Cell data processing method and apparatus, electronic device, and storage medium

Abstract

The application discloses a cell data processing method and device, electronic equipment and storage medium, wherein the method comprises: determining a first sequence based on first information of each cell in a first cell set; the first information represents time sequence data of each index in at least two indexes in the corresponding cell; the first sequence represents a hole rate sequence determined based on the period and time granularity of the time sequence data in the first information; constructing a first processing model based on at least two parallel hole convolutional neural networks; each element in the first sequence is used to configure a hole rate of a hole convolutional neural network; calling the first processing model to process the first information to obtain second information corresponding to each cell; the second information represents multi-scale features of the corresponding cell corresponding to each index in the at least one index.

Description

Technical Field

[0001] This application relates to the field of wireless technology, and in particular to a cell data processing method, apparatus, electronic device, and storage medium. Background Technology

[0002] Resource scheduling, anomaly detection, and quality early warning in wireless cells all rely on accurate multi-indicator spatiotemporal prediction. The prerequisite for multi-indicator spatiotemporal prediction is accurate feature extraction from the cell's multi-indicator time-series data. However, current technologies that rely on feature extraction from the cell's multi-indicator time-series data cannot yield accurate multi-indicator spatiotemporal prediction results. Summary of the Invention

[0003] To address the related technical issues, embodiments of this application provide a cell data processing method, apparatus, electronic device, and storage medium.

[0004] The technical solution of this application embodiment is implemented as follows: This application provides a cell data processing method, including: Based on the first information of each cell in the first cell set, a first sequence is determined; the first information represents the time series data of each of at least two indicators in the corresponding cell; the first sequence represents the hole rate sequence determined based on the period and time granularity of the time series data in the first information. A first processing model is constructed based on at least two parallel dilated convolutional neural networks; each element in the first sequence is used to configure the dilation rate of a dilated convolutional neural network. The first processing model is invoked to process the first information to obtain the second information corresponding to each cell; the second information represents the multi-scale features of each of the at least one index of the corresponding cell.

[0005] The method in the above scheme further includes: Initialize the cell groups to obtain at least three second cell sets; If the first condition is met, iterate through all set pairs, and if the second condition is met, perform the following merge process: Merge the cells in the currently traversed set pair into a new second cell set, and delete the set pair that contains any second cell set in the currently traversed set pair; If the first set of conditions is not met, each cell set obtained after processing will be output as a first cell set; wherein, Each set pair contains two second cell sets; the first set condition indicates that there is no second cell set with a cell count greater than a set number; the second set condition indicates that the ratio of the first cell count to the count of any second cell count is greater than a set ratio; the first cell count indicates the number of cells in the intersection of the two second cell sets in the currently traversed set pair; the second cell count indicates the number of cells in one of the second cell sets in the currently traversed set pair.

[0006] In the above scheme, the initial grouping of cells includes: Based on the adjacency matrix corresponding to each cell, determine the number of neighbors of each cell in all cells; In the ungrouped communities, select the community with the largest number of neighbors, and the neighbor communities of the community with the largest number of neighbors, to generate a second community set.

[0007] In the above scheme, the first sequence is determined based on the maximum period and minimum time granularity of the time-series data in the first information.

[0008] The method in the above scheme further includes: The upper limit of the void ratio is determined based on the maximum period; The lower limit of the void ratio is determined based on the minimum time granularity. The first sequence is generated based on the lower limit of the void ratio, and the last item in the first sequence is less than or equal to the upper limit of the void ratio.

[0009] The method in the above scheme further includes: For the time-series data of each cell on each of the at least two indicators, the time-series data of the second indicator is padded with time granularity based on the time granularity corresponding to the first indicator; wherein, The first indicator represents the indicator with the smallest time granularity among the at least two indicators; the second indicator represents the other indicators among the at least two indicators besides the first indicator.

[0010] The method in the above scheme further includes: Calculate the autocorrelation function value of each time series data for each cell on each indicator; Based on the distance between adjacent non-negative extreme points of the autocorrelation function value corresponding to each indicator, and based on the time granularity corresponding to the time series data of each indicator, the period corresponding to each indicator of each cell is calculated.

