Method and system for constructing digital twin model of spatiotemporal data association based on dynamic form

Abstract

The embodiment of the present invention provides a method and system for constructing a spatiotemporal data-associated digital twin model based on a dynamic form, and belongs to the field of data processing technology. The method includes: responding to the interactive behavior of the target user on the interface, and obtaining the form fields adjusted by the target user according to the interactive behavior; generating a corresponding target form according to the form fields, and obtaining the initial spatiotemporal data linked to the target form from the target database; performing abnormal identification on the initial spatiotemporal data to obtain target abnormal data, and obtaining the corresponding target abnormal position according to the target abnormal data; filling the target abnormal position with data according to the initial spatiotemporal data, and obtaining the target spatiotemporal data linked to the target form; determining the corresponding initial twin model according to the form fields, and constructing the initial twin model according to the target spatiotemporal data to obtain the target twin model.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for constructing a spatiotemporal data associated digital twin model based on a dynamic form. Background Art

[0002] The twin model construction process is usually based on predetermined requirements and data sets. Specifically, developers will directly collect relevant data based on specific application scenarios or research objectives, and generate corresponding twin models based on this. Although this method can meet basic modeling needs, it has certain limitations in terms of flexibility and user-friendliness. Existing twin model construction methods lack consideration for user personalized needs, and often cannot automatically identify and adapt to user personalized needs, thereby reducing user satisfaction. Summary of the invention

[0003] The main purpose of the embodiments of the present invention is to provide a method and system for constructing a digital twin model of spatiotemporal data association based on a dynamic form, aiming to solve the problem that the existing twin model construction methods lack consideration for user personalized needs, and are often unable to automatically identify and adapt to user personalized needs, thereby reducing user satisfaction.

[0004] In a first aspect, an embodiment of the present invention provides a method for constructing a spatiotemporal data association digital twin model based on a dynamic form, comprising:

[0005] Responding to the target user's interactive behavior on the interface, and obtaining the form fields adjusted by the target user according to the interactive behavior;

[0006] Generate a corresponding target form according to the form fields, and obtain the initial spatiotemporal data linked to the target form from a target database;

[0007] Performing anomaly identification on the initial spatiotemporal data to obtain target anomaly data, and obtaining a corresponding target anomaly position according to the target anomaly data;

[0008] Filling the target abnormal position with data according to the initial spatiotemporal data to obtain the target spatiotemporal data linked to the target form;

[0009] The corresponding initial twin model is determined according to the form fields, and the initial twin model is constructed according to the target spatiotemporal data to obtain the target twin model.

[0010] In a second aspect, an embodiment of the present invention provides a system for constructing a spatiotemporal data-associated digital twin model based on a dynamic form, comprising:

[0011] A data acquisition module, used to respond to the target user's interactive behavior on the interface and obtain the form fields adjusted by the target user according to the interactive behavior;

[0012] A data generation module, used to generate a corresponding target form according to the form fields, and obtain the initial spatiotemporal data linked to the target form from a target database;

[0013] An anomaly identification module, used to perform anomaly identification on the initial spatiotemporal data to obtain target anomaly data, and obtain a corresponding target anomaly position according to the target anomaly data;

[0014] A data filling module, used to fill the target abnormal position with data according to the initial spatiotemporal data, and obtain the target spatiotemporal data linked to the target form;

[0015] A model building module is used to determine the corresponding initial twin model according to the form fields, and to build the initial twin model according to the target spatiotemporal data to obtain the target twin model.

[0016] The embodiment of the present invention provides a method and system for constructing a spatiotemporal data-associated digital twin model based on a dynamic form, the method comprising: by responding to the interactive behavior of the target user on the interface, and then obtaining the form field adjusted by the target user according to the interactive behavior, so that the form field can be adjusted in real time to meet the personalized needs of the user. This interactive operation mode can significantly improve the user's experience, is more flexible and close to the actual needs of the user, and then obtains the initial spatiotemporal data linked to the adjusted form field from the target database, which can ensure the accuracy and pertinence of the data. This avoids the risk of using wrong or outdated data, and provides a reliable basis for subsequent data processing and model construction. Then, by performing abnormal identification on the initial spatiotemporal data, the target abnormal data and its corresponding target abnormal position can be discovered and located in time, so that the target abnormal position is filled with data according to the initial spatiotemporal data, and more complete and comprehensive target spatiotemporal data can be obtained. This helps to make up for the impact of data missing or abnormal, improve the quality and availability of data, and finally determine the corresponding initial twin model according to the form field, and build the model in combination with the target spatiotemporal data, so as to obtain a more accurate and effective target twin model. This model building method based on actual data and user needs can significantly improve the practicality and performance of the model, and provide strong support for subsequent applications and decision-making. This method has brought significant beneficial effects in improving user experience, improving data accuracy, enhancing data integrity, optimizing model building, etc. It also solves the problem that the existing twin model building methods lack consideration for user personalized needs, often fail to automatically identify and adapt to user personalized needs, thereby reducing user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A schematic diagram of a process for constructing a spatiotemporal data-associated digital twin model based on a dynamic form provided in an embodiment of the present invention;

[0019] Figure 2 A schematic diagram of the module structure of a system for constructing a spatiotemporal data associated digital twin model based on a dynamic form provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0022] It should be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0023] The embodiment of the present invention provides a method and system for constructing a digital twin model of spatiotemporal data association based on a dynamic form. The method for constructing a digital twin model of spatiotemporal data association based on a dynamic form can be applied to a terminal device, which can be an electronic device such as a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device. The terminal device can be a server or a server cluster.

[0024] Some embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0025] Please refer to Figure 1 , Figure 1 A flowchart of a method for constructing a spatiotemporal data association digital twin model based on a dynamic form is provided in an embodiment of the present invention.

[0026] like Figure 1 As shown, the method for constructing a spatiotemporal data associated digital twin model based on a dynamic form includes steps S101 to S105.

[0027] Step S101: respond to the target user's interactive behavior on the interface, and obtain the form fields adjusted by the target user according to the interactive behavior.

