2) Construct a new LSTM model.

[0116] 3) Train the new LSTM model based on the input-output pairs created in 1).

[0117] The model obtained in 3) above is the final berth demand model. Subsequently, this berth demand model can be deployed to the port environment for real-time prediction, and the prediction results are visualized on the GIS platform.

[0118] When deploying the final berth demand model to the port environment for real-time prediction, optimal berth allocation suggestions can be given based on the prediction results and the user's historical behavior and preferences, that is, step S700.

[0119] This step is specifically carried out in the following manner:

[0120] S710: Collect the user's berth usage history and preference information; the usage history includes the frequency and time period of the user's past berth usage, etc.; the preference generally refers to the berth location and berth type preferred by the user, etc.

[0121] S720: Determine the available state of berths in the future period based on the prediction results of the berth demand model; formulate a berth allocation strategy for specific users in combination with the collected user berth usage history and preference information;

[0122] S730: Solve the berth allocation optimization based on linear programming:

[0123]

[0124] subject to ,

[0125] ,

[0126] ;

[0127] where, is the cost vector, is the decision variable vector, and and are the coefficient matrices of the equality constraints and inequality constraints respectively, and are the right - hand - side vectors of the equality constraints and inequality constraints respectively, and are the lower and upper bounds of the variables.

[0128] In the above linear programming:

[0129] is the objective function, which represents the quantity we hope to minimize;

[0130] Equality constraint means that the berth demand is equal to the predicted demand, where is a matrix, each row corresponding to a berth location and each column corresponding to a user; each row of is a unit vector, representing the allocation of berth locations, is a vector, representing the predicted demand value for each berth location.

[0131] Inequality constraint means that the berth allocation does not exceed the available berth quantity, where, is a matrix, each row corresponding to a berth location and each column corresponding to a user; each row of is a unit vector, representing the allocation of berth locations; is a vector, representing the maximum available berth quantity for each berth location;

[0132] Lower bound is a vector, each element being 0, indicating that the berth allocation quantity cannot be less than 0;

[0133] Upper bound is a vector, each element representing the maximum available berth number for each berth location.

[0134] The above step S700 can give the optimal berth allocation suggestion based on the historical behaviors and preferences of historical users. However, for a certain port, there will inevitably be some new users. In the technical solution of this application, considering the problem that new users do not have historical behaviors and preferences, an allocation strategy for new users is given.

[0135] The berth allocation strategy for new users is recorded in step S800. Specifically, it can be carried out in the following way:

[0136] S810: Data collection and pre - processing

[0137] Collect the behavior data of all historical users, including berth reservation records, berth usage frequencies, docking time lengths, etc.;

[0138] Based on the collected data, a user preference model can be constructed. The clustering algorithm can be used to divide users into different groups, and each group has similar berth demand preferences.

[0139] S820: New User Feature Collection

[0140] For new users, their needs can be initially understood by asking new users about information such as their vessel type, cargo type, expected docking time, etc.

[0141] S830: User Similarity Calculation

[0142] Use the information provided by the new user to match with the preference models of historical users to find the most similar group of users;

[0143] If the new user already has some initial behavior data (such as past berth reservations), the collaborative filtering method can be adopted to find other users with similar behaviors to the new user.

[0144] S840: Berth Recommendation Generation

[0145] Use the previously established berth demand model to predict the berth demand situation;

[0146] Combining the characteristics of the new user and the behavior data of historical users, generate a recommended berth list for the new user, which can be based on factors such as berth availability, location, facility configuration, etc.

[0147] S850: Recommendation Result Optimization

[0148] Based on the additional information provided by the new user (such as special needs, time window, etc.), further adjust the recommendation list, and encourage the new user to provide feedback for further optimizing the recommendation algorithm in the future.