[0011] The method in the above scheme further includes: By upsampling or downsampling the features at each scale in the second information, the features at each scale are converted into features at at least three predetermined scales. The features at the at least three defined scales are spliced ​​together to obtain fused multi-scale features.

[0012] This application also provides a cell data processing apparatus, including: The first determining unit is used to determine a first sequence based on the first information of each cell in the first cell set; the first information represents the time-series data of each of at least two indicators in the corresponding cell; the first sequence represents the hole rate sequence determined based on the period and time granularity of the time-series data in the first information. The first building unit is used to build a first processing model based on at least two parallel dilated convolutional neural networks; each element in the first sequence is used to configure the dilation rate of a dilated convolutional neural network. The first calling unit is used to call the first processing model to process the first information and obtain the second information corresponding to each cell; the second information represents the multi-scale features of each of the at least one index of the corresponding cell.

[0013] This application also provides an electronic device, including: a first processor and a first communication interface; wherein, The first processor is configured to determine a first sequence based on first information of each cell in the first cell set; the first information represents the time-series data of each of at least two indicators in the corresponding cell; the first sequence represents the hole rate sequence determined based on the period and time granularity of the time-series data in the first information. The first processor is further configured to construct a first processing model based on at least two parallel dilated convolutional neural networks; each element in the first sequence is configured to correspond to the dilation rate of a dilated convolutional neural network. The first processor is further configured to call the first processing model to process the first information and obtain the second information corresponding to each cell; the second information represents the multi-scale features of each of the at least one index of the corresponding cell.

[0014] This application also provides an electronic device, including: a first processor and a first memory for storing a computer program capable of running on the processor. Wherein, when the first processor is used to run the computer program, it executes the steps of any of the above methods.

[0015] This application also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of any of the above methods.

[0016] The cell data processing method, apparatus, electronic device, and storage medium provided in this application first determine a first sequence based on the first information of each cell in a first cell set. Here, the first information represents the time-series data of each of at least two indicators in the corresponding cell; the first sequence represents the dilatancy rate sequence determined based on the period and time granularity of the time-series data in the first information. Secondly, a first processing model is constructed based on at least two parallel dilated convolutional neural networks, wherein each element in the first sequence is used to configure the dilatancy rate of a corresponding dilated convolutional neural network. Finally, the first processing model is invoked to process the first information, obtaining multi-scale features of each cell corresponding to each of the at least one indicator. Based on the above scheme, efficient and accurate feature extraction of multi-indicator time-series data of cells can be achieved, and multi-indicator spatiotemporal prediction can be performed on this basis, improving the accuracy of prediction. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart of a cell data processing method according to an embodiment of this application; Figure 2 This is a schematic diagram illustrating the principle and process of multi-scale feature fusion in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a cell data processing device according to an embodiment of this application; Figure 4 This is a schematic diagram of the electronic device structure according to an embodiment of this application. Detailed Implementation

[0018] Resource scheduling, anomaly detection, and quality early warning in wireless cells all rely on accurate multi-indicator spatiotemporal prediction. A key aspect is the efficient and accurate extraction of multi-indicator time-series data features from cells. Related technologies include methods for extracting multi-indicator time-series data features, such as: 1. Extract statistical features from multi-indicator time-series data, such as mean and variance. This method is relatively fast, but sudden fluctuations in cell data can severely affect the calculation of statistical features. Furthermore, the aforementioned statistical features are difficult to reflect the essential information of the data and cannot improve the accuracy of multi-indicator spatiotemporal prediction.

[0019] Second, using a learnable matrix of coefficients to extract multi-indicator time-series data features. This method does not consider the correlation between multi-indicator data features in different cells, and therefore cannot improve the accuracy of multi-indicator spatiotemporal prediction.

