System and method for mobile traffic prediction based on mobile traffic pattern extraction and clustering

Abstract

The present invention relates to a technology for collecting a traffic amount by time for a base station and converting the traffic amount into an image so as to identify a pattern of mobile network traffic generation amounts, wherein a system for predicting the amount of mobile traffic according to an embodiment of the present invention includes: a data collection unit for collecting time-series data as a traffic amount by time of base stations in a mobile network; an image conversion unit for converting the collected time-series data into an image including a generation pattern according to the traffic amount by time; a clustering processing unit for clustering the converted image on the basis of the generation pattern to form at least one group of base stations; and a traffic prediction unit for predicting a traffic amount for each formed group through an artificial intelligence model.

Description

Mobile traffic prediction system and method based on mobile traffic pattern extraction and clustering.

[0001] The present invention is a technology for collecting the hourly traffic volume of a base station, converting it into an image, and identifying the pattern of mobile traffic generation volume. The technology utilizes an autoencoder to reduce the irregularity of traffic, while at the same time forming a cluster by gathering base stations showing similar pattern characteristics, and then predicting the mobile traffic volume for each clustered group, thereby improving the performance of mobile traffic volume prediction.

[0002] Mobile network traffic refers to the amount of traffic generated when multiple people access mobile networks using mobile devices like mobile phones and tablets. Traffic processed on mobile networks, such as 3G / 4G / 5G, is processed through base stations installed by mobile carriers, and the total amount of this data is referred to as base station mobile traffic.

[0003] Mobile network traffic volumes exhibit different patterns depending on the regional characteristics of the base station. For example, a base station located in an office might see higher traffic during working hours and lower traffic during non-working hours like evenings and weekends. However, a base station located in a shopping mall might see higher traffic on weekends than weekdays. In this way, mobile network traffic differs from time-series data such as electricity and voice signals in that it exhibits regional and time-specific characteristics.

[0004] Mobile network traffic must also consider the time of day it was measured. For example, even with the same pattern, regional characteristics can differ depending on whether it occurs in the evening or morning, or on a weekday or weekend. However, time-series data such as electricity, voice, and vibration analyze only the data's patterns and shapes. This means that mobile network traffic differs from typical time-series data, which doesn't consider the time of measurement.

[0005] Furthermore, mobile traffic differs in that, rather than processing traffic volume over time, it is predicted by summing the traffic generated within a certain timeframe. For example, in the present invention, traffic is collected hourly and then classified and predicted. Because signals from multiple users converge at a base station, predicting the instantaneous data volume itself is not feasible. Instead, data over a certain period of time is aggregated for analysis and prediction.

[0006] Data on mobile network traffic from a single base station may not be sufficient to predict traffic volume. However, using data from all base stations can result in poor prediction accuracy, as training and prediction are based on data from base stations with different traffic patterns.

[0007] Current methods for predicting traffic volumes involve learning all mobile network traffic volumes through an AI model. However, this method requires learning all mobile network traffic volume patterns, making the AI ​​model heavy and requiring long training times. Furthermore, the characteristics of the predicted mobile network traffic volume patterns can be influenced by other mobile network traffic volume patterns, degrading prediction performance.

[0008] The purpose of the present invention is to extract patterns of mobile traffic generation from numerous base stations, perform clustering according to classification, standardize mobile traffic generation patterns, and make predictions based on the extracted patterns.

[0009] The present invention aims to obtain both accuracy and efficiency of prediction by forming a cluster by grouping base stations with similar traffic characteristics such as traffic generation patterns and training using mobile network traffic data for each cluster.

[0010] The purpose is to further improve clustering accuracy by visualizing time series data to classify traffic characteristics, and applying an autoencoder to classify them to enhance the characteristics of these images and reduce irregularity.

