Communication flow intelligent prediction method based on big data analysis
By decomposing and correlating the communication traffic data, the optimal feature set is constructed, and multiple primary prediction models are used for parallel prediction, the problem of insufficient prediction accuracy in the existing technology is solved, and higher prediction accuracy and reliability are achieved.
Patent Information
- Application Number
- CN202411922523.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-25
- Publication Date
- 2025-05-02
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing communication traffic prediction methods are difficult to adapt to complex and diverse network traffic changes, and lack real-time and dynamics, resulting in insufficient prediction accuracy and accuracy.
The communication traffic intelligent prediction method based on big data analysis is adopted, and the optimal feature set is constructed through the decomposition and correlation processing of trend factors, seasonal factors and periodic factors, and multiple primary prediction models are used for parallel prediction, and the ultimate prediction result of communication traffic is finally integrated to obtain the ultimate prediction result.
Through segmentation processing and feature correlation, the accuracy and reliability of the prediction model are improved, real-time and dynamics are enhanced, and communication traffic can be predicted more accurately.
Smart Images
Figure CN119922090A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of traffic prediction, and in particular relates to a communication traffic intelligent prediction method based on big data analysis. Background Art
[0002] With the rapid development of the Internet and the gradual maturity of IoT technology, the scale of networks, the number of users, and the amount of data transmission are all showing an exponential growth trend. According to recent statistics, the number of Internet users worldwide has exceeded billions, and the number of IoT devices is growing at an alarming rate. The large amount of network data traffic has brought many problems, such as network congestion, low bandwidth utilization, and network security issues. These problems not only affect the user's network experience, but also bring huge challenges to network management and optimization.
[0003] Network traffic forecasting helps network administrators to properly allocate and manage network traffic and ensure network stability and reliability. By accurately predicting future network traffic, network resources can be adjusted in advance to avoid network congestion and bandwidth waste.
[0004] However, with the continuous growth and change of network traffic, data has become more and more complex and diverse. Traditional prediction methods, such as time series analysis and regression analysis, are difficult to adapt to this complexity and diversity. They often need to analyze and process a large amount of historical data, but when the scale of network traffic increases, this method is difficult to meet the needs. Network traffic changes in real time and is affected by many factors, such as user behavior, network topology, service quality, etc. Therefore, the prediction model needs to be real-time and dynamic, and be able to adaptively adjust the prediction model to adapt to different network environments. However, existing prediction methods often lack this real-time and dynamic nature. Network traffic prediction is a complex task that is affected by many factors. Although many prediction methods and techniques have been proposed, their precision and accuracy are still facing challenges. Especially when dealing with large-scale and complex data, the precision and accuracy of the prediction model are more easily affected.
[0005] In summary, although intelligent prediction of communication traffic plays an important role in network management and optimization, there are still many limitations and challenges. In order to solve these problems, it is necessary to continuously explore new prediction methods and technologies, improve the precision and accuracy of prediction models, and enhance their real-time and dynamic nature. At the same time, it is also necessary to pay attention to the updating and development of technology to improve the adaptability and scalability of prediction models. Summary of the invention
[0006] In order to overcome the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a communication traffic intelligent prediction method based on big data analysis.
[0007] In order to achieve the above object, the present invention adopts the following technical solutions:
[0008] The intelligent prediction method of communication traffic based on big data analysis has the following process:
[0009] Step 1: Obtain the communication traffic data of the area to be predicted, perform trend decomposition, seasonal decomposition, and periodic decomposition on the communication traffic data to obtain trend factors, seasonal factors, and periodic factors respectively;
[0010] Step 2: Extract the characteristics of trend factors, including long-term growth trend characteristics and short-term fluctuation characteristics; extract the characteristics of periodic factors, including daily periodicity characteristics, weekly periodicity characteristics, and monthly periodicity characteristics; extract the characteristics of seasonal factors, including holiday effect characteristics and seasonal change characteristics;
[0011] Step 3: Correlation processing is performed on the features of trend factors, seasonal factors, and periodic factors, and feature variables with correlation greater than a set threshold are screened out and integrated into an available feature set, where correlation refers to the correlation between the feature and the target feature of communication traffic prediction; the available feature set refers to the feature set required for communication traffic prediction;
[0012] Step 4: Screen the available feature sets to select the optimal feature set;
[0013] Step 5: input the optimal feature subset into the pre-trained communication traffic first primary prediction model, second primary prediction model and third primary prediction model, and output the first primary prediction result, the second primary prediction result and the third primary prediction result respectively;
[0014] Step 6: Integrate the first primary prediction result, the second primary prediction result and the third primary prediction result to obtain the ultimate prediction result of communication traffic.