[0020] Third, using convolutional neural networks to extract features from multi-indicator time-series data. This method does not consider the relationships between multiple indicator data within a single cell or the relationships between individual indicator data between different cells, and therefore cannot improve the accuracy of multi-indicator spatiotemporal prediction.

[0021] In summary, among the related technologies, feature extraction from multi-indicator time-series data of a community often fails to yield accurate spatiotemporal prediction results for multiple indicators.

[0022] Based on this, in each embodiment of this application, a first sequence is first determined based on the first information of each cell in the first cell set. Here, the first information represents the time-series data of each of at least two indicators in the corresponding cell; the first sequence represents the dilatancy rate sequence determined based on the period and time granularity of the time-series data in the first information. Secondly, a first processing model is constructed based on at least two parallel dilated convolutional neural networks, wherein each element in the first sequence is used to configure the dilatancy rate of a dilated convolutional neural network. Finally, the first processing model is invoked to process the first information, obtaining multi-scale features of each cell corresponding to each of the at least one indicator. Based on the above scheme, efficient and accurate feature extraction of multi-indicator time-series data of cells can be achieved, and multi-indicator spatiotemporal prediction based on this can improve the accuracy of prediction.

[0023] The present application will now be described in further detail with reference to the accompanying drawings and embodiments.

[0024] First, for ease of explanation, let's define the following application scenario: We need to extract features from N cells, each with M indicators. The maximum length of the indicator sequence is L, and the adjacency matrix of the cells is A. For example, considering the actual needs of a certain province, we need to extract features from approximately 10,000 cells, each with dozens of indicators, to support subsequent indicator prediction and other requirements.

[0025] This application provides a cell data processing method, see [link to relevant documentation]. Figure 1 The methods include: Step 101: Determine the first sequence based on the first information of each cell in the first cell set.

[0026] The first information represents the time-series data of each of the at least two indicators in the corresponding cell, such as the time-series data of indicators such as access latency, link layer quality, and throughput in each cell; the first sequence represents the hole rate sequence determined based on the period and time granularity of the time-series data in the first information.

[0027] In practical applications, since the time series data of different indicators often have large numerical differences, in order to ensure the convergence of the subsequent model, the time series data of different indicators can be normalized separately before determining the first sequence. Then, the first sequence can be determined based on the normalized time series data.

[0028] Specifically, the z-score standardization method can be used, that is, for each metric, normalization is performed separately in the following way:

[0029] in, i =1, 2, …, M j =1, 2, …, N, Characterizing the first j The first in the community i Time series data for each indicator Characterizing the first i The mean of a metric in time-series data across N cells. Characterizing the first i The variance of a given indicator in time-series data across N cells. The normalized representation of the first j The first in the community i Time series data for each indicator.

[0030] Here, the first sequence is a void rate sequence determined based on the period and time granularity of the time-series data in the first information. In one embodiment, the first sequence is determined based on the maximum period and minimum time granularity of the time-series data in the first information. The method for calculating the corresponding void rate sequence further includes: The upper limit of the void ratio is determined based on the maximum period; The lower limit of the void ratio is determined based on the minimum time granularity. The first sequence is generated based on the lower limit of the void ratio, and the last item in the first sequence is less than or equal to the upper limit of the void ratio.

[0031] For example: First, calculate the upper limit of the void ratio using the following formula:

[0032] in, This represents the maximum value of the time series data for each indicator.

[0033] Next, take the minimum value of the time granularity of each indicator. Calculate the lower limit of the void ratio:

[0034] Thus, the void ratio sequence is obtained:

[0035] Where the last term of the sequence is less than or equal to .

[0036] In one embodiment, the method further includes calculating the period of the time-series data in the first information: Calculate the autocorrelation function value of each time series data for each cell on each indicator; Based on the distance between adjacent non-negative extreme points of the autocorrelation function value corresponding to each indicator, and based on the time granularity corresponding to the time series data of each indicator, the period corresponding to each indicator of each cell is calculated.

[0037] Specifically, no. j The first community i The period calculation for the time series data of each indicator is as follows:

[0038] in, and Respectively characterize the first j The first in the community i The mean and variance of each indicator.

