CNN-based user classification method and system using log spectrum analysis

Abstract

Provided in the embodiments of the present invention are a CNN-based user classification method and system using log spectrum analysis. The method comprises: when a user classification request is detected, acquiring target historical data in historical log data of a target user; determining a request factor and user classification data according to a request target, acquiring a data time of the user classification data, and generating a time series of the target user on the basis of the user classification data and the corresponding data time; dividing the time series into time series sets according to time intervals of the data time, standardizing the time series sets, and transforming the time series sets into corresponding spectrograms by means of Fourier transform; and building a corresponding convolutional neural network (CNN) model, inputting the spectrograms into the model for classification training, and determining a classification result on the basis of a classification training result. In this way, log data analysis of a user can be shifted from a time domain to a frequency domain, thereby avoiding the loss of time information and further improving the accuracy of a classification effect.

Description

A user classification method and system based on CNN usage log spectrum analysis Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a user classification method and system based on CNN usage log spectrum analysis. Background Art

[0002] The Internet has a huge user base, especially social, e-commerce, and financial service platforms. Whether it is to promote user activity on the platform, or to recommend corresponding products to corresponding users, or to match friends in social networks, user classification is involved. It can be said that user classification is one of the most important technologies on the Internet.

[0003] Currently, because user log data is too large and there is time correlation between log data, it is impossible to directly use log data to train the model. However, the existing technology is based on log statistics. As a result, most of the time information will be lost after the statistics are completed, resulting in inaccurate classification results.

[0004] Summary of the Invention

[0005] In response to the problems existing in the prior art, an embodiment of the present invention provides a user classification method and system based on CNN usage log spectrum analysis.

[0006] An embodiment of the present invention provides a user classification method based on CNN usage log spectrum analysis, including:

[0007] When a user classification request is detected, a target user and a request target corresponding to the user classification request are obtained, a corresponding request time period is determined based on the request target, and target historical data in the historical log data of the target user is obtained based on the request time period;

[0008] Determining a corresponding request factor according to the request target, determining user classification data in the target historical data based on the request factor, obtaining a data time corresponding to the user classification data, and generating a time series corresponding to the target user based on the user classification data and the corresponding data time;

[0009] Dividing the time series into time series sets according to the time interval of the data time, standardizing the time series sets to obtain standardized time series sets, and converting the standardized time series sets into corresponding frequency spectrograms through Fourier transform;

[0010] Based on the request purpose and request factors, a corresponding CNN convolutional neural network model is built, the spectrum graph is input into the CNN convolutional neural network model for classification training, and the classification result of the target user is determined based on the classification training result. In one embodiment, the method further includes:

[0011] Obtaining the number of user classification requests containing the request target within a preset period, determining the priority of the user classification request based on the number of requests, and determining a corresponding user classification request sequence based on the priority;

[0012] The priority of the corresponding request target is determined based on the priority in the user classification request sequence, and the learning weight of the convolution layer in the CNN convolutional neural network is adjusted based on the priority of the request target.

[0013] In one embodiment, the method further comprises:

[0014] Based on the classification result, a corresponding classification label is generated, and a corresponding update period is determined based on the requested time period;

[0015] The classification label is saved in the historical log data of the corresponding target user, and the classification label is updated based on the update cycle.

[0016] In one embodiment, the method further comprises:

[0017] Acquire a historical comparison set based on the user classification data, and divide the historical comparison set into a standard set and a verification set according to a preset ratio, wherein the historical comparison set is historical data associated with the user classification data in a database and has been classified;

[0018] Generate a standard spectrum graph corresponding to the standard set data, and input the standard matching graph and the corresponding classification results into the CNN convolutional neural network as standard parameters of the neural network model;

[0019] Obtain the classification training results of the CNN convolutional neural network model, and compare the classification training results with the validation set data. If the comparison result meets the preset accuracy requirements, determine the classification result of the target user corresponding to the classification training.

[0020] In one embodiment, the method further comprises:

[0021] The normalization is Z-score normalization.

