Smart city data processing method based on Beidou positioning

Abstract

The invention discloses a smart city data processing method based on Beidou positioning, which relates to the technical field of electric digital data processing, and comprises the following steps: collecting data from a Beidou positioning module and an auxiliary sensor, and preprocessing the collected data; and performing event detection on the preprocessed data, and executing intelligent response operation according to an event detection result. And performing data integrity verification and privacy protection based on a cryptographic algorithm on the data. A data processing task is divided into a plurality of sub-tasks, parallel processing is carried out on the sub-tasks at the same time on a multi-core processor or a distributed computing platform through a multi-thread scheduling algorithm, and processing resources are distributed to the sub-tasks according to a dynamic scheduling strategy. And performing time sequence analysis on the historical data and the real-time data, and constructing a prediction decision model. According to the method, on the basis of a traditional data processing technology, comprehensive analysis and prediction of various sensor data of a city are realized by introducing multi-model hybrid prediction, meta-model dynamic weight calculation and residual correction.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic digital data processing, and in particular, to a method for processing smart city data based on Beidou positioning. Background Art

[0002] In recent years, with the continuous advancement of smart city construction, various sensors, positioning systems (such as Beidou positioning system), and Internet of Things devices have been widely deployed, and urban operation data has shown explosive growth. Existing data processing methods mainly rely on static task scheduling and prediction of a single model, and have limitations in processing multi-source and real-time data, making it difficult to meet the requirements for data processing accuracy and response speed in a dynamic environment. In addition, traditional prediction models often lack an adaptive adjustment mechanism when facing complex scenarios such as urban traffic and environmental monitoring, and it is difficult to accurately estimate future data traffic and processing requirements. Summary of the Invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the present invention proposes a method for processing smart city data with closed-loop feedback management. This method constructs a hybrid prediction model by combining the ARIMA model and the long short-term memory network (LSTM), and introduces a dynamic weight calculation and residual correction module based on a meta-model to achieve comprehensive modeling of linear and non-linear features. At the same time, the system automatically adjusts data processing parameters (such as sampling frequency, filtering threshold) and task scheduling strategies (such as task allocation ratio and priority) according to the prediction results, and records the adjustment information and actual data feedback for periodically updating the prediction model parameters. Through this adaptive feedback mechanism, technical problems such as static models, fixed scheduling, and prediction failure in traditional methods are solved.

[0005] To solve the above technical problems, the present invention provides the following technical solution, a method for processing smart city data based on Beidou positioning, including: Collect data from the Beidou positioning module and auxiliary sensors, and preprocess the collected data; perform event detection on the preprocessed data, and execute intelligent response operations according to the event detection results; perform data integrity verification and privacy protection on the data based on cryptographic algorithms; divide the data processing tasks into multiple subtasks, and use a multi-threaded scheduling algorithm to simultaneously process each subtask in parallel on a multi-core processor or a distributed computing platform, and allocate processing resources to each subtask according to a dynamic scheduling strategy; perform time series analysis on historical data and real-time data, and construct a prediction decision model.

[0006] As a preferred solution of the smart city data processing method based on Beidou positioning according to the present invention, wherein: the preprocessing of the collected data includes noise filtering, data format conversion and data standardization processing of the collected data.

[0007] As a preferred solution of the smart city data processing method based on Beidou positioning according to the present invention, wherein: the event detection includes dimensionality reduction representation and reconstruction of the input data, extraction of features and calculation of the reconstruction error. For the input data it is processed using an autoencoder model as follows: ; wherein, represents the data vector at time , d is the dimension, is the encoder function, obtaining the low-dimensional feature representation ; using the decoder function to reconstruct the data: ; is the data reconstructed by the autoencoder; to quantify the reconstruction effect, the Euclidean distance between the input data and the reconstructed data is calculated as the reconstruction error: ; On the basis of obtaining the reconstruction error , key indicators in the input data are further extracted for preliminary anomaly judgment. A fixed threshold is set, and the key indicator is extracted from , and the judgment is made through an indicator function: ; wherein, when is greater than , the value of the indicator function is 1, otherwise it is 0; In order to further enhance the attention to local data features, an attention mechanism is introduced on the basis of the low-dimensional representation extracted by the autoencoder, and the attention score calculation formula is as follows: ; wherein, is the attention score at time t, is the weight matrix of the attention layer, is the bias vector, is the activation function, and is the attention vector, mapping the activation result to a scalar score; Normalize the scores at all times and calculate the attention weights: ; Among them, represents the total number of times considered within the current window; is the normalized attention weight at time t, obtained by exponentiating and normalizing the scores at all times.