[0015] Preferably, the trend factor reflects the overall upward or downward trend of the communication traffic data over a long period of time; the periodic factor reflects the pattern or trend of the communication traffic data that repeats within a fixed time interval; and the seasonal factor reflects the specific pattern or trend of the communication traffic data that changes with the seasons within a year.
[0016] As a preferred method, trend decomposition, seasonal decomposition, and periodic decomposition are as follows:
[0017] The trend factor is calculated using the moving average method. The calculation formula is as follows:
[0018]
[0019] Where, t=T, t=T+1, ..., M, M represents the last time point of the communication flow sequence; A tIt represents the cumulative displacement value at time t in the communication flow sequence; It represents the trend item flow corresponding to time t in the communication flow sequence; T represents the period, and its value is selected according to different time modes;
[0020] Then, separate the period factor C, period factor = communication flow sequence divided by trend factor T;
[0021] After decomposing the trend factor T and the cycle factor C, the trend factor T and the cycle factor C are combined into a sequence TC, which represents the long-term trend and periodic fluctuation in the original time series;
[0022] Finally, the time series data are grouped by month, and the average value of each seasonal unit is calculated. These seasonal average values are divided by the average value of the entire communication traffic series to obtain the seasonal index S.
[0023] As a preference, step three is specifically as follows:
[0024] Substitute the feature variable A and the target feature variable B into the association formula, the formula is as follows:
[0025]
[0026] in, and are the means of feature variable A and target feature variable B respectively, α is the correlation coefficient, and the value range of α is [-1,1]. The larger the |α| is, the stronger the correlation between feature variable A and target feature variable B is. When the |α| value is between 0.8-1.0, the feature variable A and target feature variable B are judged to have an extremely strong correlation; when the |α| value is between 0.6-0.8, the feature variable A and target feature variable B are judged to have a strong correlation; when the |α| value is between 0.4-0.6, the feature variable A and target feature variable B are judged to have a medium correlation; when the |α| value is between 0.2-0.4, the feature variable A and target feature variable B are judged to have a weak correlation; when the |α| value is between 0.0-0.2, the feature variable A and target feature variable B are judged to have a weak correlation or no correlation;
[0027] Features with absolute values of correlation coefficients |α| between 0.6 and 1.0 were selected and integrated to generate a usable feature set.
[0028] As a preference, step four is specifically as follows:
[0029] When there are two or more features in the available feature set that carry almost the same information, the features are marked as redundant features, and the redundant features are removed to screen out the optimal features. The available feature variable set C = (C1, C2, C3, ..., C m ), target feature variable set B = (B1, B2, B3, ..., Bm ); The calculation formula of mutual information I(B;C) is as follows:
[0030]
[0031] Among them, P(B m ,C m ) is the joint distribution probability of the target feature variable set B and the available feature variable set C, P(B m ) and P(C m ) is the marginal distribution probability; mutual information I(B;C) is the joint distribution P(B m , C m ) and marginal distribution P(B m ) and P(C m )'s relative entropy;
[0032] S401, filter out the feature with the largest mutual information value with the target feature in the available feature set through the correlation calculation formula, and obtain the available feature subset. The calculation formula is as follows:
[0033]
[0034] S402, filter out the feature with the smallest mutual information value with the target feature variable in the available feature subset through the redundancy calculation formula, the calculation formula is as follows:
[0035]
[0036] S403. Combine maxD(S,C) and minR(S) by a combination formula to obtain the correlation between the available features and the target features. The combination formula is as follows:
[0037] maxΦ(D,R),Φ=DR;
[0038] S404, sorting the features according to the calculated relevance, ensuring that the features most relevant to the target variable and with the lowest redundancy are placed in front;
[0039] S405, selecting a feature with the greatest correlation and the lowest redundancy with other selected features from the sorted feature list, and adding it to the optimal feature set;
[0040] S406. Repeat steps S401 to S405 until a preset number of features are selected to form an optimal feature set.
[0041] Preferably, the first primary prediction model is a periodic factor prediction model, which is trained based on multiple historical periodic factor data and is used to output a communication traffic prediction periodic factor; the second primary prediction model is a seasonal factor prediction model, which is trained based on multiple historical seasonal factor data and is used to output a communication traffic prediction seasonal factor; the third primary prediction model is a trend factor prediction model, which is trained based on multiple historical trend factor data and is used to output a communication traffic prediction trend factor.