[0022] An embodiment of the present invention provides a user classification system based on CNN usage log spectrum analysis, including:

[0023] a detection module configured to, upon detecting a user classification request, obtain a target user and a request target corresponding to the user classification request, determine a corresponding request time period based on the request target, and obtain target historical data from the historical log data of the target user based on the request time period;

[0024] a sequence module, configured to determine a corresponding request factor according to the request target, determine user classification data in the target historical data based on the request factor, obtain a data time corresponding to the user classification data, and generate a time series corresponding to the target user based on the user classification data and the corresponding data time;

[0025] a spectrum module, configured to divide the time series into time series sets according to the time intervals of the data time, standardize the time series sets to obtain standardized time series sets, and convert the standardized time series sets into corresponding spectrum graphs through Fourier transform;

[0026] A training module is used to build a corresponding CNN convolutional neural network model based on the request purpose and request factor, input the spectrum graph into the CNN convolutional neural network model for classification training, and determine the classification result of the target user based on the classification training result.

[0027] In one embodiment, the system further comprises:

[0028] An acquisition module is used to obtain the number of user classification requests containing the request target within a preset period, determine the priority of the user classification request based on the number of requests, and determine the corresponding user classification request sequence based on the priority; a weight module is used to determine the priority of the corresponding request target based on the priority in the user classification request sequence, and adjust the learning weight of the convolution layer in the CNN convolutional neural network based on the priority of the request target.

[0029] In one embodiment, the system further comprises:

[0030] a label module, configured to generate a corresponding classification label based on the classification result, and determine a corresponding update period based on the requested time period;

[0031] An updating module is used to save the classification label to the historical log data of the corresponding target user and update the classification label based on the update cycle.

[0032] An embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the user classification method based on CNN usage log spectrum analysis are implemented.

[0033] An embodiment of the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the user classification method based on CNN usage log spectrum analysis are implemented.

[0034] The embodiment of the present invention provides a user classification method and system based on CNN usage log spectrum analysis. When a user classification request is detected, the target user and request target corresponding to the user classification request are obtained, the corresponding request time period is determined based on the request target, and the target historical data in the historical log data of the target user is obtained based on the request time period; the corresponding request factor is determined according to the request target, the user classification data in the target historical data is determined based on the request factor, and the data time corresponding to the user classification data is obtained. Based on the user classification data and the corresponding data time, a time series corresponding to the target user is generated; the time series is divided into a time series set according to the time interval of the data time, and the time series set is standardized to obtain a standardized time series set, and the time series set is converted into a corresponding spectrum graph through Fourier transform; a corresponding CNN convolutional neural network model is built based on the request purpose and request factor, the spectrum graph is input into the CNN convolutional neural network model for classification training, and the classification result of the target user is determined based on the classification training result. In this way, the user log data analysis can be transferred from the time domain to the frequency domain for analysis, thereby avoiding the loss of time information and further improving the accuracy of the classification effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] FIG1 is a flow chart of a user classification method based on CNN usage log spectrum analysis according to an embodiment of the present invention;

[0037] FIG2 is a structural diagram of a time series to frequency spectrum diagram according to an embodiment of the present invention;

[0038] FIG3 is a flow chart of a CNN convolutional neural network model classification training according to an embodiment of the present invention;

[0039] FIG4 is a structural diagram of a user classification system based on CNN usage log spectrum analysis according to an embodiment of the present invention;

[0040] FIG5 is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0042] Figure 1 is a flow chart of a user classification method based on CNN usage log spectrum analysis provided by an embodiment of the present invention. As shown in Figure 1, an embodiment of the present invention provides a user classification method based on CNN usage log spectrum analysis: Step S101, when a user classification request is detected, the target user and request target corresponding to the user classification request are obtained, the corresponding request time period is determined based on the request target, and the target historical data in the historical log data of the target user is obtained based on the request time period.