[0008] Combine the fixed threshold judgment and the reconstruction error after attention weighting, and fuse the two parts of information in the form of a linear combination to obtain a composite anomaly score. The specific formula is as follows: ; Among them, is the result of the fixed threshold judgment, 0 or 1; is the composite anomaly score; is the weight coefficient, with a value within ; According to the composite anomaly score and the preset anomaly judgment threshold compare to judge whether the data is abnormal: ; When exceeds , it is determined that the data point is abnormal.

[0009] As a preferred solution of the smart city data processing method based on Beidou positioning described in the present invention, wherein: the intelligent response operation includes switching the data acquisition mode when detecting abnormal data, starting the data correction operation, performing resampling and redundancy verification on the data, and recording the operation process at the same time.

[0010] As a preferred solution of the smart city data processing method based on Beidou positioning described in the present invention, wherein: the data integrity verification and privacy protection based on cryptographic algorithms include, to prevent the same data from generating the same digest repeatedly and resist pre-computation attacks, concatenate the input data with a random salt value, generate a data digest through a secure hash function, sign the data digest using a quantum-secure digital signature, verify the digital signature using the public key during data transmission and subsequent processing, and encrypt the fields containing sensitive information using desensitization processing and a symmetric encryption algorithm.

[0011] As a preferred solution of the smart city data processing method based on Beidou positioning described in the present invention, wherein: the parallel processing includes dividing the preprocessed data into multiple subtasks according to the data type and processing requirements, and using a multi-threaded scheduling algorithm to perform parallel processing on each subtask simultaneously on a multi-core processor or a distributed computing platform.

[0012] As a preferred solution of the smart city data processing method based on Beidou positioning according to the present invention, wherein: the dynamic scheduling strategy includes real-time monitoring of the load status and response time of each processing unit. When the load status of a certain processing unit exceeds a predetermined load threshold, the subtasks in this processing unit are reassigned to other processing units whose load status does not exceed the predetermined load threshold, and the status of the reassigned subtasks is monitored.

[0013] As a preferred solution of the smart city data processing method based on Beidou positioning according to the present invention, wherein: the construction of the prediction decision model includes linear modeling of the standardized time series data, using the ARIMA model. The ARIMA model expression is: ; Wherein, represents the standardized observed value at time ; represents the mean of the data series, indicating the average level of the long-term trend; represents the lag operator, defined as , and is used to represent the data at the past time; is the autoregressive polynomial, where are the autoregressive coefficients, is the autoregressive order; represents the moving average polynomial, are the moving average coefficients, is the moving average order; represents the white noise sequence, representing the random error; The ARIMA model captures the linear trend and periodic changes in the time series, generating a linear prediction result based on historical data, denoted as ; To capture the non-linear features in the data, a long short-term memory network is used for modeling. The LSTM model updates the hidden state and output prediction in the following form: ; Wherein, represents the LSTM hidden state at time , representing the feature vector memorized by the model; represents the LSTM cell, which receives the current input and the previous hidden state as inputs, and outputs the current hidden state ; represents the output layer weight matrix, which is used to map the hidden state to the prediction space; represents the output layer bias vector; Denote the predicted value generated by the LSTM model at time ; To adaptively adjust the contributions of the ARIMA and LSTM models in the hybrid prediction, a meta-model is constructed to calculate the dynamic weights. First, construct the feature vector: ; where the prediction error and the change in error fluctuation are defined as follows: ; where is the standard deviation of the prediction error within a time window at time ; Calculate the dynamic weights through the meta-model: ; where, represents the input feature vector of the meta-model; represents the absolute value of the ARIMA prediction error; represents the absolute value of the LSTM prediction error; represents the true observed value; represents the ARIMA predicted value; represents the LSTM predicted value; represents the change in the standard deviation of the prediction error; represents the weight matrix of the meta-model, used to perform weighted summation on the features; represents the bias of the meta-model; represents the Sigmoid activation function, mapping the output to the interval; Denote the dynamic weight calculated at time , representing the contribution ratio of the ARIMA prediction result in the hybrid prediction; Fuse the two prediction results by weighting to obtain the preliminary predicted value: ; where, represents the preliminary predicted value of the hybrid prediction model; To further correct the systematic bias in the preliminary prediction, calculate the prediction residuals and use a feed-forward neural network to correct them. Define the prediction residuals as: ; Construct a feed-forward neural network using the historical residual sequence to obtain the correction term: ; where, represents the prediction residual at time , that is, the difference between the actual observed value and the preliminary predicted value; represents a feedforward neural network function for learning the mapping relationship between historical residuals and current prediction errors; represents the residual sequence at time as the input to the network; represents the number of time steps of historical residuals, used to capture short-term residual characteristics; represents the residual correction term, output by the neural network, used to compensate the preliminary prediction; ; where represents at time the final predicted value, the output after hybrid prediction and residual correction.