[0039] Next, the non-negative extreme points in the autocorrelation function are extracted, and the average distance between adjacent extreme points is calculated to obtain the first extreme point. j The first in the community i Number of data points within a period of time series data for each indicator , thus calculating the first i Average duration of one indicator over one period :

[0040]

[0041] in, Characterizing the first i The time granularity of the time series data corresponding to each indicator.

[0042] In practical applications, considering that different indicators may have different time granularities, the time-series data of different indicators in the first information can be aligned in terms of time granularity before determining the first sequence. Based on this, in one embodiment, the method further includes: For the time-series data of each cell on each of the at least two indicators, the time-series data of the second indicator is padded with time granularity based on the time granularity corresponding to the first indicator; wherein, The first indicator represents the indicator with the smallest time granularity among the at least two indicators; the second indicator represents the other indicators among the at least two indicators besides the first indicator.

[0043] Specifically, the first step is to select a benchmark time granularity. For example, the smallest time granularity among all indicators is selected as the benchmark time granularity. Then, with Based on this, the time when all indicator data is completed is used, with missing values ​​filled with 0, thus obtaining fine-grained multi-indicator data for the community. ,in, l Characterizes the length of the input sequence.

[0044] Step 102: Construct the first processing model based on at least two parallel dilated convolutional neural networks.

[0045] Each element in the first sequence is used to configure the dilation rate of a dilated convolutional neural network.

[0046] Step 103: Call the first processing model to process the first information and obtain the second information corresponding to each cell.

[0047] The second information characterization corresponds to the multi-scale features of each of the at least one index.

[0048] Here, multiple parallel dilated convolutional neural networks are constructed. Each network uses one element from the first sequence as the dilation limit, expanding the convolutional kernel in both time and metric dimensions, performing three consecutive dilated convolutions. Different dilation rates produce receptive fields of different sizes, resulting in multi-scale initial features after convolution. In this way, dilated convolutional neural networks can expand their receptive field according to the dilation rate while maintaining the same computational cost.

[0049] In practical applications, the number of cells can be enormous, leading to a huge computational cost if all cells are processed directly. Therefore, in this embodiment, the cells are divided into multiple groups, and corresponding operations are performed on each group. This not only improves the correlation between cells within a group but also reduces the overall computational cost.

[0050] Specifically, when grouping cells, the method further includes: Initialize the cell groups to obtain at least three second cell sets; If the first condition is met, iterate through all set pairs, and if the second condition is met, perform the following merge process: Merge the cells in the currently traversed set pair into a new second cell set, and delete the set pair that contains any second cell set in the currently traversed set pair; If the first set condition is not met, each cell set obtained from the processing will be output as a first cell set.

[0051] Each set pair contains two second cell sets; the first set condition indicates that there is no second cell set with a cell count greater than a set number; the second set condition indicates that the ratio of the number of first cells to the number of any second cells is greater than a set ratio; the first cell count indicates the number of cells in the intersection of the two second cell sets in the currently traversed set pair; the second cell count indicates the number of cells in one of the second cell sets in the currently traversed set pair.

[0052] The initialization of cell grouping includes: Based on the adjacency matrix corresponding to each cell, determine the number of neighbors of each cell in all cells; In the ungrouped communities, select the community with the largest number of neighbors, and the neighbor communities of the community with the largest number of neighbors, to generate a second community set.

[0053] Specifically, the grouping method is as follows: 1. Set the upper limit for the number of cells in a group. All cells are marked as "ungrouped". At the same time, the number of neighbors of each cell is calculated using the adjacency matrix A.

[0054] 2. Add the neighborhood with the most neighbors and its neighbors into a group and mark it as "Grouped". Repeat step 2 until there are no more "Ungrouped" neighborhoods.

[0055] At this point, several initial groups are obtained, and there may be overlap between different initial groups.

[0056] 3. Calculate the number of elements in the intersection between each pair of initial groups.