[0011] A mobile traffic volume prediction system according to one embodiment may include a data collection unit that collects time series data as the hourly traffic volume of base stations in a mobile network, an image conversion unit that converts the collected time series data into an image including the hourly traffic occurrence pattern, a clustering processing unit that clusters the converted image based on the occurrence pattern to form at least one group of base stations, and a traffic prediction unit that predicts the traffic volume for each formed group through each artificial intelligence model.

[0012] According to one embodiment, the image conversion unit may convert the traffic amount by base station into an image after normalizing the traffic amount by base station in order to reduce the irregularity of traffic and to analyze the characteristics of the traffic generation pattern because the traffic amount itself is different by base station.

[0013] The clustering processing unit according to one embodiment can generate an output image that reinforces an occurrence pattern of the image using an encoder of an autoencoder, reduces traffic irregularity, and classify the generated output image into at least one group using a clustering algorithm.

[0014] The clustering processing unit according to one embodiment learns the hourly traffic volume data of the imaged base stations through the autoencoder, and can cluster the data using a K-means clustering algorithm or the like by utilizing the encoder result value of the learned autoencoder to classify the data into at least one group.

[0015] The clustering processing unit according to one embodiment can classify cluster groups according to regional characteristics, including at least one group among offices, residences, roads, shopping malls, commercial areas, and airports.

[0016] According to one embodiment, the image conversion unit may convert the time series data into an image using at least one algorithm among methods such as Recurrence Plot, Markov Transition Field, and Gramian Angular Field.

[0017] An operating method of a mobile traffic volume prediction system according to an embodiment may include a step of collecting time series data as hourly traffic volumes of base stations in a mobile network, a step of converting the collected time series data into an image including an occurrence pattern according to the hourly traffic volume after normalizing the collected time series data, a step of forming at least one group of base stations by clustering the converted image based on the occurrence pattern, and a step of predicting the traffic volume for each formed group.

[0018] The step of forming at least one group of base stations by clustering the transformed image based on the occurrence pattern according to one embodiment may include the step of strengthening the occurrence pattern of the image using an encoder of an autoencoder and generating an output image with reduced traffic irregularity, and the step of classifying the generated output image into at least one group using a clustering algorithm.

[0019] The step of generating an output image that reinforces the occurrence pattern of the image and reduces the irregularity of traffic using the encoder of the autoencoder according to one embodiment includes the step of learning the hourly traffic volume data of the imaged base stations through the autoencoder, and the step of classifying the generated output image into at least one group using a clustering algorithm may include the step of clustering using the result value of the encoder of the learned autoencoder using a K-means clustering algorithm or the like.

[0020] The step of classifying into at least one group using a clustering algorithm according to an embodiment may include a step of classifying into cluster groups according to regional characteristics, including at least one group among offices, residences, roads, shopping malls, commerce, and airports.

[0021] The step of converting the collected time series data according to one embodiment into an image including an occurrence pattern according to the amount of traffic by time may include a step of converting the time series data into an image using at least one algorithm among Recurrence Plot, Markov Transition Field, and Gramian Angular Field.

[0022] According to one embodiment, the mobile network traffic generation pattern of numerous base stations can be classified and clustered to standardize the mobile network traffic generation pattern and make predictions based on the same.

[0023] In one embodiment, both prediction accuracy and efficiency can be achieved by training using mobile network traffic data from base stations with similar traffic generation patterns.

[0024] Figure 1 is a drawing explaining the entire process according to the present invention.

[0025] Figure 2 is a diagram illustrating a clustered pattern.

[0026] FIG. 3 is a diagram illustrating a mobile traffic volume prediction system according to an embodiment.

[0027] Figure 4 is a diagram explaining the operation of an autoencoder.

[0028] Figure 5 is a diagram explaining the operation of training and predicting for each group compared to training and predicting with all data.

[0029] Figure 6 is a diagram explaining the LSTM prediction results trained with the entire data.

[0030] Figure 7 is a diagram explaining the LSTM prediction results trained with road data.