[0014] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the smart city data processing method based on Beidou positioning.

[0015] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps of the smart city data processing method based on Beidou positioning.

[0016] Advantages of the present invention: Based on traditional data processing technologies, the present invention realizes the comprehensive analysis and prediction of various types of urban sensor data by introducing multi-model hybrid prediction, meta-model dynamic weight calculation, and residual correction modules. At the same time, the closed-loop feedback management mechanism based on prediction errors enables the system to automatically adjust data processing parameters and task scheduling strategies during operation, and uses feedback records to periodically update the prediction model, thereby improving the self-adaptability of the system to data processing, scheduling, and prediction. That is, on the basis of realizing the interaction of multi-level information, this closed-loop feedback mechanism ensures that the model and system parameters can always be synchronized with the real-time data state, meeting the actual needs of complex and dynamic data processing in the urban environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0018] Figure 1 FIG. is a schematic flowchart of a smart city data processing method based on Beidou positioning provided by an embodiment of the present invention. Detailed Embodiments

[0019] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following provides a detailed description of the specific embodiments of the present invention in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all 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 scope of protection of the present invention.

[0020] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0021] Secondly, the so-called "one embodiment" or "embodiment" herein refers to specific features, structures, or characteristics that may be included in at least one implementation manner of the present invention. The phrase "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that mutually excludes other embodiments.

[0022] The present invention is described in detail in conjunction with schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views showing the device structure will be enlarged locally in a non-general proportion, and the schematic diagrams are only examples and should not limit the scope of protection of the present invention herein. In addition, in actual production, three-dimensional spatial dimensions including length, width, and depth should be included.

[0023] At the same time, in the description of the present invention, it should be noted that the orientation or positional relationships indicated by terms such as "upper, lower, inner, and outer" are based on the orientation or positional relationships shown in the drawings. This is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention. In addition, the terms "first, second, or third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0024] Unless otherwise clearly defined and limited in the present invention, the terms "mounted, connected, and coupled" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can also be a mechanical connection, an electrical connection, or a direct connection, and can also be indirectly connected through an intermediate medium, or it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0025] Embodiment 1, referring to Figure 1, which is the first embodiment of the present invention. This embodiment provides a method for processing smart city data based on Beidou positioning, including: S1: Collect data from the Beidou positioning module and auxiliary sensors, and preprocess the collected data.

[0026] S2: Perform event detection on the preprocessed data, and execute intelligent response operations according to the event detection results.

[0027] S3: Implement data integrity verification and privacy protection for the data based on cryptographic algorithms.

[0028] S4: Divide the data processing tasks into multiple subtasks, and use multi-threaded scheduling algorithms to perform parallel processing on each subtask simultaneously on a multi-core processor or distributed computing platform, and allocate processing resources to each subtask according to dynamic scheduling strategies.

[0029] S5: Perform time series analysis on historical data and real-time data, and build a prediction and decision-making model.