[0057] 4. Starting from the initial group with the largest intersection, traverse in descending order. If the number of elements in the intersection is greater than 20% of the elements in the currently traversed initial group, delete the other initial group, and then retrieve the elements from the group that contains the elements in the first initial group. 5. For each cell group, if the number of cells in the group is greater than 1, repeat the above steps until the number of cells in the group is less than 1. .

[0058] This gives us the grouping of all the cells:

[0059]

[0060] In other words, in this embodiment of the application, the first cell set can be any one of the cell sets obtained by grouping N cells according to the above cell grouping method.

[0061] Thus, in this embodiment of the application, the convolution kernel is expanded in both time and index dimensions to extract long-term temporal relationships and spatial cross-index and cross-cell relationships.

[0062] Specifically, the dilated convolution formula is as follows:

[0063] in, , , represents the multi-indicator data of the cells in each group; k represents the convolution kernel, with an initial size of 3. 3; Characterizes the void ratio.

[0064] In one embodiment, the method further includes: By upsampling or downsampling the features at each scale in the second information, the features at each scale are converted into features at at least three predetermined scales. The features at the at least three defined scales are spliced ​​together to obtain fused multi-scale features.

[0065] Combination Figure 2 The diagram illustrates the principle of multi-scale feature fusion. Specifically, it selects a minimum scale of 1 and a medium scale of 1. Maximum scale Then, by upsampling or downsampling, features at other scales are sampled to the above three scales respectively, and then added point-to-point to obtain features at the three scales. Finally, the features from the three scales were concatenated to obtain the multi-index features. .

[0066] Based on the embodiments of this application, efficient and accurate feature extraction can be performed on multi-indicator time series data of multiple cells. Specifically, the data processing process has low complexity and fast processing speed. Furthermore, during the feature extraction process, it can capture the correlation between time series data of different indicators within a single cell, as well as the correlation between time series data of different indicators in different cells, thereby achieving accurate feature extraction. Based on this, spatiotemporal prediction of multiple indicators can be performed to obtain more accurate prediction results.

[0067] To implement the cell data processing method of this application embodiment, this application embodiment also provides a cell data processing apparatus, such as... Figure 3 As shown, the device includes: The first determining unit 301 is used to determine a first sequence based on the first information of each cell in the first cell set; the first information represents the time-series data of each of at least two indicators in the corresponding cell; the first sequence represents the hole rate sequence determined based on the period and time granularity of the time-series data in the first information. The first building unit 302 is used to build a first processing model based on at least two parallel dilated convolutional neural networks; each element in the first sequence is used to configure the dilation rate of a dilated convolutional neural network. The first calling unit 303 is used to call the first processing model to process the first information and obtain the second information corresponding to each cell; the second information represents the multi-scale features of each of the at least one index of the corresponding cell.

[0068] In one embodiment, the device further includes: Grouping units, used for: Initialize the cell groups to obtain at least three second cell sets; If the first condition is met, iterate through all set pairs, and if the second condition is met, perform the following merge process: Merge the cells in the currently traversed set pair into a new second cell set, and delete the set pair that contains any second cell set in the currently traversed set pair; If the first set of conditions is not met, each cell set obtained after processing will be output as a first cell set; wherein, Each set pair contains two second cell sets; the first set condition indicates that there is no second cell set with a cell count greater than a set number; the second set condition indicates that the ratio of the first cell count to the count of any second cell count is greater than a set ratio; the first cell count indicates the number of cells in the intersection of the two second cell sets in the currently traversed set pair; the second cell count indicates the number of cells in one of the second cell sets in the currently traversed set pair.

[0069] In one embodiment, the packetization unit initializes cell packetization, including: Based on the adjacency matrix corresponding to each cell, determine the number of neighbors of each cell in all cells; In the ungrouped communities, select the community with the largest number of neighbors, and the neighbor communities of the community with the largest number of neighbors, to generate a second community set.

[0070] In one embodiment, the first sequence is determined based on the maximum period and minimum time granularity of the time-series data in the first information.