[0031] Figure 8 is a diagram showing the clustering results.

[0032] Figure 9 is a diagram comparing clustering by time series and clustering by the method of the present invention.

[0033] Figure 10 is a diagram showing COEX base station traffic (left) and a cluster tied to offices (right).

[0034] Figure 11 is a diagram explaining the analysis of clustering results.

[0035] Fig. 12 is a drawing explaining an operation method of a mobile traffic volume prediction system according to an embodiment.

[0036] Specific structural or functional descriptions of embodiments according to the concept of the present invention disclosed in this specification are merely illustrative for the purpose of explaining embodiments according to the concept of the present invention, and embodiments according to the concept of the present invention may be implemented in various forms and are not limited to the embodiments described in this specification.

[0037] Embodiments according to the concept of the present invention may have various modifications and take various forms, and thus, embodiments are illustrated in the drawings and described in detail in this specification. However, this is not intended to limit embodiments according to the concept of the present invention to specific disclosed forms, but rather includes modifications, equivalents, or alternatives that fall within the spirit and technical scope of the present invention.

[0038] While terms such as "first" or "second" may be used to describe various components, these components should not be limited by these terms. These terms are intended solely to distinguish one component from another. For example, a first component may be referred to as a "second component," and similarly, a second component may also be referred to as a "first component," without departing from the scope of the invention.

[0039] When a component is referred to as being "connected" or "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but that there may be other components in between. Conversely, when a component is referred to as being "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between. Expressions that describe relationships between components, such as "between," "immediately between," or "directly adjacent to," should be interpreted similarly.

[0040] The terminology used herein is for the purpose of describing specific embodiments only and is not intended to limit the present invention. The singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, it should be understood that the terms "comprises" or "has" are intended to specify the presence of a described feature, number, step, operation, component, part, or combination thereof, but do not preclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0041] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and will not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0042]

[0043] Hereinafter, embodiments will be described in detail with reference to the attached drawings. However, the scope of the patent application is not limited or restricted by these embodiments. The same reference numerals in each drawing represent the same components.

[0044] Figure 1 is a drawing explaining the entire process according to the present invention.

[0045] The present invention is a technology that collects base station traffic volumes over time, converts them into images, and identifies patterns in mobile network traffic generation. Furthermore, it utilizes an autoencoder to reduce traffic irregularity and cluster base stations exhibiting similar pattern characteristics. By then predicting mobile network traffic volume for each clustered group, the invention can enhance the performance of mobile network traffic prediction.

[0046] To this end, the present invention collects hourly traffic volume data (110) and visualizes the traffic volume data using Gramian Angular Field (120). The visualized traffic volume data is trained (130) using an autoencoder, and the encoder results of the trained autoencoder are utilized to cluster the data using a K-means clustering algorithm (140). Furthermore, performance can be improved by predicting the clustered groups using a separate artificial intelligence model (150).

[0047] This allows us to classify and cluster mobile network traffic patterns across numerous base stations, formalizing and predicting these patterns. Furthermore, by training using mobile network traffic data from base stations with similar traffic patterns, we can achieve both prediction accuracy and efficiency.

[0048] The present invention classifies and clusters patterns of mobile network traffic generation from numerous base stations to standardize mobile network traffic generation patterns and predict traffic volume based on the patterns.

[0049] Image processing refers to converting time-series data into images. Converting time-series data into images offers the advantage of facilitating at-a-glance identification of patterns and classification.

[0050] In the present invention, as an example of visualizing time-series data, time-series data representing the amount of mobile traffic per hour at a base station is converted into an image. Depending on the prediction cycle, the time can be set to 10 minutes, 30 minutes, one hour, or two hours.

[0051] Figure 2 is a drawing illustrating a clustered pattern (200).

[0052] We clustered using actual mobile network traffic data from mobile carriers, and analyzed and investigated each cluster group to confirm that the cluster groups were divided according to regional characteristics such as offices, residences, roads, shopping malls, and commercial areas around the base station.