[0028] Exemplarily, an intuitive and user-friendly interface layout is designed to ensure that all form fields are displayed and selectable in the interface. In addition, clear labels and instructions can be used to help users understand the function and purpose of each field, and then event listeners are integrated into the interface to capture various interactive behaviors of the target user, such as clicks, inputs, selections, etc., so as to monitor the user's operations on the interface in real time to obtain interactive behaviors, and then obtain the form fields selected by the target user in the interface based on the interactive behaviors.

[0029] Step S102: Generate a corresponding target form according to the form fields, and obtain initial spatiotemporal data linked to the target form from a target database.

[0030] Exemplarily, the name, type and value of each field are determined according to the form field, and then the parsed form field is mapped to the predefined form structure. Ensure that each form field has a corresponding definition and position in the target form. Thus, a suitable form template is selected according to the type and content of the form field to generate the corresponding target form.

[0031] Exemplarily, the initial spatiotemporal data corresponding to each form field is obtained from the target database, and then the initial spatiotemporal data corresponding to each form field is stored in the corresponding target form, thereby linking the data of the form field in the target database with the target form.

[0032] Step S103: performing anomaly identification on the initial spatiotemporal data to obtain target anomaly data, and obtaining a corresponding target anomaly position according to the target anomaly data.

[0033] Exemplarily, an unsupervised learning algorithm, such as cluster analysis (such as K-means, DBSCAN) or anomaly detection algorithm (such as isolation forest, one-class support vector machine) is used to train the model to identify anomalies of the initial spatiotemporal data to obtain target anomaly data, and then obtain the corresponding target anomaly position based on the target anomaly data.

[0034] In some embodiments, the anomaly identification of the initial spatiotemporal data to obtain target anomaly data includes: determining a target window and a target step size, and performing data segmentation on the initial spatiotemporal data according to the target window and the target step size to obtain multiple window spatiotemporal data; performing feature extraction on the window spatiotemporal data to obtain a target representation vector corresponding to the window spatiotemporal data; performing distance calculation according to the target representation vector to obtain first adjacent spatiotemporal data corresponding to the window spatiotemporal data; calculating an outlier representation value corresponding to the window spatiotemporal data according to the first adjacent spatiotemporal data, and performing anomaly identification according to the outlier representation value to obtain the target anomaly data.

[0035] Exemplarily, according to the characteristics of the initial spatiotemporal data and business requirements, the target window is selected to ensure that the data in the window is rich enough to reflect the spatiotemporal characteristics, and the target step size of the target window is determined. The size of the target step size affects the granularity of data segmentation and the efficiency of calculation. The target step size can be the same as or different from the target window size. The minimum target step size is equal to 1, the maximum target step size is equal to the length of the target window, and the target step size is a positive integer.

[0036] Exemplarily, the sliding window technology is used to segment the initial spatiotemporal data according to the target window and the target step size to generate multiple window spatiotemporal data. Each window spatiotemporal data is a sub-dataset, which contains the initial spatiotemporal data corresponding to the form field under the target window.

[0037] Exemplarily, feature extraction is performed on each window spatiotemporal data using a feature extraction model to obtain a target feature vector corresponding to the window spatiotemporal data, and then a distance measurement method, such as Euclidean distance, Manhattan distance, cosine similarity, etc., is used to calculate the similarity between the window spatiotemporal data, so as to filter out the first adjacent spatiotemporal data corresponding to each window spatiotemporal data when the similarity is high. The first adjacent window spatiotemporal data refers to the window data that is closest to the current window spatiotemporal data in data distribution.

[0038] Exemplarily, various outlier detection methods (such as Z-score, local outlier factor LOF, isolated forest, etc.) are used to calculate the outlier characterization value of the first adjacent spatiotemporal data. The outlier characterization value reflects the degree of abnormality of the first adjacent spatiotemporal data relative to its surrounding data. Due to the high correlation between the first adjacent spatiotemporal data and the current window spatiotemporal data, it can be reasonably assumed that the outlier characterization value of the first adjacent spatiotemporal data can represent the degree of outliers of the current window spatiotemporal data to a certain extent. Therefore, the calculated outlier characterization value of the first adjacent spatiotemporal data is assigned to the current window spatiotemporal data as a quantitative indicator of its degree of outliers.

[0039] Exemplarily, a threshold of the outlier characterization value is set. Data points whose outlier characterization values ​​exceed the threshold are considered as target abnormal data, so that the window spatiotemporal data whose outlier characterization values ​​exceed the threshold are identified as target abnormal data.

[0040] In some embodiments, the calculating the outlier characterization value corresponding to the window spatiotemporal data according to the first adjacent spatiotemporal data includes: obtaining first distance information between the first adjacent spatiotemporal data and the window spatiotemporal data according to the target characterization vector; summing the first distance information to obtain a first sum value, and dividing the adjacent number corresponding to the first adjacent spatiotemporal data by the first sum value to obtain first density information corresponding to the window spatiotemporal data; obtaining second adjacent spatiotemporal data corresponding to the first adjacent spatiotemporal data, and obtaining second density information corresponding to the first adjacent spatiotemporal data according to the second adjacent spatiotemporal data; determining the outlier characterization value corresponding to the window spatiotemporal data according to the second density information and the first density information; wherein the outlier characterization value is obtained according to the following formula:

[0041]

[0042] in, represents the outlier characterization value corresponding to the i-th window spatiotemporal data, represents the second density information corresponding to the j-th first adjacent spatiotemporal data, represents the first density information corresponding to the i-th window spatiotemporal data, and num represents the adjacent number corresponding to the first adjacent spatiotemporal data.

[0043] Exemplarily, according to the target characterization vector, the first distance information between the window spatiotemporal data and the first adjacent spatiotemporal data is calculated, and then all the first distance information is summed to obtain a first sum value, so as to divide the adjacent number of the first adjacent spatiotemporal data by the first sum value to obtain the first density information of the window spatiotemporal data. The first density information reflects the compactness of the data around the window spatiotemporal data.