[0149] The above are only the preferred embodiments of the present application and are not used to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A method for constructing a port geographic information service platform based on GIS, which is applied to a device for constructing a port geographic information service platform based on GIS, and is characterized in that, Including: S100: Construct a GIS data model; S200: Receive multi-source collection of port geographical information data from multiple collectors, respectively extract m features related to berth demand prediction from the multi-source data, fuse the data extracted from multiple data sources, and store the fused data in the GIS data model; S300: Analyze the fused data, sequentially select the eigenvectors corresponding to k largest eigenvalues as the first principal component column according to the eigenvalue magnitudes, and construct a berth demand model based on the first principal component column; Among them, analyzing the fused data includes: S310: Standardize the fused data; S320: Calculate the covariance matrix, and calculate the corresponding eigenvalues and eigenvectors based on the covariance matrix; S330: Sort the eigenvalues in descending order, and select the eigenvectors corresponding to k largest eigenvalues that can explain at least 80% of the variance as the principal components to form the first principal component column; S340: Select the first principal component column as the new feature axis, and project the original data into the low-dimensional space corresponding to the k largest eigenvalues; Among them, the constructing of the berth demand model includes: S350: Create input-output pairs from the data projected onto the low-dimensional space corresponding to the k largest eigenvalues, where the input corresponds to the values of the past p time steps and the output is the value of the next time step; and divide the input-output pairs into a training data set Y 1 and a test data set Y 2; S360: Construct an LSTM model; S370: Train the LSTM model based on the training dataset; S400: Based on the SHAP value analysis method, identify the key factors affecting port berth demand and their interactions, judge the influence of the key factors and the interactions on the berth demand model, and give the second principal component column based on the key factors and their interactions; Among them, the step S400 includes: S410: Calculate the SHAP values based on the test dataset; determine k key features according to the absolute value magnitudes of the SHAP values; S420: Calculate the SHAP interaction values between the k key features. If the SHAP interaction value between any two key features is not greater than the preset value, then use the k key features determined by sorting according to the absolute value magnitudes of the SHAP values as the second principal component column; S430: If the SHAP interaction value between two key features A and B is greater than the preset threshold, establish a new key feature AB based on the key feature A and the key feature B; after establishing the new key feature, re-select the key features according to the eigenvalue magnitudes so that the number of key features remains k, and these k key features form the second principal component column; S500: Judge whether the eigenvectors corresponding to the second principal component column and the first principal component column are consistent; if so, still use the original berth demand model; if not, then use the second principal component column as the new feature axis, project the original data and adjust the berth demand model; S600: Deploy the final berth demand model to the port environment for real-time prediction, and visualize the prediction results on the GIS platform for decision support; S700: Give the optimal berth allocation suggestion based on the prediction results and the user's historical behavior and preferences.

2. The method according to claim 1, characterized in that, In the step S200, after storing the fused data into the GIS data model, an indexing mechanism for the corresponding data based on the GIS data model is established.

3. The method according to claim 1, characterized in that, In the step S500, the judgment of whether the eigenvectors corresponding to the second principal component column and the first principal component column are consistent includes: Whether the eigenvectors in the second principal component column are all of the eigenvectors in the first principal component column; Whether the order of the eigenvectors in the second principal component column is the same as the order of the eigenvectors in the first principal component column; If the eigenvectors in the second principal component column are not all of the eigenvectors in the first principal component column, and / or the order of the eigenvectors in the second principal component column is inconsistent with the order of the eigenvectors in the first principal component column, it is determined that the eigenvectors corresponding to the second principal component column and the first principal component column are inconsistent.

4. The method according to claim 1, wherein The step S700 includes: S710: Collect the historical berth usage and preference information of users; S720: Determine the available status of berths in a future period based on the prediction result of the berth demand model; formulate a berth allocation strategy for specific users in combination with the collected historical berth usage and preference information of users; S730: Solve the berth allocation optimization based on linear programming: ； subject to , ， ； Among them, is the cost vector, is the decision variable vector, is the objective function; and are the coefficient matrices of the equality constraint and the inequality constraint respectively, and are the right - hand side vectors of the equality constraint and the inequality constraint respectively, represents the demand forecast value for each berth location, represents the maximum available berth number for each berth location; Equality constraint Indicates that the berth demand is equal to the predicted demand, inequality constraint Indicates that the berth allocation does not exceed the available berth quantity; and are the lower and upper bounds of the variable. The lower bound is a vector with each element being 0, indicating that the berth allocation quantity cannot be less than 0; the upper bound is a vector with each element representing the maximum available number of berths at each berth location.

5. The method according to claim 1, wherein It further includes: S800: Generate a recommendation system based on the preferences of historical users; Determine the available status of berths in a future period based on the prediction result of the berth demand model, and formulate a berth allocation strategy for new users in combination with the recommendation system.

6. A device for constructing a port geographic information service platform based on GIS, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-5.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1-5.

Images