[0053] In addition, we confirmed that the prediction performance of the model predicted using data targeting the mobile traffic volume of the entire base station without grouping was superior to that of the model predicted using data targeting the mobile traffic volume of the entire base station without grouping.

[0054] To view the patterns of mobile network traffic, the changing patterns of traffic are more important than the absolute traffic volume, so the traffic volume per base station can be normalized and then visualized.

[0055] Methods for converting time-series data into images include recurrence plots, Markov transition fields, and Gramian angular fields (GAFs). In this invention, GAFs were used for classification. However, the same method as in this invention can be applied to other image conversion methods.

[0056] FIG. 3 is a diagram illustrating a mobile traffic volume prediction system (300) according to one embodiment.

[0057] A mobile traffic volume prediction system (300) according to one embodiment classifies and clusters mobile network traffic volume patterns from numerous base stations, standardizes mobile network traffic generation patterns, and predicts based on these patterns. Furthermore, by training using mobile network traffic data from base stations with similar traffic generation patterns, the system can achieve both the necessary data volume and accuracy for prediction.

[0058] To this end, a mobile traffic volume prediction system (300) according to one embodiment may include a data collection unit (310), an image conversion unit (320), a clustering processing unit (330), a traffic prediction unit (340), and a control unit (350).

[0059] A data collection unit (310) according to one embodiment collects time series data as the hourly traffic volume of base stations in a mobile network.

[0060] An image conversion unit (320) according to one embodiment converts the collected time series data into an image including an occurrence pattern according to the amount of traffic per hour.

[0061] The image conversion unit (320) can convert the traffic amount by base station into an image after normalizing the traffic amount by base station to reduce the irregularity of the traffic, since the traffic amount itself may be different for each base station.

[0062] For example, the image conversion unit (320) can convert the time series data into an image using at least one algorithm among methods such as Recurrence Plot, Markov Transition Field, and Gramian Angular Field.

[0063] These algorithms transform time-series data into visual images, helping analyze and interpret mobile network traffic patterns. Recurrence plots are used to visualize recurring patterns in time-series data, while Markov Transition Fields use Markov chains to represent data transition patterns as images. Finally, Gramian Angular Fields transform time-series data into angular information to visually represent relative patterns between data points. These image transformation algorithms enable better understanding and prediction of mobile traffic data.

[0064] A clustering processing unit (330) according to one embodiment clusters base stations based on occurrence patterns for converted images to form at least one group.

[0065] This clustering process can be performed using images that visually analyze the network traffic patterns of each base station. During this process, base stations with similar patterns are grouped into clusters, allowing for the identification of diverse traffic patterns and characteristics occurring on the network. This clustering allows for a more efficient understanding of diverse mobile traffic patterns, enabling prediction and management accordingly.

[0066] The clustering processing unit (330) uses the encoder of the autoencoder to enhance the characteristics of the occurrence pattern for the image, generates an output image with reduced traffic irregularity, and classifies the generated output image into at least one group using a clustering algorithm.

[0067] In this process, an autoencoder compresses the input image and transforms it into a latent representation to highlight traffic patterns, and a clustering algorithm forms similar groups based on these enhanced traffic patterns. This allows for efficient understanding of the diverse traffic patterns occurring on the network and enables prediction and management based on these patterns.

[0068] The clustering processing unit (330) learns the hourly traffic volume data of imaged base stations through an autoencoder, and can cluster the data using a K-means clustering algorithm or the like by utilizing the encoder result value of the learned autoencoder to classify the data into at least one group.

[0069] An autoencoder transforms base station traffic data into characteristic images, and the K-means algorithm groups base stations with similar traffic patterns. This allows for efficient understanding of the diverse traffic patterns occurring in mobile networks and clustering them into relevant groups.