[0044] Exemplarily, for each first adjacent spatiotemporal data, its corresponding second adjacent spatiotemporal data (i.e., other data adjacent to the first adjacent spatiotemporal data) is found, so as to calculate the second density information of the first adjacent spatiotemporal data based on the second adjacent spatiotemporal data. This further describes the data distribution around the first adjacent spatiotemporal data.

[0045] Exemplarily, the outlier characterization value of the window spatiotemporal data is calculated using the following formula in combination with the first density information and the second density information. The outlier characterization value quantifies the degree of outlier of the window spatiotemporal data relative to its surrounding data:

[0046]

[0047] in, represents the outlier representation value corresponding to the spatiotemporal data of the i-th window, represents the second density information corresponding to the jth first adjacent spatiotemporal data, It represents the first density information corresponding to the i-th window spatiotemporal data, and num represents the adjacent number corresponding to the first adjacent spatiotemporal data.

[0048] Exemplarily, by considering the relationship between the window spatiotemporal data and the first adjacent spatiotemporal data, and the relationship between the first adjacent spatiotemporal data and the second adjacent spatiotemporal data, the local structure and distribution characteristics of the data can be captured more accurately, thereby improving the accuracy of anomaly identification. Using density information instead of a single distance threshold to judge outliers makes the method more robust to local changes in the data. Even in the case of uneven data density, abnormal data can be effectively identified.

[0049] In some embodiments, obtaining the corresponding target abnormal position according to the target abnormal data includes: determining the newly added window data corresponding to the target abnormal data under the target window and the target step size; and obtaining the corresponding target abnormal position from the initial spatiotemporal data according to the newly added window data.

[0050] Exemplarily, the target abnormal data is the window spatiotemporal data whose outlier characterization value exceeds a preset threshold. In order to further determine the specific abnormal data in the window spatiotemporal data, it is necessary to determine the corresponding window new data based on the data gradually added to the window spatiotemporal data under the target window and target step size.

[0051] For example, the size of the target window determines the amount of data analyzed each time, while the target step size controls the degree of overlap or the speed of movement between data windows. Assume that we have set the size of the target window to 10, which means that in each analysis cycle, windowed spatiotemporal data containing 10 data points will be processed. At the same time, the target step size is set to 2, which indicates that in consecutive analysis cycles, the window will move at a speed of 2 data points. Now, consider a specific scenario in which the windowed spatiotemporal data w1 is identified as target abnormal data. This means that the data points contained in w1 exhibit characteristics that are significantly different from the normal pattern, which may indicate an abnormal event or behavior. According to the setting of a target step size of 2, when transitioning from the current analysis cycle to the next analysis cycle, the windowed spatiotemporal data will be updated to include the latest two data and exclude the oldest two data points. In this case, the two data newly added to the windowed spatiotemporal data w1 can be determined as "new window data."

[0052] Exemplarily, a target abnormal position corresponding to the window's newly added data is obtained from the initial spatiotemporal data according to the window's newly added data.

[0053] Step S104: filling the target abnormal position with data according to the initial spatiotemporal data to obtain the target spatiotemporal data linked to the target form.

[0054] Exemplarily, a machine learning model such as a regression model or a time series model is constructed, and the model is trained using the data of the position preceding the target abnormal position, and the model is used to predict the value of the target abnormal position, thereby filling the predicted value into the target abnormal position, and then obtaining the target spatiotemporal data of the target form link.

[0055] In some embodiments, the method of filling data in the target abnormal position according to the initial spatiotemporal data to obtain the target spatiotemporal data linked to the target form includes: obtaining first sequence data corresponding to the left position in the target abnormal position and second sequence data corresponding to the right position in the target abnormal position from the initial spatiotemporal data; obtaining first similar data corresponding to the first sequence data from the initial spatiotemporal data and second similar data corresponding to the second sequence data from the initial spatiotemporal data; filling data in the target abnormal position according to the first similar data to obtain first filling data; filling data in the target abnormal position according to the second similar data to obtain second filling data; obtaining target filling data corresponding to the target abnormal position according to the first filling data and the second filling data; and determining the target spatiotemporal data linked to the target form according to the target filling data and the target abnormal position.

[0056] Exemplarily, data on the left side of the target abnormal position is extracted from the initial spatiotemporal data to form a first sequence of data, and data on the right side of the target abnormal position is extracted from the initial spatiotemporal data to form a second sequence of data.

[0057] Exemplarily, the data point closest to the first sequence data is found from the initial spatiotemporal data and determined as the first similar data, and the data point closest to the second sequence data is found from the initial spatiotemporal data and determined as the second similar data.

[0058] Exemplarily, the data corresponding to the next position in the first similar data is obtained from the initial spatiotemporal data, and the data corresponding to the next position in the first similar data is determined as the first filling data. The data corresponding to the previous position in the second similar data is obtained from the initial spatiotemporal data, and the data corresponding to the previous position in the second similar data is determined as the second filling data.

[0059] Exemplarily, based on the credibility or proximity of the first filling data and the second filling data, a first weight corresponding to the first filling data and a second weight corresponding to the second filling data are calculated, so that the first filling data and the second filling data are fused according to the first weight and the second weight to generate target filling data.

[0060] For example, multi-directional filling can more accurately fill in the target abnormal position and improve the integrity of spatiotemporal data. Using similar data on the left and right sides for filling can reduce the error that may be introduced by unidirectional filling and improve the accuracy of the data.