[0043] Specifically, after receiving a user classification request to classify historical users in a database, the target user and request target corresponding to the user classification request are determined, wherein the target user is the user or user group pointed to by this user classification request, which may include all historical users, or additional filtering conditions may be added, such as all male users, all users under 18 years old, etc. The request target is the request purpose corresponding to the user classification request, for example, the target of user classification may be to make corresponding product recommendations, friend matching and other request targets, and then the corresponding request time period is determined based on the request target, and the target historical data in the historical log data of the target user is obtained based on the request time period, wherein the request time period can be divided according to the data update frequency in the log data, or according to the most recent time when the user's log data was last updated. In the database, user log data is usually sorted according to data time. After receiving the user classification request, the target historical data in the historical log data of the target user is obtained according to the request time period while consuming very little computing resources.

[0044] Step S102: determine the corresponding request factor according to the request target, determine the user classification data in the target historical data based on the request factor, obtain the data time corresponding to the user classification data, and generate a time series corresponding to the target user based on the user classification data and the corresponding data time.

[0045] Specifically, after determining the request target of the user classification request, determine the request factor corresponding to the request target. For example, when the request target is product recommendation, the request factor includes product attributes, such as product type, product price, etc. When the request target is friend matching, the request factor includes friend attributes, such as friend gender, friend age, friend hobbies, etc. Classify the target historical data according to the request factor, determine the corresponding user classification data, and obtain the data time of the log data corresponding to the user classification data. Based on the user classification data and the corresponding data time, generate the time series corresponding to the target user.

[0046] Step S103 , dividing the time series into time series sets according to the time interval of the data time, and standardizing the time series sets to obtain standardized time series sets, and converting the standardized time series sets into corresponding frequency spectrograms through Fourier transform.

[0047] Specifically, the time series is divided into time series sets corresponding to different time periods based on the time interval of the data. Shorter time series sets can meet the requirements of the Fourier transform in the subsequent steps. The time series sets are then normalized. Normalization can be performed using, for example, Z-score normalization. Normalization transforms the data to conform to a distribution with a mean of 0 and a variance of 1. Z-score normalization only transforms the data to a mean of 0 and a variance of 1, without changing the original data distribution. After normalization, the time series set is converted into a corresponding spectrogram using a Fourier transform. The Fourier transform is used to determine the sinusoidal frequency and phase content of the local portion of the signal that varies over time. The Fourier transform process divides the longer signal into shorter segments of equal length and then calculates the Fourier transform on each segment. This reveals the Fourier spectrum of each segment. The changing spectrum is then typically plotted as a function of time, called a spectrogram. After the short-time Fourier transform, as shown in Figure 2, the user time series set is converted from the time domain to the frequency domain. The spectrogram is saved as a user feature map representing the user classification characteristics.

[0048] Step S104: Build a corresponding CNN convolutional neural network model based on the request purpose and request factors, input the spectrum graph into the CNN convolutional neural network model for classification training, and determine the classification result of the target user based on the classification training result.

[0049] Specifically, the training items of the convolutional neural network model are determined based on the request target and the request factor, and the spectrum graph is input into the CNN convolutional neural network model for classification training. The CNN convolutional neural network model performs a convolution operation on the image. The training process of the CNN convolutional neural network model for classification training is shown in Figure 3. The classification result corresponding to the spectrum graph in the classification training result is determined, which is the classification result of the target user.

[0050] In addition, based on the classification results, a classification label corresponding to the target user can be generated, and the corresponding update cycle can be determined based on the request time period. The classification label can then be saved to the historical log data of the corresponding target user, and the classification label can be updated based on the update cycle. This makes it convenient to determine the classification of the corresponding user based on the classification label when a user classification request is received in the future, and to update the user's classification label in a timely manner.

[0051] In addition, in step S101 above, the number of user classification requests containing the request target within a preset period is obtained, and then the priority of the user classification request is determined based on the number of requests. That is, the urgency of the target user classification tendency is determined based on the number of requests, and the corresponding user classification request sequence is determined based on the priority. Then, the priority of the corresponding request target is determined based on the priority in the user classification request sequence. The learning weights of the convolutional layer in the CNN convolutional neural network are adjusted based on the priority of the request target. This can more specifically allocate model training resources, allowing model training resources to focus more on processing higher-priority user classification requests, thereby achieving better user classification results that are more consistent with user characteristics.