[0071] In one embodiment, the first determining unit 301 is specifically used for: The upper limit of the void ratio is determined based on the maximum period; The lower limit of the void ratio is determined based on the minimum time granularity. The first sequence is generated based on the lower limit of the void ratio, and the last item in the first sequence is less than or equal to the upper limit of the void ratio.

[0072] In one embodiment, the device further includes: A time granularity alignment unit is used to fill the time granularity of the time-series data of the second indicator for each cell's time-series data across the at least two indicators, using the time granularity corresponding to the first indicator as a reference; wherein, The first indicator represents the indicator with the smallest time granularity among the at least two indicators; the second indicator represents the other indicators among the at least two indicators besides the first indicator.

[0073] In one embodiment, the device further includes: Periodic calculation unit, used for: Calculate the autocorrelation function value of each time series data for each cell on each indicator; Based on the distance between adjacent non-negative extreme points of the autocorrelation function value corresponding to each indicator, and based on the time granularity corresponding to the time series data of each indicator, the period corresponding to each indicator of each cell is calculated.

[0074] In one embodiment, the device further includes: The feature fusion unit is used to convert the features at each scale into features at at least three predetermined scales by upsampling or downsampling the features at each scale in the second information. The features at the at least three defined scales are spliced ​​together to obtain fused multi-scale features.

[0075] In practical applications, the above-mentioned units can be implemented by the processor in the cell data processing device in combination with the communication interface.

[0076] It should be noted that the cell data processing device provided in the above embodiments is only illustrated by the division of the above program modules when performing cell data processing. In actual applications, the above processing can be assigned to different program modules as needed, that is, the internal structure of the device can be divided into different program modules to complete all or part of the processing described above. In addition, the cell data processing device and the cell data processing method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0077] Based on the hardware implementation of the above program modules, and in order to implement the method on the electronic device side of the embodiments of this application, the embodiments of this application also provide an electronic device, such as... Figure 4 As shown, the electronic device 400 includes: The first communication interface 401 is capable of exchanging information with other network nodes; The first processor 402 is connected to the first communication interface 401 to enable information interaction with other network nodes and to execute the methods provided by one or more technical solutions on the electronic device side when running a computer program. The computer program is stored in the first memory 403.

[0078] Specifically, the first processor 402 is used for: Based on the first information of each cell in the first cell set, a first sequence is determined; the first information represents the time series data of each of at least two indicators in the corresponding cell; the first sequence represents the hole rate sequence determined based on the period and time granularity of the time series data in the first information. A first processing model is constructed based on at least two parallel dilated convolutional neural networks; each element in the first sequence is used to configure the dilation rate of a dilated convolutional neural network. The first processing model is invoked to process the first information to obtain the second information corresponding to each cell; the second information represents the multi-scale features of each of the at least one index of the corresponding cell.

[0079] In one embodiment, the first processor 402 is further configured to: Initialize the cell groups to obtain at least three second cell sets; If the first condition is met, iterate through all set pairs, and if the second condition is met, perform the following merge process: Merge the cells in the currently traversed set pair into a new second cell set, and delete the set pair that contains any second cell set in the currently traversed set pair; If the first set of conditions is not met, each cell set obtained after processing will be output as a first cell set; wherein, Each set pair contains two second cell sets; the first set condition indicates that there is no second cell set with a cell count greater than a set number; the second set condition indicates that the ratio of the first cell count to the count of any second cell count is greater than a set ratio; the first cell count indicates the number of cells in the intersection of the two second cell sets in the currently traversed set pair; the second cell count indicates the number of cells in one of the second cell sets in the currently traversed set pair.

[0080] In one embodiment, the first processor 402 initializes cell groups, including: Based on the adjacency matrix corresponding to each cell, determine the number of neighbors of each cell in all cells; In the ungrouped communities, select the community with the largest number of neighbors, and the neighbor communities of the community with the largest number of neighbors, to generate a second community set.

[0081] In one embodiment, the first sequence is determined based on the maximum period and minimum time granularity of the time-series data in the first information.