[0061] In some embodiments, obtaining the first similar data corresponding to the first sequence data from the initial spatiotemporal data and obtaining the second similar data corresponding to the second sequence data from the initial spatiotemporal data includes: obtaining the data length corresponding to the first sequence data, and performing data interception from the initial spatiotemporal data according to the data length to obtain the first intercepted data; obtaining the first distribution probability corresponding to the first sub-data at the first position in the first sequence data and the second distribution probability corresponding to the second sub-data at the first position in the first intercepted data; determining the first weight information corresponding to the first position according to the first distribution probability and the second distribution probability; obtaining the first abnormal position corresponding to the target abnormal data in the first intercepted data and obtaining the second abnormal position corresponding to the target abnormal data in the first sequence data; determining the first normal data amount corresponding to the first intercepted data and the first sequence data according to the first abnormal position and the second abnormal position; calculating the weight information and the first normal data amount between the first sequence data and the first intercepted data. the first similarity of the target abnormal data in the second intercepted data; determining the first similar data corresponding to the first sequence data according to the first similarity and the first intercepted data; obtaining the second intercepted data by intercepting data from the initial spatiotemporal data according to the data length; obtaining the third distribution probability corresponding to the third sub-data at the second position in the second sequence data and the fourth distribution probability corresponding to the fourth sub-data at the second position in the second intercepted data; determining the second weight information corresponding to the second position according to the third distribution probability and the fourth distribution probability; obtaining the third abnormal position corresponding to the target abnormal data in the second intercepted data and obtaining the fourth abnormal position corresponding to the target abnormal data in the second sequence data; determining the second normal data amount corresponding to the second intercepted data and the second sequence data according to the third abnormal position and the fourth abnormal position; calculating the second similarity between the second sequence data and the second intercepted data according to the second weight information and the second normal data amount; determining the second similarity corresponding to the second sequence data according to the second similarity and the second intercepted data.

[0062] Exemplarily, the length of the first sequence data, that is, the number of data points included in the first sequence data, is calculated to obtain the data length, and then the data length of the first sequence data is used to extract data of the data length from the initial spatiotemporal data to generate first intercepted data.

[0063] Exemplarily, the first distribution probability of the first sub-data corresponding to each first position in the first sequence data in the initial spatiotemporal data is counted. The first distribution probability can be represented by the frequency or probability density function of the data. The second distribution probability of the second sub-data corresponding to the first position in the first intercepted data in the initial spatiotemporal data is counted. Similarly, the second distribution probability can be represented by the frequency or probability density function of the data.

[0064] Exemplarily, determine whether the first sub-data at the first position is normal data and whether the second sub-data at the first position is normal data. If the first sub-data at the first position is normal data and the second sub-data at the first position is normal data, the first weight information at the first position is the product of the first distribution probability and the second distribution probability; if the first sub-data at the first position is abnormal data or the second sub-data at the first position is abnormal data, the first weight information at the first position is 0.

[0065] Exemplarily, the position of the target abnormal data is found in the first intercepted data according to the target abnormal data and marked as the first abnormal position, and the position of the target abnormal data is found in the first sequence data and marked as the second abnormal position.

[0066] Exemplarily, the first normal data amount is obtained by determining the number of normal data in the first intercepted data and the first sequence data according to the first abnormal position and the second abnormal position. The first normal data amount can be expressed as the number of normal data in the first intercepted data and the first sequence data.

[0067] Exemplarily, the first similarity between the first sequence data and the first intercepted data is calculated using the first weight information and the first normal data amount according to the following formula:

[0068]

[0069] in, represents the first similarity between the first sequence data and the i-th first intercepted data, length1 represents the data length corresponding to the first intercepted data, Indicates the first normal data volume, represents the first sub-data corresponding to the j-th first position in the first sequence data, represents the second sub-data corresponding to the j-th first position in the i-th first intercepted data, Represents first weight information between the first sub-data and the second sub-data at the j-th first position.

[0070] Exemplarily, the first similar data corresponding to the first sequence data is obtained by screening the first intercepted data according to the first similarity and the similarity threshold. The first similar data may be the data with the highest similarity to the first sequence data in the first intercepted data.

[0071] Exemplarily, the length of the second sequence data, i.e., the number of data points contained in the second sequence data, is calculated to obtain the data length, and then the data length of the second sequence data is used to extract data of the data length from the initial spatiotemporal data to generate second extracted data.

[0072] Exemplarily, the third distribution probability of the third sub-data corresponding to each second position in the second sequence data in the initial spatiotemporal data is counted. The third distribution probability can be represented by the frequency or probability density function of the data. The fourth distribution probability of the fourth sub-data corresponding to the second position in the second intercepted data in the initial spatiotemporal data is counted. Similarly, the fourth distribution probability can be represented by the frequency or probability density function of the data.

[0073] Exemplarily, determine whether the third sub-data at the second position is normal data and whether the fourth sub-data at the second position is normal data. If the third sub-data at the second position is normal data and the fourth sub-data at the second position is normal data, the second weight information at the second position is the product of the third distribution probability and the fourth distribution probability; if the third sub-data at the second position is abnormal data or the fourth sub-data at the second position is abnormal data, the second weight information at the second position is 0.

[0074] Exemplarily, the position of the target abnormal data is found in the second intercepted data according to the target abnormal data and marked as the third abnormal position, and the position of the target abnormal data is found in the second sequence data and marked as the fourth abnormal position.

[0075] Exemplarily, the number of the second intercepted data and the second sequence data when both are normal data is determined according to the third abnormal position and the fourth abnormal position, that is, the second normal data amount is obtained. The second normal data amount can be expressed as the number of the second intercepted data and the second sequence data when both are normal data.

[0076] Exemplarily, the second similarity between the second sequence data and the second intercepted data is calculated using the second weight information and the second normal data amount according to the following formula:

[0077]

[0078] in, represents the second similarity between the second sequence data and the t-th second intercepted data, length2 represents the data length corresponding to the second intercepted data, Indicates the second normal data volume, represents the third sub-data corresponding to the mth second position in the second sequence data, represents the fourth sub-data corresponding to the m-th second position in the t-th second intercepted data, Represents the second weight information between the third sub-data and the fourth sub-data at the m-th second position.

[0079] Exemplarily, the second similar data corresponding to the second sequence data is obtained by screening the second intercepted data according to the second similarity and the similarity threshold. The second similar data may be the data with the highest similarity to the second sequence data in the second intercepted data.