[0052] The embodiment of the present invention provides a user classification method based on CNN usage log spectrum analysis, which configures a basic access threshold in a cache queue, and after filtering the input data through the basic access threshold to determine the cached data, calculates the data hit rate of the cached data when a data request is received, and iteratively adjusts the basic access threshold, calculates the adjusted hit rate, and determines a mapping relationship based on the adjusted standard access threshold and the adjusted hit rate; repeats the iterative adjustment process, and adjusts the data attributes of the input data during the iterative adjustment process, combines the Markov model, and determines a mapping relationship between the standard access threshold and the adjusted hit rate under different data attributes; when the target data is received, obtains the data attributes of the target data, and calculates the adjusted hit rate based on the target data; The mapping relationship corresponding to the data attributes of the data determines the target adjustment threshold, and the target data is filtered through the target adjustment threshold to determine the target cache data; the data classification and data scenario corresponding to the target data are obtained, and the selection result and priority sorting result of the target data are determined based on the data classification and data scenario, and the target cache data is selected based on the selection result, and the target cache data after data selection is prioritized according to the priority sorting result to determine the priority cache data after priority sorting; the metadata in the priority cache data is obtained, the throughput corresponding to the metadata is counted, and the metadata in the priority cache data is filtered based on the throughput to determine the final cache data after filtering, and the final cache data is queued into the cache queue. In this way, the optimal cache level for active data can be automatically determined, the object hit rate can be optimized, and the application execution speed can be improved. Through priority sorting and selective allocation, storage elements that can improve performance can be identified and cached. For metadata classification cache, the performance of specific scenarios can be improved.

[0053] In another embodiment, the user classification method based on CNN usage log spectrum analysis further includes:

[0054] Acquire a historical comparison set based on the user classification data, and divide the historical comparison set into a standard set and a verification set according to a preset ratio, wherein the historical comparison set is historical data associated with the user classification data in a database and has been classified;

[0055] Generate a standard spectrum graph corresponding to the standard set data, and input the standard matching graph and the corresponding classification results into the CNN convolutional neural network as standard parameters of the neural network model;

[0056] Obtain the classification training results of the CNN convolutional neural network model, and compare the classification training results with the validation set data. If the comparison result meets the preset accuracy requirements, determine the classification result of the target user corresponding to the classification training.

[0057] In this embodiment, in order to further improve the accuracy of the CNN convolutional neural network model training results, when training the CNN convolutional neural network model, a historical control set is obtained based on the user classification data, and the historical control set is divided into a standard set and a verification set according to a preset ratio, such as a ratio of 8:2. The historical control set is historical data in the database that is associated with the user classification data and has been classified, such as historical data that is the same or similar in type to the user classification data. Then, a standard spectrum graph corresponding to the standard set data is generated, and the standard matching graph and the corresponding classification result are input into the CNN convolutional neural network as standard parameters of the neural network model. The classification training result of the CNN convolutional neural network model is obtained, and the classification training result is compared with the verification set data. If the comparison result meets the preset accuracy requirement, such as the accuracy is greater than 90%, the classification result of the target user corresponding to the classification training is determined. In this way, the accuracy of the classification result can be further improved through the standard set and the data set.

[0058] FIG4 shows a user classification system based on CNN usage log spectrum analysis provided by an embodiment of the present invention, including: a detection module S201, a sequence module S202, a spectrum module S203, and a training module S204, wherein:

[0059] The detection module S201 is used to obtain the target user and request target corresponding to the user classification request when a user classification request is detected, determine the corresponding request time period based on the request target, and obtain the target historical data in the historical log data of the target user based on the request time period.

[0060] The sequence module S202 is used to determine the corresponding request factor according to the request target, determine the user classification data in the target historical data based on the request factor, obtain the data time corresponding to the user classification data, and generate a time series corresponding to the target user based on the user classification data and the corresponding data time.