[0082] In one embodiment, the first processor 402 is further configured to: The upper limit of the void ratio is determined based on the maximum period; The lower limit of the void ratio is determined based on the minimum time granularity. The first sequence is generated based on the lower limit of the void ratio, and the last item in the first sequence is less than or equal to the upper limit of the void ratio.

[0083] In one embodiment, the first processor 402 is further configured to: For the time-series data of each cell on each of the at least two indicators, the time-series data of the second indicator is padded with time granularity based on the time granularity corresponding to the first indicator; wherein, The first indicator represents the indicator with the smallest time granularity among the at least two indicators; the second indicator represents the other indicators among the at least two indicators besides the first indicator.

[0084] In one embodiment, the first processor 402 is further configured to: Calculate the autocorrelation function value of each time series data for each cell on each indicator; Based on the distance between adjacent non-negative extreme points of the autocorrelation function value corresponding to each indicator, and based on the time granularity corresponding to the time series data of each indicator, the period corresponding to each indicator of each cell is calculated.

[0085] In one embodiment, the first processor 402 is further configured to: By upsampling or downsampling the features at each scale in the second information, the features at each scale are converted into features at at least three predetermined scales. The features at the at least three defined scales are spliced ​​together to obtain fused multi-scale features.

[0086] It should be noted that the specific processing procedures of the first processor 402 and the first communication interface 401 can be understood by referring to the above method.

[0087] Of course, in practical applications, the various components in electronic device 400 are coupled together through bus system 404. It can be understood that bus system 404 is used to realize the connection and communication between these components. In addition to a data bus, bus system 404 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, in... Figure 4 The general designated all buses as Bus System 404.

[0088] The first memory 403 in this embodiment is used to store various types of data to support the operation of the electronic device 400. Examples of such data include any computer program used to operate on the electronic device 400.

[0089] The methods disclosed in the embodiments of this application can be applied to the first processor 402, or implemented by the first processor 402. The first processor 402 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware or by instructions in the form of software in the first processor 402. The first processor 402 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The first processor 402 can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in the first memory 403. The first processor 402 reads the information in the first memory 403 and completes the steps of the aforementioned method in combination with its hardware.

[0090] In an exemplary embodiment, the electronic device 400 may be implemented by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components to perform the aforementioned method.

[0091] It is understood that the first memory 403 in this embodiment can be volatile memory or non-volatile memory, or both. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), ferromagnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); the magnetic surface memory can be disk storage or magnetic tape storage. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), SyncLink Dynamic Random Access Memory (SLDRAM), and Direct Rambus Random Access Memory (DRRAM).The memories described in the embodiments of this application are intended to include, but are not limited to, these and any other suitable types of memories.

[0092] In an exemplary embodiment, this application also provides a storage medium, namely a computer storage medium, specifically a computer-readable storage medium, such as a first memory 403 storing a computer program, which can be executed by a first processor 402 of an electronic device 400 to complete the steps described in the aforementioned electronic device-side method. The computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, or CD-ROM.

[0093] It should be noted that terms such as "first" and "second" are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0094] Furthermore, the technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0095] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application.

Claims

1. A method for processing community data, characterized in that, include: Based on the first information of each cell in the first cell set, determine the first sequence; The first information represents the time-series data of each of at least two indicators in the corresponding cell; The first sequence represents a void rate sequence generated based on the lower limit of void rate, arranged in ascending order of elements. The last item of the first sequence is less than or equal to the upper limit of void rate. The lower limit of void rate is calculated based on the minimum time granularity of the time series data in the first information, and the upper limit of void rate is determined based on the maximum period of the time series data in the first information. A first processing model is constructed based on at least two parallel dilated convolutional neural networks; each element in the first sequence is used to configure the dilation rate of a dilated convolutional neural network. The first processing model is invoked to process the first information and obtain the second information corresponding to each cell. The second information characterization corresponds to the multi-scale features of each of the at least one index in the cell.