[0080] Specifically, in the similarity calculation, the weight is adjusted according to the distribution probability of the data point. Data points with high distribution probability have higher weights in the similarity calculation because they better represent the actual distribution of the data. Thus, the similarity calculation is performed by combining the missing data and the data distribution information. A higher weight is given to the non-missing data, a lower weight is given to the missing data, and the weight is adjusted according to the distribution probability of the data point to generate the similarity. Then, by comprehensively considering the missing data and the data distribution information, the calculation accuracy of the similar data can be significantly improved. The filling and weight adjustment of the missing data reduce the impact of the missing values ​​on the similarity calculation, while the consideration of the distribution probability more accurately reflects the actual distribution of the data. Thus, the missing data and the data distribution information can be effectively considered when calculating the similar data, thereby improving the accuracy and reliability of the calculation, and providing strong support for the subsequent data analysis and application.

[0081] In some embodiments, data filling is performed on the target abnormal position according to the first similar data to obtain first filling data, and data filling is performed on the target abnormal position according to the second similar data to obtain second filling data, including: performing data fitting on the first similar data and the first sequence data to obtain a first fitting model corresponding to the first sequence data; performing data prediction on the target abnormal position according to the first fitting model to obtain the first filling data; performing data fitting on the second similar data and the second sequence data to obtain a second fitting model corresponding to the second sequence data; performing data prediction on the target abnormal position according to the second fitting model to obtain the second filling data.

[0082] Exemplarily, after data fusion of the first similar data and the first sequence data, data fitting is performed on the fused data to construct a first fitting model corresponding to the first sequence data, so as to use the first fitting model to perform data prediction on the target abnormal position to obtain the first filling data. For example, the timestamp or position information of the target abnormal position is input into the first fitting model to obtain the first filling data.

[0083] Exemplarily, after data fusion of the second similar data and the second sequence data, data fitting is performed on the fused data to construct a second fitting model corresponding to the second sequence data, so as to use the second fitting model to perform data prediction on the target abnormal position to obtain the second filling data. For example, the timestamp or position information of the target abnormal position is input into the second fitting model to obtain the second filling data.

[0084] Specifically, by fitting and predicting the sequence data from the left and right sides, different filling results can be obtained, providing more references for the final filling decision. The first filling data and the second filling data of the target abnormal position are effectively obtained from the initial spatiotemporal data, thus providing a solid foundation for subsequent data integration and analysis.

[0085] Step S105: determine the corresponding initial twin model according to the form field, and construct the initial twin model according to the target spatiotemporal data to obtain the target twin model.

[0086] Exemplarily, the form fields are combined to obtain the corresponding initial twin model, and then the target spatiotemporal data is used to construct the initial twin model to obtain the target twin model.

[0087] In some embodiments, determining the corresponding initial twin model based on the form fields includes: determining the associated fields required for the preset twin model; analyzing the associated fields according to principal component analysis to obtain the first field and the second field corresponding to the preset twin model, wherein the first field is used to characterize the necessary fields corresponding to the preset twin model, and the second field is used to characterize the fields required for the preset twin model when performing model optimization on the first field; establishing a model mapping table based on the first field, the second field and the preset twin model; and determining the initial twin model based on the form fields and the model mapping table.

[0088] For example, the key fields required for the preset twin model are clearly defined. The key fields are the basis for building the twin model. According to the application scenarios and goals of the preset twin model, principal component analysis is used to analyze the associated fields to determine which fields are necessary and which fields are optional.

[0089] Exemplarily, principal component analysis is used to reduce the dimension and extract features of the associated fields to identify which fields contribute most to the performance of the model, thereby obtaining the first field and the second field. The first field represents the necessary fields corresponding to the preset twin model. The first field is the core of the operation of the preset twin model. The second field represents the fields required for the preset twin model to optimize the model on the first field. The second field may be an auxiliary feature used to further optimize the performance of the model.

[0090] Exemplarily, a model mapping table is established according to the first field, the second field and the preset twin model. Thus, the form field is queried in the model mapping table to determine the initial twin model. The initial twin model is a preliminary model constructed based on the form field and the model mapping table.

[0091] Exemplarily, when the first fields corresponding to the preset twin model all exist in the form field, the preset twin model is determined as the initial twin model; when the first fields corresponding to the preset twin model do not all exist in the form field, the preset twin model cannot be determined as the initial twin model; when the first fields corresponding to the preset twin model do not all exist in the form field but the second fields all exist in the form field, the preset twin model cannot be determined as the initial twin model.

[0092] In some embodiments, the method also includes: determining the remaining fields in the form fields that fail to form the initial twin model according to the model mapping table; obtaining the relevant fields corresponding to the remaining fields according to the model mapping table, generating corresponding target reminder information according to the relevant fields, and displaying the target reminder information to the target user; obtaining the target interaction behavior performed by the target user under the target reminder information; modeling the remaining fields according to the target interaction behavior to obtain the model building results corresponding to the remaining fields.

[0093] Exemplarily, according to the model mapping table, the remaining fields in the form fields that fail to form the initial twin model are determined. The remaining fields are because there is no corresponding mapping relationship in the model mapping table, or the fields are missing, so that the corresponding preset twin model cannot be found.

[0094] Exemplarily, the relevant fields corresponding to the remaining fields are determined according to the model mapping table. The relevant fields are fields that can find the corresponding preset twin model in the model mapping table after being combined with the remaining fields, so as to generate corresponding target reminder information based on the relevant fields. The target reminder information can prompt the target user that the corresponding twin model cannot be built according to the remaining fields because the relevant fields have not been selected, so as to suggest the target user to add the relevant fields, and then display these target reminder information to the target user so that the user can understand the status of the current model construction and the fields that may need to be processed.

[0095] Exemplarily, the target interaction behavior of the target user under the target reminder information is obtained. The target interaction behavior can be operations such as clicking, selecting, editing, etc. These behaviors will be used to further process the remaining fields.

[0096] Exemplarily, the remaining fields are modeled according to the target interaction behavior of the target user to obtain the model building results corresponding to the remaining fields. For example, if the target user chooses to map a remaining field to a related field in the model, then the related field and the remaining field can obtain a corresponding twin model, and then the twin model can be incorporated into the model building process to obtain the model building results corresponding to the remaining fields.