[0030] The preprocessing of the collected data includes noise filtering, data format conversion, and data standardization processing of the collected data.

[0031] The event detection includes dimensionality reduction representation and reconstruction of the input data, extraction of features and calculation of reconstruction errors. For the input data It is processed using an autoencoder model, specifically as follows: ; Among them, represents the data vector at time , d is the dimension, is the encoder function, obtaining the low-dimensional feature representation ; Using the decoder function to reconstruct the data: ; is the data reconstructed by the autoencoder; To quantify the reconstruction effect, calculate the Euclidean distance between the input data and the reconstructed data as the reconstruction error: ; On the basis of obtaining the reconstruction error , further extract key indicators in the input data for preliminary anomaly judgment, set a fixed threshold , and extract the key indicator from , and make a judgment through the indicator function: ; Among them, when is greater than When, the value of the indicator function is 1, otherwise 0; To further enhance the attention to local data features, an attention mechanism is introduced based on the low-dimensional representation extracted by the autoencoder. The formula for calculating the attention score is as follows: ; Among them, is the attention score at time t, is the weight matrix of the attention layer, is the bias vector, is the activation function, and is the attention vector, which maps the activation result to a scalar score; Normalize the scores at all times to calculate the attention weights: ; Among them, represents the total number of times considered within the current window; is the normalized attention weight at time t, obtained by exponential normalization of the scores at all times; Integrate the fixed threshold judgment and the reconstruction error after attention weighting, and fuse the two parts of information in the form of a linear combination to obtain a composite anomaly score. The specific formula is as follows: ; Among them, is the result of the fixed threshold judgment, 0 or 1; is the composite anomaly score; is the weight coefficient, with a value within and is used to balance the fixed threshold judgment and the attention-weighted reconstruction error in the anomaly score.

[0032] According to the comparison between the composite anomaly score and the preset anomaly judgment threshold , determine whether the data is abnormal: ; When exceeds , it is determined that the data point is abnormal.

[0033] The intelligent response operation includes, after detecting abnormal data, determining a switching strategy for the data acquisition mode according to the anomaly score and anomaly type; when the abnormal data only slightly exceeds the predetermined threshold, the system switches the data acquisition mode from the standard mode to the high-frequency sampling mode, and this switching process includes issuing a switching instruction, confirming the sensor status, and confirming the mode activation; when the abnormal data significantly exceeds the predetermined range, the system activates the backup sensor and switches to the dual-channel acquisition mode; subsequently, the system starts the data correction operation, where for systematic offset anomalies, an adaptive Kalman filter-based correction algorithm is used to calculate the correction coefficient through historical data and perform real-time correction on the currently acquired data; for random noise or sudden interference anomalies, a locally weighted regression correction method is used to calculate the correction value by statistically calculating the local mean and variance of the data within the neighboring time window and adjust the abnormal data; After the correction operation is started, the system performs resampling on the same data source within a predetermined time window, and calls the redundancy verification module to compare the original data and the resampled data point by point, and uses the chi-square test and the Z-test to perform statistical tests on the data distribution and the mean respectively. At the same time, the system records the timestamp of the data acquisition mode switch, the issued instruction, the selected correction algorithm and its parameters, the resampled data, and the comparison results of the statistical tests, and writes the operation information into the operation log.

[0034] Furthermore, the data acquisition mode switching strategy is based on the anomaly score specific value and the determination of the anomaly type. First, a composite anomaly score is used to obtain a numerical representation of the current data state. According to two preset thresholds and , the anomaly score is classified: when is in the interval of , it is determined as a minor anomaly; when exceeds , it is determined as a significant anomaly. In terms of anomaly types, according to the performance of the abnormal data, it is divided into three categories: one is a systematic offset anomaly, that is, the data continuously deviates from the normal range, usually related to sensor calibration; one is a random noise anomaly, that is, the data has short-term random fluctuations, manifested as slight noise interference; one is a sudden interference anomaly, that is, the data has short-term large deviations, which may be caused by external environmental interference or sensor failure.