[0061] The spectrum module S203 is used to divide the time series into time series sets according to the time interval of the data time, standardize the time series sets to obtain standardized time series sets, and convert the standardized time series sets into corresponding spectrum graphs through Fourier transform.

[0062] The training module S204 is used to build a corresponding CNN convolutional neural network model based on the request purpose and request factor, input the spectrum graph into the CNN convolutional neural network model for classification training, and determine the classification result of the target user based on the classification training result.

[0063] In one embodiment, the system further comprises:

[0064] An acquisition module is configured to obtain the number of user classification requests containing the request target within a preset period, determine the priority of the user classification request based on the number of requests, and determine a corresponding user classification request sequence based on the priority. A weighting module is configured to determine the priority of the corresponding request target based on the priority in the user classification request sequence, and adjust the learning weights of the convolutional layer in the CNN convolutional neural network based on the priority of the request target.

[0065] In one embodiment, the system further comprises:

[0066] The label module is used to generate a corresponding classification label based on the classification result and determine a corresponding update period based on the request time period.

[0067] An updating module is used to save the classification label to the historical log data of the corresponding target user and update the classification label based on the update cycle.

[0068] For the specific limitations of the user classification system based on CNN usage log spectrum analysis, please refer to the limitations of the user classification method based on CNN usage log spectrum analysis above, which will not be repeated here. The various modules in the above-mentioned user classification system based on CNN usage log spectrum analysis can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0069] Figure 5 illustrates a schematic diagram of the physical structure of an electronic device. As shown in Figure 5, the electronic device may include: a processor (processor) 301, a memory (memory) 302, a communication interface (Communications Interface) 303 and a communication bus 304, wherein the processor 301, the memory 302, and the communication interface 303 communicate with each other through the communication bus 304. The processor 301 can call the logic instructions in the memory 302 to execute the following method: when a user classification request is detected, the target user and the request target corresponding to the user classification request are obtained, the corresponding request time period is determined based on the request target, and the target historical data in the historical log data of the target user is obtained based on the request time period; the corresponding request factor is determined according to the request target, the user classification data in the target historical data is determined based on the request factor, and the data time corresponding to the user classification data is obtained, and the time series corresponding to the target user is generated based on the user classification data and the corresponding data time; the time series is divided into a time series set according to the time interval of the data time, and the time series set is standardized to obtain the standardized time series set, and the time series set is converted into a corresponding spectrum graph through Fourier transform; the corresponding CNN convolutional neural network model is built based on the request purpose and the request factor, the spectrum graph is input into the CNN convolutional neural network model for classification training, and the classification result is determined based on the classification training result.

[0070] In addition, the logic instructions in the above-mentioned memory 302 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0071] On the other hand, an embodiment of the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the transmission method provided by the above-mentioned embodiments, for example, including: when a user classification request is detected, obtaining the target user and request target corresponding to the user classification request, determining the corresponding request time period based on the request target, and obtaining the target historical data in the historical log data of the target user based on the request time period; determining the corresponding request factor according to the request target, determining the user classification data in the target historical data based on the request factor, and obtaining the data time corresponding to the user classification data, and generating a time series corresponding to the target user based on the user classification data and the corresponding data time; dividing the time series into a time series set according to the time interval of the data time, and standardizing the time series set to obtain the standardized time series set, and converting the time series set into a corresponding spectrum graph through Fourier transform; building a corresponding CNN convolutional neural network model based on the request purpose and request factor, inputting the spectrum graph into the CNN convolutional neural network model for classification training, and determining the classification result based on the classification training result.

[0072] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0073] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A user classification method based on CNN usage log spectrum analysis, characterized in that: include: When a user classification request is detected, a target user and a request target corresponding to the user classification request are obtained, a corresponding request time period is determined based on the request target, and target historical data in the historical log data of the target user is obtained based on the request time period; Determine a corresponding request factor according to the request target, determine user classification data in the target historical data based on the request factor, obtain data time corresponding to the user classification data, and generate a time series corresponding to the target user based on the user classification data and the corresponding data time; Dividing the time series into time series sets according to the time interval of the data time, and standardizing the time series sets to obtain standardized time series sets, and converting the standardized time series sets into corresponding frequency spectrograms through Fourier transform; A corresponding CNN convolutional neural network model is built based on the request purpose and request factors, the spectrum graph is input into the CNN convolutional neural network model for classification training, and the classification result of the target user is determined based on the classification training result.