[0097] Exemplarily, if no relevant fields are chosen to be added in the target interactive behavior, data analysis will be performed based on the target spatiotemporal data corresponding to the remaining fields, so that the data change information of the target spatiotemporal data corresponding to the remaining fields is determined as the model building results corresponding to the remaining fields.

[0098] See also Figure 2 , Figure 2 A system 200 for constructing a spatiotemporal data-associated digital twin model based on a dynamic form is provided in an embodiment of the present application. The system 200 for constructing a spatiotemporal data-associated digital twin model based on a dynamic form includes a data acquisition module 201, a data generation module 202, an anomaly recognition module 203, a data filling module 204, and a model construction module 205, wherein the data acquisition module 201 is used to respond to the interactive behavior of the target user on the interface and obtain the form fields adjusted by the target user according to the interactive behavior; the data generation module 202 is used to generate a corresponding target form according to the form fields, And obtain the initial spatiotemporal data of the target form link from the target database; an exception identification module 203 is used to perform exception identification on the initial spatiotemporal data to obtain target exception data, and obtain the corresponding target exception position according to the target exception data; a data filling module 204 is used to fill the target exception position with data according to the initial spatiotemporal data to obtain the target spatiotemporal data of the target form link; a model construction module 205 is used to determine the corresponding initial twin model according to the form field, and construct the initial twin model according to the target spatiotemporal data to obtain the target twin model.

[0099] In some implementations, the anomaly identification module 203, during the process of performing anomaly identification on the initial spatiotemporal data to obtain target anomaly data, performs:

[0100] Determine a target window and a target step size, and perform data segmentation on the initial spatiotemporal data according to the target window and the target step size to obtain a plurality of window spatiotemporal data;

[0101] Performing feature extraction on the window spatiotemporal data to obtain a target representation vector corresponding to the window spatiotemporal data;

[0102] Performing distance calculation according to the target characterization vector to obtain first adjacent spatiotemporal data corresponding to the window spatiotemporal data;

[0103] An outlier characterization value corresponding to the window spatiotemporal data is calculated according to the first adjacent spatiotemporal data, and anomaly identification is performed according to the outlier characterization value to obtain the target anomaly data.

[0104] In some implementations, the anomaly identification module 203, during the process of calculating the outlier characterization value corresponding to the window spatiotemporal data according to the first adjacent spatiotemporal data, performs:

[0105] Obtaining first distance information between the first adjacent spatiotemporal data and the window spatiotemporal data according to the target representation vector;

[0106] Summing the first distance information to obtain a first sum value, and dividing the adjacent number corresponding to the first adjacent spatiotemporal data by the first sum value to obtain first density information corresponding to the window spatiotemporal data;

[0107] Obtaining second adjacent spatiotemporal data corresponding to the first adjacent spatiotemporal data, and obtaining second density information corresponding to the first adjacent spatiotemporal data according to the second adjacent spatiotemporal data;

[0108] Determine the outlier characterization value corresponding to the window spatiotemporal data according to the second density information and the first density information;

[0109] The outlier characterization value is obtained according to the following formula:

[0110]

[0111] in, represents the outlier characterization value corresponding to the i-th window spatiotemporal data, represents the second density information corresponding to the j-th first adjacent spatiotemporal data, represents the first density information corresponding to the i-th window spatiotemporal data, and num represents the adjacent number corresponding to the first adjacent spatiotemporal data.

[0112] In some implementations, the anomaly identification module 203, in the process of obtaining the corresponding target anomaly position according to the target anomaly data, performs:

[0113] Determine the newly added data in the window corresponding to the target abnormal data under the target window and the target step size;

[0114] The corresponding target abnormal position is obtained from the initial spatiotemporal data according to the newly added data in the window.

[0115] In some implementations, the data filling module 204 performs, during the process of filling the target abnormal position with data according to the initial spatiotemporal data to obtain the target spatiotemporal data linked to the target form:

[0116] Obtaining, from the initial spatiotemporal data, a first sequence of data corresponding to a left position in the target abnormal position and a second sequence of data corresponding to a right position in the target abnormal position;

[0117] Obtaining first similar data corresponding to the first sequence data from the initial spatiotemporal data and obtaining second similar data corresponding to the second sequence data from the initial spatiotemporal data;

[0118] Filling the target abnormal position with data according to the first similar data to obtain first filling data;

[0119] Filling the target abnormal position with data according to the second similar data to obtain second filling data;

[0120] Obtaining target filling data corresponding to the target abnormal position according to the first filling data and the second filling data;

[0121] The target spatiotemporal data linked to the target form is determined according to the target filling data and the target abnormal position.

[0122] In some implementations, the data filling module 204, in the process of obtaining the first similar data corresponding to the first sequence data from the initial spatiotemporal data and obtaining the second similar data corresponding to the second sequence data from the initial spatiotemporal data, performs:

[0123] Obtaining a data length corresponding to the first sequence data, and performing data interception from the initial spatiotemporal data according to the data length to obtain first intercepted data;

[0124] Obtaining a first distribution probability corresponding to the first sub-data at the first position in the first sequence data and a second distribution probability corresponding to the second sub-data at the first position in the first intercepted data;

[0125] Determine first weight information corresponding to the first position according to the first distribution probability and the second distribution probability;

[0126] Obtaining a first abnormal position corresponding to the target abnormal data in the first intercepted data and obtaining a second abnormal position corresponding to the target abnormal data in the first sequence data;

[0127] Determine a first normal data amount corresponding to the first intercepted data and the first sequence data according to the first abnormal position and the second abnormal position;

[0128] Calculate a first similarity between the first sequence data and the first intercepted data according to the first weight information and the first normal data amount;

[0129] Determine the first similar data corresponding to the first sequence data according to the first similarity and the first intercepted data;

[0130] According to the data length, data is intercepted from the initial spatiotemporal data to obtain second intercepted data;

[0131] Obtain a third distribution probability corresponding to the third sub-data at the second position in the second sequence data and a fourth distribution probability corresponding to the fourth sub-data at the second position in the second intercepted data;

[0132] Determine second weight information corresponding to the second position according to the third distribution probability and the fourth distribution probability;

[0133] Obtaining a third abnormal position corresponding to the target abnormal data in the second intercepted data and obtaining a fourth abnormal position corresponding to the target abnormal data in the second sequence data;

[0134] Determine a second normal data amount corresponding to the second intercepted data and the second sequence data according to the third abnormal position and the fourth abnormal position;

[0135] Calculate a second similarity between the second sequence data and the second intercepted data according to the second weight information and the second normal data amount;

[0136] The second similar data corresponding to the second sequence data is determined according to the second similarity and the second intercepted data.