[0070] The clustering processing unit (330) can classify cluster groups according to regional characteristics, including at least one group among offices, residences, roads, shopping malls, commerce, and airports.

[0071] These groups are formed based on the environment and circumstances in which each base station is located. For example, office and commercial areas are expected to see high traffic primarily during the day, while residential areas will have more stable traffic patterns at night. Base stations near roads will experience greater variability due to mobile users, and airports will typically see high traffic during specific times. These clustering groups based on regional characteristics allow for a better understanding and prediction of traffic patterns in each area.

[0072] A traffic prediction unit (340) according to an embodiment predicts the amount of traffic for each formed group through each artificial intelligence model.

[0073] These models provide improved predictions by considering the characteristics and patterns of a given group. For example, Long Short-Term Memory (LSTM) neural networks, which handle time-series data, can be effective in predicting traffic by considering temporal patterns. Furthermore, Fast Fourier Transform (FFT) and frequency-domain neural network models, which are used to extract and predict features in the frequency domain, can also be used. By leveraging these various models to predict traffic for each group, network resources can be optimized and service quality improved.

[0074] The control unit (350) according to one embodiment can be interpreted as a central processing unit (CPU) and can perform various operations and process data within the system.

[0075] In particular, the control unit (350) can read commands from memory, interpret and execute the commands, and can also perform arithmetic operations such as addition, subtraction, multiplication, and division.

[0076] In addition, the control unit (350) can handle data storage and retrieval, and can also perform the function of reading data from memory and storing the results of performing operations back into memory.

[0077] In addition, the control unit (350) can manage the execution flow of the program, and in particular, can control the flow of the program using commands such as conditional statements (if-else) or iterative statements (for, while). In addition, the control unit (350) can have a small and fast memory device called a register placed inside, and this register can be used to temporarily store data or perform operations.

[0078] The control unit (350) can process and take appropriate action when an external event or exceptional situation occurs, and can quickly access data and instructions by using cache memory that is faster than the main memory.

[0079] In addition, the control unit (350) can use a system bus to communicate with memory or input / output devices, and can provide various power management functions to minimize power consumption.

[0080] Figure 4 is a diagram explaining the operation of an autoencoder.

[0081] Since the irregularity of the data is too great and it is difficult to standardize it to directly image the time series data and classify its characteristics, an autoencoder (400) can be applied in the present invention to extract the characteristics of the data.

[0082] An autoencoder (400) is a technology primarily applied to unsupervised training artificial neural network models, and consists of an encoder and a decoder. The encoder compresses input data, and the decoder learns to restore the compressed data to the original input data. However, the present invention extracts traffic patterns using only the encoder to reduce traffic irregularity and extract similar features. The extracted patterns have clearer features and less irregularity, which can improve classification and clustering performance.

[0083] In the present invention, patterns of mobile network traffic data converted into images are extracted using an encoder in an autoencoder (400). Then, based on the extracted patterns, mobile network traffic data to be clustered is input to the autoencoder (400), resulting in mobile network traffic data (images) with enhanced patterns and reduced randomness. Clustering images using a clustering algorithm based on the resulting mobile network traffic data (images) allows for accurate clustering based on the characteristics of mobile network traffic volume while being minimally affected by randomness. By subsequently utilizing an artificial intelligence model for each clustered group, the prediction performance of mobile network traffic can be enhanced.

[0084] Methods for clustering time-series data, including mobile network traffic volume data, include using correlation values ​​between time-series data, clustering based on the amount of change in data values, or clustering geographically close base stations. However, these methods often struggle to identify individual time-series data patterns. Furthermore, similar patterns can be clustered into completely different groups due to randomness at each base station or time-axis shift. While issues caused by time-axis shift can be mitigated with Dynamic Time Warping (DTW), it remains vulnerable to noise and randomness.

[0085] Figure 5 is a diagram explaining the operation of training and predicting for each group compared to training and predicting with all data.