[0035] The data integrity verification and privacy protection based on cryptographic algorithms include, to prevent the same data from generating the same digest repeatedly and resist precomputation attacks, first, the input data is concatenated with a random salt value and then the data digest is generated through a secure hash function, and the formula is as follows: ; where, A bit string representing the original input data; Represents a random salt value of length ; Represents a concatenation operation that concatenates and into a new bit string; Is a secure hash function (SHA-256) that maps an input of any length to a bit string of a fixed length ; Is the generated data digest; When generating a signature, the current timestamp is introduced, and a hybrid quantum-secure digital signature algorithm is used to generate the signature. The formula is as follows: ; Where, Represents the timestamp of the current time; Represents a signature function using a quantum-secure digital signature algorithm (a signature algorithm based on lattice theory or a hash function); In the data transmission and subsequent processing stages, the receiver uses the public key corresponding to the private key to verify the digital signature to confirm the authenticity of the data digest. The formula is as follows: ; Where, Is the public key corresponding to the signer; Is the digital signature; Represents a digital signature verification algorithm; Is the verification result, 1 indicates verification passed, and 0 indicates verification failed; For the sensitive information field contained in the data , the system processes it using a predefined desensitization function to convert the original sensitive information into a form that cannot be directly recognized. The formula is as follows: ; Where, Represents the original sensitive information field; Is the desensitization function; Is the desensitized sensitive information; To protect the desensitized sensitive information , a symmetric encryption algorithm is used for encryption, and a dynamic key derivation mechanism is introduced to make the encryption key related to the timestamp and other auxiliary information. First, a dynamic key is generated through a key derivation function: ; Where, Is the initial key or master key; Is the auxiliary information extracted from the non-sensitive data as an additional parameter for key derivation; Represents a key derivation function (based on PBKDF2 or HKDF) for generating a dynamic key related to the current session; Is the derived symmetric encryption key; Using the generated key Encrypts the desensitized sensitive information To obtain the ciphertext : ; Among them, Represents a function for encryption using the symmetric encryption algorithm (AES); Is the encrypted ciphertext.

[0036] The parallel processing includes that the preprocessed data set Is divided into multiple subtasks according to the data type and processing requirements. Let the preprocessed data be: ; Define a partitioning function: ; So that each subtask corresponds to a subset: ; At the same time, for each subtask Define its computational load as ; Let the available processing units be a set: ; Define a static scheduling function: ; Allocate the subtask To the processing unit , and the total load on the processing unit Is: ; To adapt to the real-time change of the load during the task execution, introduce a dynamic priority weight For each subtask, where Represents the discrete time (or scheduling round), and define the dynamic effective load of the subtask as: ; The processing unit At time The dynamic total load on is: ; The scheduling goal changes to minimizing the maximum dynamic load within each time Instant, that is, solve: ; Let the average load be: ; For the task assigned to a certain processing unit above , adjust its weight using the following update rule: ; Among them, represents the preprocessed data set; represents a single data unit in the data set, i = 1,..., N; represents the data partitioning function; K represents the total number of subtasks; represents the data set of the k-th subtask; represents the subtask 's computational load; represents the set of available processing units; represents the static task scheduling function; represents the processing unit 's static total load; represents the dynamic effective load of subtask k at time t; represents the processing unit 's dynamic total load at time t; represents the average dynamic load of all processing units at time t; is the learning rate parameter for weight update; When a certain processing unit has a high load, that is , the corresponding task weight increases, thereby increasing the priority of this task in the next scheduling round, which helps the scheduling algorithm to reassign tasks to processing units with lower loads; conversely, when , the weight will decrease.