[0042] As a preferred embodiment, the integration process is as follows:
[0043] Corresponding weights d, e and f are assigned to the first primary prediction result, the second primary prediction result and the third primary prediction result respectively, and substituted into the prediction formula. The prediction formula is as follows:
[0044] Final prediction result = (first primary prediction result * d) + (second primary prediction result * e) + (third primary prediction result * f);
[0045] Among them, d+e+f=1, and d, e and f are all greater than 0.
[0046] As a preference, the daily periodicity feature reflects the daily changes in the communication traffic data; the weekly periodicity feature reflects the weekly changes in the communication traffic data; the monthly periodicity feature reflects the monthly changes in the communication traffic data; the long-term growth trend feature reflects that the communication traffic data generally presents a long-term growth trend; the short-term fluctuation feature reflects the short-term fluctuation in the communication traffic data; the holiday effect feature reflects that the communication traffic data shows a significant increase during holidays; and the seasonal change feature reflects the seasonal distribution of the communication traffic data.
[0047] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:
[0048] 1. The present invention can more deeply understand the inherent laws of communication traffic by performing trend factorization, seasonal factorization and periodic factorization on communication traffic data. This subdivision processing method can reveal hidden patterns in the data, thereby more accurately predicting future communication traffic;
[0049] 2. The present invention can construct a more accurate feature set by correlating the features of trend factors, seasonal factors and periodic factors, and screen out feature variables with strong correlation, thereby further improving the accuracy and reliability of the prediction model;
[0050] 3. Before constructing the prediction model, the present invention first screens the feature set to select the optimal feature set, which can exclude features with weak correlation with the prediction target and avoid overfitting problems. At the same time, it reduces the complexity of model training and improves the operating efficiency of the prediction model. The first primary prediction model, the second primary prediction model and the third primary prediction model are used for parallel prediction, which can make full use of the advantages of different models and improve the performance of the overall prediction model through the complementary effect between models. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0052] Figure 1 A flow chart of a communication traffic intelligent prediction method based on big data analysis of the present invention is shown;
[0053] Figure 2 The model prediction flow chart of the present invention is shown. DETAILED DESCRIPTION
[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0055] Example 1
[0056] See also Figure 1 As shown, the communication traffic intelligent prediction method based on big data analysis of this embodiment has the following process:
[0057] Step 1: Obtain the communication traffic data of the area to be predicted, perform trend decomposition, seasonal decomposition, and periodic decomposition on the communication traffic data to obtain trend factors, seasonal factors, and periodic factors, respectively.
[0058] Among them, the trend factor reflects the overall upward or downward trend of communication traffic data over a long period of time; the periodic factor reflects the pattern or trend of communication traffic data that repeats at fixed time intervals; the seasonal factor reflects the specific pattern or trend of communication traffic data that changes with the seasons within a year;
[0059] First, the trend factor is calculated using the moving average method. The calculation formula is as follows:
[0060]
[0061] Where, t=T, t=T+1, ..., M, M represents the last time point of the communication flow sequence; A t It represents the cumulative displacement value at time t in the communication flow sequence; It represents the trend item flow corresponding to time t in the communication flow sequence; T represents the period, and its value is selected according to different time modes;
[0062] Then, separate the period factor C, period factor = communication flow sequence divided by trend factor T;
[0063] After decomposing the trend factor T and the cycle factor C, the trend factor T and the cycle factor C are merged into a sequence TC, which represents the long-term trend and periodic fluctuations in the original time series.
[0064] Finally, the time series data are grouped by month, and the average value of each seasonal unit is calculated. These seasonal average values are divided by the average value of the entire communication traffic series to obtain the seasonal index S.
[0065] Step 2: Extract the characteristics of trend factors, including long-term growth trend characteristics and short-term fluctuation characteristics; extract the characteristics of periodic factors, including daily periodic characteristics, weekly periodic characteristics, and monthly periodic characteristics; extract the characteristics of seasonal factors, including holiday effect characteristics and seasonal change characteristics.
[0066] Daily periodicity: For example, traffic is usually higher during peak hours in the morning and evening (such as 8:00 to 10:00 a.m. and 6:00 to 8:00 p.m.), while traffic is relatively lower during other hours. This reflects the periodic changes in human activities and network usage habits.
[0067] Weekly periodicity: Traffic patterns on weekdays are usually different from those on weekends. For example, daytime traffic on weekdays may be higher because more people use the Internet at work or school, while weekend traffic peaks may occur in the afternoon or evening because people have more free time for Internet activities.
[0068] Monthly periodicity characteristics: The monthly periodicity of traffic may be related to holidays, special events, or user payment cycles (such as monthly bill settlement dates).