2. The user classification method based on CNN usage log spectrum analysis according to claim 1 is characterized in that: The method further comprises: Obtaining the number of user classification requests containing the request target within a preset period, determining the priority of the user classification request based on the number of requests, and determining a corresponding user classification request sequence based on the priority; The priority of the corresponding request target is determined based on the priority in the user classification request sequence, and the learning weight of the convolution layer in the CNN convolutional neural network is adjusted based on the priority of the request target.

3. The user classification method based on CNN usage log spectrum analysis according to claim 1 is characterized in that: The method further comprises: Based on the classification result, a corresponding classification label is generated, and based on the request time period, a corresponding update period is determined; The classification label is saved to the historical log data of the corresponding target user, and the classification label is updated based on the update cycle.

4. The user classification method based on CNN usage log spectrum analysis according to claim 1 is characterized in that: The step of inputting the spectrum graph into a CNN convolutional neural network model for classification training comprises: Acquire a historical comparison set based on the user classification data, and divide the historical comparison set into a standard set and a verification set according to a preset ratio, wherein the historical comparison set is historical data in a database that is associated with the user classification data and has been classified; Generate a standard spectrum graph corresponding to the standard set data, and input the standard matching graph and the corresponding classification result into the CNN convolutional neural network as standard parameters of the neural network model; Obtain the classification training result of the CNN convolutional neural network model, and compare the classification training result with the validation set data. If the comparison result meets the preset accuracy requirement, determine the classification result of the target user corresponding to the classification training.

5. The user classification method based on CNN usage log spectrum analysis according to claim 1 is characterized in that: The method further comprises: The standardization is Z-score standardization.

6. A user classification system based on CNN usage log spectrum analysis, characterized in that: The system comprises: A detection module, configured to, when a user classification request is detected, obtain a target user and a request target corresponding to the user classification request, determine a corresponding request time period based on the request target, and obtain target historical data in the historical log data of the target user based on the request time period; A sequence module, used to determine a corresponding request factor according to the request target, determine user classification data in the target historical data based on the request factor, obtain a data time corresponding to the user classification data, and generate a time series corresponding to the target user based on the user classification data and the corresponding data time; A spectrum module, used for dividing the time series into time series sets according to the time interval of the data time, and standardizing the time series sets to obtain standardized time series sets, and converting the standardized time series sets into corresponding spectrum graphs through Fourier transform; A training module is used to build a corresponding CNN convolutional neural network model based on the request purpose and request factors, input the spectrum graph into the CNN convolutional neural network model for classification training, and determine the classification result of the target user based on the classification training result.

7. The user classification system based on CNN usage log spectrum analysis according to claim 6 is characterized in that: The system further comprises: an acquisition module, configured to acquire the number of user classification requests containing the request target within a preset period, determine the priority of the user classification request based on the number of requests, and determine a corresponding user classification request sequence based on the priority; A weight module is used to determine the priority of the corresponding request target based on the priority in the user classification request sequence, and adjust the learning weight of the convolution layer in the CNN convolutional neural network based on the priority of the request target.

8. The user classification system based on CNN usage log spectrum analysis according to claim 6 is characterized in that: The system further comprises: A label module, used to generate a corresponding classification label based on the classification result, and determine a corresponding update period based on the request time period; An updating module is used to save the classification label to the historical log data of the corresponding target user, and to update the classification label based on the updating cycle.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the user classification method based on CNN usage log spectrum analysis as described in any one of claims 1 to 5 are implemented.

10. A non-transitory 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 user classification method based on CNN usage log spectrum analysis as described in any one of claims 1 to 5 are implemented.

Images