2. The method according to claim 1, characterized in that, The method further includes: Initialize the cell groups to obtain at least three second cell sets; If the first condition is met, iterate through all set pairs, and if the second condition is met, perform the following merge process: Merge the cells in the currently traversed set pair into a new second cell set, and delete the set pair that contains any second cell set in the currently traversed set pair; If the first set of conditions is not met, each cell set obtained after processing will be output as a first cell set; wherein, Each set pair contains two second cell sets; the first set condition indicates that there is no second cell set with a cell count greater than a set number; the second set condition indicates that the ratio of the first cell count to the count of any second cell count is greater than a set ratio; the first cell count indicates the number of cells in the intersection of the two second cell sets in the currently traversed set pair; the second cell count indicates the number of cells in one of the second cell sets in the currently traversed set pair.

3. The method according to claim 2, characterized in that, The initialization of cell grouping includes: Based on the adjacency matrix corresponding to each cell, determine the number of neighbors of each cell in all cells; In the ungrouped communities, select the community with the largest number of neighbors, and the neighbor communities of the community with the largest number of neighbors, to generate a second community set.

4. The method according to claim 1, characterized in that, The method further includes: For the time-series data of each cell on each of the at least two indicators, the time-series data of the second indicator is padded with time granularity based on the time granularity corresponding to the first indicator; wherein, The first indicator represents the indicator with the smallest time granularity among the at least two indicators; the second indicator represents the other indicators among the at least two indicators besides the first indicator.

5. The method according to claim 1, characterized in that, The method further includes: Calculate the autocorrelation function value of each time series data for each cell on each indicator; Based on the distance between adjacent non-negative extreme points of the autocorrelation function value corresponding to each indicator, and based on the time granularity corresponding to the time series data of each indicator, the period corresponding to each indicator of each cell is calculated.

6. The method according to claim 1, characterized in that, The method further includes: By upsampling or downsampling the features at each scale in the second information, the features at each scale are converted into features at at least three predetermined scales. The features at the at least three defined scales are spliced ​​together to obtain fused multi-scale features.

7. A community data processing device, characterized in that, include: The first determining unit is used to determine a first sequence based on the first information of each cell in the first cell set; The first information represents the time-series data of each of at least two indicators in the corresponding cell; The first sequence represents a void rate sequence generated based on the lower limit of void rate, arranged in ascending order of elements. The last item of the first sequence is less than or equal to the upper limit of void rate. The lower limit of void rate is calculated based on the minimum time granularity of the time series data in the first information, and the upper limit of void rate is determined based on the maximum period of the time series data in the first information. The first building unit is used to build a first processing model based on at least two parallel dilated convolutional neural networks; each element in the first sequence is used to configure the dilation rate of a dilated convolutional neural network. The first calling unit is used to call the first processing model to process the first information and obtain the second information corresponding to each cell; The second information characterization corresponds to the multi-scale features of each of the at least one index in the cell.

8. An electronic device, characterized in that, include: A first processor and a first communication interface; wherein... The first processor is configured to determine a first sequence based on first information of each cell in the first cell set; the first information represents the time-series data of each of at least two indicators in the corresponding cell; the first sequence represents a hole rate sequence generated based on the lower limit of hole rate in ascending order, wherein the last item of the first sequence is less than or equal to the upper limit of hole rate, the lower limit of hole rate is calculated based on the minimum time granularity of the time-series data in the first information, and the upper limit of hole rate is determined based on the maximum period of the time-series data in the first information; The first processor is further configured to construct a first processing model based on at least two parallel dilated convolutional neural networks; each element in the first sequence is configured to correspond to the dilation rate of a dilated convolutional neural network. The first processor is further configured to invoke the first processing model to process the first information. The second information corresponding to each cell is obtained; the second information represents the multi-scale features of each of the at least one index of the corresponding cell.

9. An electronic device, characterized in that, include: A first processor and a first memory for storing computer programs capable of running on the processor. Wherein, when the first processor is used to run the computer program, it performs the steps of the method according to any one of claims 1 to 6.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images