[0037] The dynamic scheduling strategy includes continuously collecting real-time load data of each processing unit through a monitoring module. The real-time load data includes CPU utilization rate, memory occupancy rate, task queue length, and response latency, and calculating the comprehensive load index using the preset weight ; comparing the comprehensive load index with the preset threshold , when the of a certain processing unit exceeds , the system adopts different strategies according to the duration of overload; If the overload is only a short-term fluctuation, the system starts a cooling timer, delays the task migration operation, and continues to observe the load changes of the unit in subsequent monitoring cycles; if the overload duration exceeds a predetermined window, the scheduling module screens the tasks to be migrated in the processing unit, determines candidate tasks according to the task execution progress, priority, data dependency, and estimated execution time, and selects in other processing units lower than the target unit, saves a snapshot of the execution context of the candidate task, migrates the task to the target unit through an asynchronous communication protocol, and starts a new thread to execute in the target unit; after migration, the system continues to monitor the status of these tasks, records the task start time, execution progress, response latency, and the current load of the target processing unit, and at the same time uses the update formula to update the dynamic priority weight of the migrated task, where represents the weight of the task at the scheduling moment ; represents the weight of the task at the scheduling moment at the next moment; is the average comprehensive load of all processing units at the current moment, is the learning rate parameter; this dynamic weight is used to adjust task allocation in subsequent scheduling cycles, so that the scheduling process performs adaptive task reallocation according to the real-time load situation. The whole process is executed in continuous monitoring cycles, and the migration and execution status of each task are recorded.

[0038] The construction of the prediction and decision model includes linear modeling of the standardized time series data, using the ARIMA model. The ARIMA model expression is: ; where represents the standardized observation value at time ; represents the mean of the data series, indicating the average level of the long-term trend; represents the lag operator, defined as , and is used to represent the data at the past time; is the autoregressive polynomial, where are the autoregressive coefficients, is the autoregressive order; represents the moving average polynomial, are the moving average coefficients, is the moving average order; represents the white noise sequence, representing the random error; Capture the linear trend and periodic changes in the time series through the ARIMA model, generate the linear prediction results based on historical data, denoted as ; To capture the non-linear features in the data, a long short-term memory network is used for modeling. The LSTM model updates the hidden state and output prediction in the following form: ; Among them, represents the LSTM hidden state at time , which represents the feature vector memorized by the model; represents the LSTM cell, which accepts the current input and the previous hidden state as inputs and outputs the current hidden state ; represents the output layer weight matrix, which is used to map the hidden state to the prediction space; represents the output layer bias vector; represents the predicted value generated by the LSTM model at time ; To adaptively adjust the contributions of the ARIMA and LSTM models in the hybrid prediction, a meta-model is constructed to calculate the dynamic weights; first, construct the feature vector: ; Among them, the prediction error and the change in error fluctuation are defined as follows: ; Among them is the standard deviation of the prediction error within a time window at time ; Calculate the dynamic weights through the meta-model: ; Among them, represents the input feature vector of the meta-model; represents the absolute value of the ARIMA prediction error; represents the absolute value of the LSTM prediction error; represents the true observed value; represents the ARIMA predicted value; represents the LSTM predicted value; represents the change in the standard deviation of the prediction error; represents the meta-model weight matrix, which is used to perform weighted summation on the features; represents the meta-model bias; represents the Sigmoid activation function, which maps the output to the interval; represents at time The calculated dynamic weight represents the contribution ratio of the ARIMA prediction result in the hybrid prediction; The two prediction results are weighted and fused to obtain a preliminary prediction value: ; Among them, represents the preliminary prediction value of the hybrid prediction model; To further correct the systematic bias in the preliminary prediction, the prediction residual is calculated and corrected using a feedforward neural network. The prediction residual is defined as: ; A feedforward neural network is constructed using the historical residual sequence to obtain a correction term: ; Among them, represents the prediction residual at time , that is, the difference between the actual observed value and the preliminary prediction value; represents the feedforward neural network function, which is used to learn the mapping relationship between the historical residual and the current prediction error; represents the previous time residual sequence, which is used as the input of the network; represents the time step of the historical residual, which is used to capture the short-term residual characteristics; represents the residual correction term, which is output by the neural network and used to compensate the preliminary prediction; The residual correction term is added to the preliminary prediction value to obtain the final prediction result: ; Among them, represents the final prediction value at time , which is the output after hybrid prediction and residual correction.

[0039] After obtaining the prediction result, the system uses the prediction value output by the hybrid prediction model and the actual observed value to compare. By calculating the prediction error and related statistics (mean and variance), the future data processing requirements and task scheduling situations are estimated.