[0137] In some implementations, the data filling module 204 performs, during the process of filling the target abnormal position with data according to the first similar data to obtain the first filling data and filling the target abnormal position with data according to the second similar data to obtain the second filling data:

[0138] Performing data fitting according to the first similar data and the first sequence data to obtain a first fitting model corresponding to the first sequence data;

[0139] Performing data prediction on the target abnormal position according to the first fitting model to obtain the first filling data;

[0140] Performing data fitting according to the second similar data and the second sequence data to obtain a second fitting model corresponding to the second sequence data;

[0141] The second filling data is obtained by performing data prediction on the target abnormal position according to the second fitting model.

[0142] In some implementations, the model building module 205 performs, during the process of determining the corresponding initial twin model according to the form fields:

[0143] Determine the associated fields required for the preset twin model;

[0144] Analyzing the associated fields according to principal component analysis to obtain a first field and a second field corresponding to the preset twin model, wherein the first field is used to characterize a necessary field corresponding to the preset twin model, and the second field is used to characterize a field required for the preset twin model to perform model optimization on the first field;

[0145] Establishing a model mapping table according to the first field, the second field and the preset twin model;

[0146] The initial twin model is determined according to the form fields and the model mapping table.

[0147] In some implementations, the model building module 205 also performs:

[0148] Determine, according to the model mapping table, the remaining fields in the form fields that fail to form the initial twin model;

[0149] Obtaining relevant fields corresponding to the remaining fields according to the model mapping table, generating corresponding target reminder information according to the relevant fields, and displaying the target reminder information to the target user;

[0150] Obtaining a target interactive behavior performed by the target user under the target reminder information;

[0151] Model building is performed on the remaining fields according to the target interaction behavior to obtain model building results corresponding to the remaining fields.

[0152] In some embodiments, the system 200 for constructing a spatiotemporal data-associated digital twin model based on a dynamic form can be applied to a terminal device.

[0153] It should be noted that technical personnel in the relevant field can clearly understand that, for the convenience and conciseness of description, the specific working process of the system 200 for constructing a digital twin model with spatiotemporal data association based on a dynamic form described above can refer to the corresponding process in the aforementioned embodiment of the method for constructing a digital twin model with spatiotemporal data association based on a dynamic form, and will not be repeated here.

[0154] An embodiment of the present invention also provides a storage medium for computer-readable storage, wherein the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement any step of a method for constructing a spatiotemporal data associated digital twin model based on a dynamic form as provided in the description of the embodiment of the present invention.

[0155] The storage medium may be an internal storage unit of the terminal device in the aforementioned embodiment, such as a hard disk or memory of the terminal device. The storage medium may also be an external storage device of the terminal device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the terminal device.

[0156] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware embodiment, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0157] It should be understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations. It should be noted that, in this article, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "including a..." does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0158] The serial numbers of the embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for constructing a spatiotemporal data-related digital twin model based on a dynamic form, characterized in that: The method comprises: Responding to the target user's interactive behavior on the interface, and obtaining the form fields adjusted by the target user according to the interactive behavior; Generate a corresponding target form according to the form fields, and obtain the initial spatiotemporal data linked to the target form from a target database; Determine a target window and a target step size, and perform data segmentation on the initial spatiotemporal data according to the target window and the target step size to obtain multiple window spatiotemporal data; perform feature extraction on the window spatiotemporal data to obtain a target characterization vector corresponding to the window spatiotemporal data; perform distance calculation according to the target characterization vector to obtain first adjacent spatiotemporal data corresponding to the window spatiotemporal data; calculate an outlier characterization value corresponding to the window spatiotemporal data according to the first adjacent spatiotemporal data, perform anomaly identification according to the outlier characterization value to obtain target anomaly data, and obtain a corresponding target anomaly position according to the target anomaly data; Obtaining first sequence data corresponding to the left position of the target abnormal position and second sequence data corresponding to the right position of the target abnormal position from the initial spatiotemporal data; obtaining first adjacent data corresponding to the first sequence data from the initial spatiotemporal data and obtaining second adjacent data corresponding to the second sequence data from the initial spatiotemporal data; performing data filling on the target abnormal position according to the first adjacent data to obtain first filling data; performing data filling on the target abnormal position according to the second adjacent data to obtain second filling data; obtaining target filling data corresponding to the target abnormal position according to the first filling data and the second filling data; determining target spatiotemporal data linked to the target form according to the target filling data and the target abnormal position; The corresponding initial twin model is determined according to the form fields, and the initial twin model is constructed according to the target spatiotemporal data to obtain the target twin model.

2. The method according to claim 1, characterized in that The calculating the outlier characterization value corresponding to the window spatiotemporal data according to the first adjacent spatiotemporal data includes: Obtaining first distance information between the first adjacent spatiotemporal data and the window spatiotemporal data according to the target representation vector; Summing the first distance information to obtain a first sum value, and dividing the adjacent number corresponding to the first adjacent spatiotemporal data by the first sum value to obtain first density information corresponding to the window spatiotemporal data; Obtaining second adjacent spatiotemporal data corresponding to the first adjacent spatiotemporal data, and obtaining second density information corresponding to the first adjacent spatiotemporal data according to the second adjacent spatiotemporal data; Determine the outlier characterization value corresponding to the window spatiotemporal data according to the second density information and the first density information; The outlier characterization value is obtained according to the following formula: ； in, represents the outlier characterization value corresponding to the i-th window spatiotemporal data, represents the second density information corresponding to the j-th first adjacent spatiotemporal data, represents the first density information corresponding to the i-th window spatiotemporal data, and num represents the adjacent number corresponding to the first adjacent spatiotemporal data.