[0086] Drawing symbol 510 illustrates a structure for training and predicting with all data, and drawing symbol 520 illustrates a structure for training and predicting for each clustered group.

[0087] An example of a result of training and prediction with all data through the structure of drawing number 510 is illustrated in FIG. 6, and an example of a result of training and prediction with all data through the structure of drawing number 520 is illustrated in FIG. 7.

[0088] Figure 6 shows the prediction results of LSTM trained with the entire data.

[0089] Figure 6 shows the mobile network traffic volume prediction results of an artificial intelligence model trained with the entire mobile network traffic volume data.

[0090] In the case of Figure 6, the entire mobile network traffic volume data was trained using an AI model. The trained model was then asked to predict one of the mobile network traffic volume data points, which included roads located around a base station. The AI ​​model was asked to predict 20 values ​​following the input time series. The red line represents the predicted result of the AI ​​model, and the blue line represents the actual value.

[0091] Figure 7 is a diagram illustrating the LSTM prediction results trained on data from base stations belonging to clusters classified by road data.

[0092] Figure 7 shows the mobile network traffic volume prediction results of an artificial intelligence model trained with mobile network traffic volume data of one cluster (group), targeting traffic data measured through a base station adjacent to a road among each clustered group.

[0093] In the case of Fig. 7, unlike Fig. 6, the entire mobile network traffic volume data was not used, but the mobile network traffic volume data of the cluster (group) to which the mobile network traffic volume data to be predicted belongs among the clusters (groups) divided according to the present invention was trained.

[0094] The AI ​​model and parameters are the same as those used in Figure 6, but the training data is different. The AI ​​model was designed to predict 20 values ​​following the input time series. The red line represents the model's predicted results, while the blue line represents the actual values.

[0095] The first graph in Fig. 6 and the first graph in Fig. 7 represent mobile network traffic volume data from the same base station. Similarly, the second graph in Fig. 6 and the second graph in Fig. 7 represent mobile network traffic volume data from the same base station. Although they are time series data for the same mobile network base station traffic volume, the time series prediction result in Fig. 7 predicted the traffic volume pattern more accurately than the time series prediction result in Fig. 6.

[0096] Figure 8 is a diagram showing the clustering results.

[0097] Shows downtown base station traffic (left) and a cluster of offices (right).

[0098] Drawing symbol 810 can be interpreted as a local area serving the first floor of a downtown office, for example, Base Station 2 / Base Station 1. Therefore, traffic from security personnel working weekends on the first floor and external traffic generated by signals going outside can be expected.

[0099] Drawing symbol 820 can be interpreted as a pattern according to the autoencoder output for drawing symbol 810.

[0100] Figure 9 is a drawing (900) that compares and explains the case of clustering by time series and the case of clustering by the method of the present invention.

[0101] Drawing symbol 910 corresponds to a pattern clustered by the method of the present invention.

[0102] Figure 10 is a drawing (1000) showing COEX base station traffic (left) and a cluster tied to offices (right).

[0103] Drawing symbol 1010 can be interpreted as station number 1, for example, a base station located at an exhibition hall. Drawing symbol 1020 can be interpreted as station number 4, which is difficult to identify, drawing symbol 1030 can be interpreted as station number 5, which is a city airport terminal, and drawing symbol 1040 can be interpreted as station number 11, which is located at an exhibition hall.

[0104] Figure 11 is a drawing (1100) explaining the analysis of clustering results.

[0105] Drawing symbol 1110 corresponds to the result of predicting mobile network traffic volume for an office by an artificial intelligence model trained with total mobile network traffic volume data.

[0106] Drawing code 1120 collects hourly traffic volume data for an office, visualizes the traffic volume data using a Gramian Angular Field, and shows the results of learning the visualized traffic volume data through an autoencoder.

[0107] Analysis of drawing number 1110 shows that the peak traffic volume occurs at 12 noon, traffic increases and decreases rapidly, and there is almost no traffic on weekends.