[0040] Based on the estimation results, the data processing parameter vector (including sampling frequency, filtering threshold in the preprocessing algorithm, etc.) and the task scheduling strategy (task allocation ratio and scheduling priority) are automatically adjusted according to predefined rules and mathematical mapping functions. Specifically, the parameter update is expressed as: ; The update of the scheduling policy is as follows: ;

[0041] Among them, the function and calculate new parameters and policies according to the prediction error and the preset mapping rules.

[0042] All adjusted data processing parameters and scheduling policies, the adjustment time, the prediction error, and the actual load data are recorded in the feedback log in the form of a time series. The recorded content includes the parameters before adjustment , the parameters after adjustment , the predicted value, the actual observed value, and the error . In a predetermined period (every time units), the system calls the model update module to perform statistical analysis on the data stored in the feedback log, and uses gradient descent, Bayesian update, or other fitting methods to update the parameters of the hybrid prediction model. The update formula is: ; Among them is the learning rate, is the cost function based on the prediction error, is its gradient.

[0043] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limitations. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention. Embodiment

[0044] The second embodiment of the present invention is different from the previous two embodiments in that: When the above-mentioned functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0045] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0046] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in this computer-readable memory generate a manufactured article including an instruction device, and this instruction device implements the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0047] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0048] Although the preferred embodiments of the present application have been described, additional changes and modifications can be made by those skilled in the art once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present application.

[0049] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

1. A method for processing smart city data based on Beidou positioning, characterized in that: including collecting data from the Beidou positioning module and auxiliary sensors, and preprocessing the collected data performing event detection on the preprocessed data, and executing intelligent response operations according to the event detection results performing data integrity verification and privacy protection on the data based on cryptographic algorithms dividing the data processing tasks into multiple subtasks, and using a multi-threaded scheduling algorithm to perform parallel processing on each subtask simultaneously on a multi-core processor or a distributed computing platform, and allocating processing resources to each subtask according to a dynamic scheduling strategy performing time series analysis on historical data and real-time data, and constructing a prediction and decision-making model 2. The method for processing smart city data based on Beidou positioning according to claim 1, wherein: The preprocessing of the collected data includes noise filtering, data format conversion, and data standardization processing of the collected data 3. A method for processing smart city data based on Beidou positioning according to claim 2, characterized in that: The event detection includes performing dimensionality reduction representation and reconstruction on the input data, extracting features and calculating the reconstruction error for the input data processing it using an autoencoder model, specifically as follows: ； Among them, represents the data vector at a moment, d is the dimension, and is the encoder function that obtains the low-dimensional feature representation ; the decoder function is used to reconstruct the data: ； Data after being reconstructed by the autoencoder; to quantify the reconstruction effect, the Euclidean distance between the input data and the reconstructed data is calculated as the reconstruction error: ； On the basis of obtaining the reconstruction error further extract key indicators from the input data for preliminary anomaly judgment, set a fixed threshold , and extract key indicators from , and make a judgment through the indicator function: ； Wherein, when is greater than , the value of the indication function is 1, otherwise it is 0; To further enhance the attention to local data features, an attention mechanism is introduced based on the low-dimensional representation extracted by the autoencoder. The formula for calculating the attention score is as follows: ； wherein, is the attention score at time t, is the weight matrix of the attention layer, is the bias vector, is the activation function, and is the attention vector, mapping the activation result to a scalar score; Performing normalization processing on all moment scores to calculate the attention weights ； Among them, represents the total number of time instances considered within the current window; is the normalized attention weight at time t, obtained by exponentiating and normalizing the scores of all time instances; Combining the fixed threshold judgment and the reconstruction error after attention weighting, and fusing the two parts of information in the form of a linear combination to obtain a composite anomaly score. The specific formula is as follows ； wherein, is the result of fixed threshold judgment, 0 or 1; is the composite anomaly score; is the weight coefficient, and the value range is inside; According to the composite anomaly score Compare with the preset anomaly judgment threshold to determine whether the data is abnormal: ； When exceeds it is determined that the data point is abnormal.