3. The method according to claim 1, characterized in that The obtaining a corresponding target abnormal position according to the target abnormal data includes: Determine the newly added data in the window corresponding to the target abnormal data under the target window and the target step size; The corresponding target abnormal position is obtained from the initial spatiotemporal data according to the newly added data in the window.

4. The method according to claim 1, characterized in that: The obtaining of first similar data corresponding to the first sequence data from the initial spatiotemporal data and obtaining of second similar data corresponding to the second sequence data from the initial spatiotemporal data includes: Obtaining a data length corresponding to the first sequence data, and performing data interception from the initial spatiotemporal data according to the data length to obtain first intercepted data; Obtaining a first distribution probability corresponding to the first sub-data at the first position in the first sequence data and a second distribution probability corresponding to the second sub-data at the first position in the first intercepted data; Determine first weight information corresponding to the first position according to the first distribution probability and the second distribution probability; Obtaining a first abnormal position corresponding to the target abnormal data in the first intercepted data and obtaining a second abnormal position corresponding to the target abnormal data in the first sequence data; Determine a first normal data amount corresponding to the first intercepted data and the first sequence data according to the first abnormal position and the second abnormal position; Calculate a first similarity between the first sequence data and the first intercepted data according to the first weight information and the first normal data amount; Determine the first similar data corresponding to the first sequence data according to the first similarity and the first intercepted data; According to the data length, data is intercepted from the initial spatiotemporal data to obtain second intercepted data; Obtain a third distribution probability corresponding to the third sub-data at the second position in the second sequence data and a fourth distribution probability corresponding to the fourth sub-data at the second position in the second intercepted data; Determine second weight information corresponding to the second position according to the third distribution probability and the fourth distribution probability; Obtaining a third abnormal position corresponding to the target abnormal data in the second intercepted data and obtaining a fourth abnormal position corresponding to the target abnormal data in the second sequence data; Determine a second normal data amount corresponding to the second intercepted data and the second sequence data according to the third abnormal position and the fourth abnormal position; Calculating a second similarity between the second sequence data and the second intercepted data according to the second weight information and the second normal data amount; The second similar data corresponding to the second sequence data is determined according to the second similarity and the second intercepted data.

5. The method according to claim 4, characterized in that Filling the target abnormal position with data according to the first similar data to obtain first filling data and filling the target abnormal position with data according to the second similar data to obtain second filling data, including: Performing data fitting according to the first similar data and the first sequence data to obtain a first fitting model corresponding to the first sequence data; Performing data prediction on the target abnormal position according to the first fitting model to obtain the first filling data; Performing data fitting according to the second similar data and the second sequence data to obtain a second fitting model corresponding to the second sequence data; The second filling data is obtained by performing data prediction on the target abnormal position according to the second fitting model.

6. The method according to claim 1, characterized in that The determining the corresponding initial twin model according to the form field includes: Determine the associated fields required for the preset twin model; Analyzing the associated fields according to principal component analysis to obtain a first field and a second field corresponding to the preset twin model, wherein the first field is used to characterize a necessary field corresponding to the preset twin model, and the second field is used to characterize a field required for the preset twin model to perform model optimization on the first field; Establishing a model mapping table according to the first field, the second field and the preset twin model; The initial twin model is determined according to the form fields and the model mapping table.

7. The method according to claim 6, characterized in that The method further comprises: Determine, according to the model mapping table, the remaining fields in the form fields that fail to form the initial twin model; Obtaining relevant fields corresponding to the remaining fields according to the model mapping table, generating corresponding target reminder information according to the relevant fields, and displaying the target reminder information to the target user; Obtaining a target interactive behavior performed by the target user under the target reminder information; Models are constructed for the remaining fields according to the target interaction behavior to obtain model construction results corresponding to the remaining fields.

8. A system for constructing a spatiotemporal data-related digital twin model based on a dynamic form, characterized in that: include: A data acquisition module, used to respond to the target user's interactive behavior on the interface and obtain the form fields adjusted by the target user according to the interactive behavior; A data generation module, used to generate a corresponding target form according to the form fields, and obtain the initial spatiotemporal data linked to the target form from a target database; An abnormality identification module, used to determine a target window and a target step size, and perform data segmentation on the initial spatiotemporal data according to the target window and the target step size to obtain a plurality of window spatiotemporal data; Performing feature extraction on the window spatiotemporal data to obtain a target characterization vector corresponding to the window spatiotemporal data; performing distance calculation based on the target characterization vector to obtain first adjacent spatiotemporal data corresponding to the window spatiotemporal data; Calculating an outlier characterization value corresponding to the window spatiotemporal data according to the first adjacent spatiotemporal data, performing anomaly identification according to the outlier characterization value to obtain target anomaly data, and obtaining a corresponding target anomaly position according to the target anomaly data; A data filling module, used to obtain a first sequence of data corresponding to a left position in the target abnormal position and a second sequence of data corresponding to a right position in the target abnormal position from the initial spatiotemporal data; Obtaining first similar data corresponding to the first sequence data from the initial spatiotemporal data and obtaining second similar data corresponding to the second sequence data from the initial spatiotemporal data; performing data filling on the target abnormal position according to the first similar data to obtain first filling data; performing data filling on the target abnormal position according to the second similar data to obtain second filling data; Obtain target filling data corresponding to the target abnormal position according to the first filling data and the second filling data; determine target spatiotemporal data linked to the target form according to the target filling data and the target abnormal position; A model building module is used to determine the corresponding initial twin model according to the form fields, and to build the initial twin model according to the target spatiotemporal data to obtain the target twin model.

Images