[0108] Fig. 12 is a drawing explaining an operation method of a mobile traffic volume prediction system according to an embodiment.

[0109] A method of operating a mobile traffic volume prediction system according to an embodiment can collect time series data as hourly traffic volumes of base stations in a mobile network (step 1201).

[0110] The method of operating a mobile traffic volume prediction system according to an embodiment can convert collected time series data into an image containing an occurrence pattern according to hourly traffic volume after normalizing it (step 1202).

[0111] According to an embodiment of the present invention, a method for operating a mobile traffic volume prediction system can form at least one group of base stations by clustering converted images based on occurrence patterns (step 1203). To this end, an encoder of an autoencoder can be used to enhance occurrence patterns of the image, generate output images with reduced traffic irregularity, and classify the generated output images into at least one group using a clustering algorithm. In particular, to generate output images with reduced traffic irregularity, hourly traffic volume data of base stations imaged through an autoencoder can be learned. Furthermore, in order to classify the images into at least one group, the output values ​​of the learned autoencoder's encoder can be utilized to cluster the images using a K-means clustering algorithm.

[0112] For example, clusters can be grouped based on regional characteristics, including at least one group among office, residential, road, shopping mall, commercial, and airport.

[0113] The method of operating a mobile traffic volume prediction system according to an embodiment can predict the traffic volume for each formed group (step 1204).

[0114] First, base stations are clustered by group to identify the traffic patterns of each group. Furthermore, traffic volume can be predicted for each group using a corresponding AI model, with the prediction model selected based on the group's characteristics and patterns.

[0115] The traffic volume prediction results for each group can be used to predict mobile traffic volume and provide it to the system.

[0116] You can monitor your mobile network in real time based on predicted traffic volumes and adjust resources as needed to optimize network performance.

[0117] The system can cluster base stations into various groups, such as offices, residential areas, roads, shopping malls, and airports, based on factors such as base station location, surrounding environment, and user traffic patterns. Each group can consist of base stations with similar traffic patterns.

[0118] For example, for office groups, traffic is expected to be high primarily during rush hour, so an LSTM neural network can be used to predict traffic volume for that group. For residential groups, traffic patterns are expected to be stable at night, so a simple linear regression model can be used to predict traffic volume.

[0119] By synthesizing the traffic volume prediction results for each group, the system predicts the mobile traffic volume for a specific region or the entire network.

[0120] Mobile networks can be monitored in real time based on predicted traffic volumes and resources can be adjusted as needed to optimize network performance. For example, if traffic exceeds expectations, additional bandwidth can be allocated or the priorities of specific base stations can be adjusted accordingly.

[0121] In the present invention, different artificial intelligence models can be applied to each cluster.

[0122] Through this method, the mobile traffic volume prediction system can operate efficiently to optimize network resources and provide the best service to users.

[0123]

[0124] The devices described above may be implemented as hardware components, software components, and / or a combination of hardware components and software components. For example, the devices and components described in the embodiments may be implemented using one or more general-purpose computers or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing instructions and responding to them. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing device is sometimes described as being used alone; however, one of ordinary skill in the art will recognize that the processing device may include multiple processing elements and / or multiple types of processing elements. For example, a processing unit may include multiple processors, or a processor and a controller. Other processing configurations, such as parallel processors, are also possible.

[0125] Software may include a computer program, code, instructions, or a combination of one or more of these, which may configure a processing device to perform a desired operation or may, independently or collectively, command the processing device. The software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave, for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on one or more computer-readable recording media.

[0126] The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program commands, data files, data structures, etc., alone or in combination. The program commands recorded on the medium may be those specially designed and configured for the embodiment or may be those known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program commands, such as ROMs, RAMs, and flash memories. Examples of the program commands include not only machine language codes generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiment, and vice versa.