4. The method for processing smart city data based on Beidou positioning according to claim 3, wherein: The intelligent response operations include switching the data collection mode when abnormal data is detected, starting the data correction operation, performing resampling and redundancy verification on the data, and recording the operation process at the same time 5. The method for processing smart city data based on Beidou positioning according to claim 4, characterized in that: The data integrity verification and privacy protection based on cryptographic algorithms include, to prevent the same data from generating the same digest repeatedly and resist precomputation attacks, concatenating the input data with a random salt value, generating a data digest through a secure hash function, signing the data digest using a quantum-secure digital signature, verifying the digital signature using the public key during data transmission and subsequent processing, and encrypting the fields containing sensitive information using desensitization processing and a symmetric encryption algorithm 6. A method for processing smart city data based on Beidou positioning according to claim 5, characterized in that: The parallel processing includes dividing the preprocessed data into multiple subtasks according to the data type and processing requirements, and using a multi-threaded scheduling algorithm to perform parallel processing on each subtask simultaneously on a multi-core processor or a distributed computing platform 7. The method for processing smart city data based on Beidou positioning according to claim 6, wherein: The dynamic scheduling strategy includes real-time monitoring of the load status and response time of each processing unit. When the load status of a certain processing unit exceeds a predetermined load threshold, the subtasks in this processing unit are reallocated to other processing units whose load status does not exceed the predetermined load threshold, and the status of the reallocated subtasks is monitored 8. The method for processing smart city data based on Beidou positioning according to claim 7, wherein: The construction of the prediction and decision-making model includes performing linear modeling on the standardized time series data, using the ARIMA model. The ARIMA model expression is ； Among them, represents the standardized observed value at time ; represents the mean of the data series, indicating the average level of the long-term trend; represents the lag operator, defined as , and is used to represent the data at the past time; is the autoregressive polynomial, where are the autoregressive coefficients, is the autoregressive order; represents the moving average polynomial, are the moving average coefficients, is the moving average order; represents the white noise sequence, representing the random error; Capture the linear trend and periodic changes in the time series through the ARIMA model, generate a linear prediction result based on historical data, denoted as ; To capture the non-linear features in the data, a long short-term memory network is used for modeling. The LSTM model updates the hidden state and output prediction in the following form ； Among them, represents the LSTM hidden state at a moment , which is the feature vector memorized by the model; represents the LSTM cell, which receives the current input and the hidden state at the previous moment as inputs, and outputs the current hidden state ; represents the output layer weight matrix, which is used to map the hidden state to the prediction space; represents the output layer bias vector; represents the prediction value generated by the LSTM model at the moment ; To adaptively adjust the contributions of the ARIMA and LSTM models in the hybrid prediction, a meta-model is constructed to calculate the dynamic weights; first, a feature vector is constructed ； where the prediction error and the change amount of error fluctuation are defined as follows ； wherein is the standard deviation of the prediction error within a time window at time ; Calculating the dynamic weights through the meta-model ； Among them, represents the input feature vector of the meta-model; represents the absolute value of the ARIMA prediction error; represents the absolute value of the LSTM prediction error; represents the true observed value; represents the ARIMA predicted value; represents the LSTM predicted value; represents the change in the standard deviation of the prediction error; represents the meta-model weight matrix, which is used to perform weighted summation on the features; represents the meta-model bias; represents the Sigmoid activation function, which maps the output to interval; represents at time The calculated dynamic weight represents the contribution ratio of the ARIMA prediction result in the hybrid prediction; Weighting and fusing the two prediction results to obtain a preliminary prediction value ； Among them, represents the preliminary prediction value of the hybrid prediction model; To further correct the systematic bias in the preliminary prediction, calculate the prediction residuals and use a feedforward neural network to correct them. The prediction residuals are defined as ； Constructing a feedforward neural network using the historical residual sequence to obtain a correction term ； Among them, represents the prediction residual at time , that is, the difference between the actual observed value and the preliminary prediction value; represents the feedforward neural network function, which is used to learn the mapping relationship between historical residuals and the current prediction error; represents the previous residual sequence at time as the input of the network; represents the residual correction term, which is output by the neural network and used to compensate the preliminary prediction; Add the residual correction term to the preliminary prediction value to obtain the final prediction result: ； Among them, represents the final predicted value at time , which is the output after hybrid prediction and residual correction.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images