[0127] Although the embodiments described above have been described with limited drawings, those skilled in the art will recognize that various modifications and variations can be made based on the above description. For example, appropriate results can still be achieved even if the described techniques are performed in a different order than described, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different manner than described, or are replaced or substituted with other components or equivalents.

[0128] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims described below.

Claims

1. A data collection unit that collects time series data as the hourly traffic volume of base stations in a mobile network; An image conversion unit that converts the collected time series data into an image containing an occurrence pattern according to the amount of traffic per hour; A clustering processing unit that clusters the base stations based on the occurrence pattern for the converted image to form at least one group; and A traffic prediction unit that predicts the amount of traffic for each group formed above using each artificial intelligence model. A mobile traffic volume prediction system including:

2. In paragraph 1, The above image conversion unit, A mobile traffic volume prediction system characterized by normalizing the traffic volume by base station and then converting it into an image to reduce traffic irregularity and identify traffic occurrence patterns rather than differences in the size of traffic volume.

3. In paragraph 1, The above clustering processing unit, A mobile traffic volume prediction system characterized in that it uses an encoder of an autoencoder to strengthen the occurrence pattern of the image, generates an output image with reduced traffic irregularity, and classifies the generated output image into at least one group using a clustering algorithm.

4. In paragraph 3, The above clustering processing unit, Learn the hourly traffic volume data of the imaged base stations through the above autoencoder, To classify into at least one group as above, A mobile traffic volume prediction system characterized in that the encoder output value of the learned autoencoder is used for clustering using a K-means clustering algorithm.

5. In paragraph 3, The above clustering processing unit, A mobile traffic volume prediction system characterized in that it classifies into cluster groups according to regional characteristics, including at least one group from office, residential, road, shopping mall, commercial, and airport.

6. In paragraph 1, The above image conversion unit, A mobile traffic volume prediction system characterized in that the time series data is converted into an image using at least one algorithm among Recurrence Plot, Markov Transition Field, and Gramian Angular Field.

7. A step of collecting time series data as the hourly traffic volume of base stations in a mobile network; A step of converting the collected time series data into an image containing an occurrence pattern according to the amount of traffic per hour after normalizing it; A step of forming at least one group of base stations by clustering the converted image based on the occurrence pattern; and Step of predicting the amount of traffic for each group formed above A method of operating a mobile traffic volume prediction system including:

8. In paragraph 7, The step of forming at least one group of base stations by clustering the above-mentioned converted image based on the above-mentioned occurrence pattern is as follows: A step of generating an output image that enhances the occurrence pattern of the image and reduces the irregularity of traffic using an encoder of an autoencoder; and A step of classifying the above generated output image into at least one group using a clustering algorithm. A method of operating a mobile traffic volume prediction system including:

9. In paragraph 8, The step of generating an output image that enhances the occurrence pattern of the image and reduces the irregularity of traffic using the encoder of the above autoencoder is as follows. Step of learning the hourly traffic volume data of imaged base stations through an autoencoder Including, For the above generated output image, the step of classifying it into at least one group using a clustering algorithm is as follows: A step of clustering using the K-means clustering algorithm by utilizing the encoder result values ​​of the above-mentioned learned autoencoder. A method of operating a mobile traffic volume prediction system including:

10. In paragraph 8, The step of classifying into at least one group using a clustering algorithm is: A step of classifying into cluster groups based on regional characteristics, including at least one group among offices, residential areas, roads, shopping malls, commercial areas, airports, etc. A method of operating a mobile traffic volume prediction system including:

11. In paragraph 7, The step of converting the collected time series data into an image containing an occurrence pattern according to the amount of traffic by hour is: A step of converting the time series data into an image using at least one algorithm among Recurrence Plot, Markov Transition Field, Gramian Angular Field, etc. A method of operating a mobile traffic volume prediction system including:

12. In paragraph 7, A method of operating a traffic volume prediction system characterized by applying different artificial intelligence models to each cluster.

Images