[0069] Characteristics of long-term growth trend: With the popularization of the Internet, the increase in mobile devices and the continuous emergence of new applications, communication traffic generally shows a long-term growth trend.
[0070] Short-term fluctuation characteristics: Based on the long-term growth trend, traffic data may experience short-term fluctuations due to seasons, events or other factors.
[0071] Characteristics of holiday effect: During important holidays (such as the Spring Festival, National Day, etc.), traffic data usually shows a significant increase due to increased activities such as travel, tourism and gatherings.
[0072] Seasonal change characteristics: With the change of seasons, people's activity patterns and network usage habits will also change, thus affecting the seasonal distribution of communication traffic. For example, in summer, traffic may increase due to factors such as student holidays and the peak tourist season.
[0073] Step 3: Correlate the features of trend factors, seasonal factors, and cycle factors, screen out feature variables with correlation greater than the set threshold, and integrate them into an available feature set, where correlation refers to the correlation between the feature and the target feature of communication traffic prediction; the available feature set refers to the feature set required for communication traffic prediction.
[0074] Specifically, the feature variable A and the target feature variable B are substituted into the association formula, and the formula is as follows:
[0075]
[0076] in, and are the means of feature variable A and target feature variable B respectively, α is the correlation coefficient, and the value range of α is [-1,1]. The larger the |α| is, the stronger the correlation between feature variable A and target feature variable B is. When the |α| value is between 0.8-1.0, the correlation between feature variable A and target feature variable B is extremely strong; when the |α| value is between 0.6-0.8, the correlation between feature variable A and target feature variable B is strong; when the |α| value is between 0.4-0.6, the correlation between feature variable A and target feature variable B is moderate; when the |α| value is between 0.2-0.4, the correlation between feature variable A and target feature variable B is weak; when the |α| value is between 0.0-0.2, the correlation between feature variable A and target feature variable B is weak or there is no correlation.
[0077] The features with the absolute value of correlation coefficient |α| between 0.6 and 1.0 are screened out and integrated to generate the available feature set C.
[0078] Step 4: Screen the available feature sets to select the optimal feature set.
[0079] Specifically, when there are two or more features in the available feature set that carry almost the same information, the features are marked as redundant features, and the redundant features are removed to screen out the optimal features. The available feature variable set C = (C1, C2, C3, ..., C m ), target feature variable set B = (B1, B2, B3, ..., B m); The calculation formula of mutual information I(B;C) is as follows:
[0080]
[0081] Among them, P(B m ,C m ) is the joint distribution probability of the target feature variable set B and the available feature variable set C, P(B m ) and P(C m ) is the marginal distribution probability; mutual information I(B;C) is the joint distribution P(B m , C m ) and marginal distribution P(B m ) and P(C m )'s relative entropy.
[0082] S401, filter out the feature with the largest mutual information value with the target feature in the available feature set through the correlation calculation formula, and obtain the available feature subset. The calculation formula is as follows:
[0083]
[0084] S402, filter out the feature with the smallest mutual information value with the target feature variable in the available feature subset through the redundancy calculation formula, the calculation formula is as follows:
[0085]
[0086] S403. Combine maxD(S,C) and minR(S) by a combination formula to obtain the correlation between the available features and the target features. The combination formula is as follows:
[0087] maxΦ(D,R),Φ=DR;
[0088] S404. Sort the features according to the calculated relevance, ensuring that the features that are most relevant to the target variable and have the lowest redundancy are placed in front.
[0089] S405. Select the feature with the greatest relevance and the lowest redundancy with other selected features from the sorted feature list, and add it to the optimal feature set.
[0090] S406. Repeat steps S401 to S405 until a preset number of features are selected to form an optimal feature set.
[0091] Step 5: Input the optimal feature subsets into the pre-trained communication traffic first primary prediction model, second primary prediction model and third primary prediction model, and output the first primary prediction result, second primary prediction result and third primary prediction result respectively.
[0092] Among them, the first primary prediction model is a periodic factor prediction model, which is trained based on multiple historical periodic factor data and is used to output the communication traffic prediction periodic factor; the second primary prediction model is a seasonal factor prediction model, which is trained based on multiple historical seasonal factor data and is used to output the communication traffic prediction seasonal factor; the third primary prediction model is a trend factor prediction model, which is trained based on multiple historical trend factor data and is used to output the communication traffic prediction trend factor.
[0093] Step 6: Integrate the first primary prediction result, the second primary prediction result and the third primary prediction result to obtain the final prediction result of the communication traffic.
[0094] See also Figure 2 As shown in the figure, the integration process is as follows:
[0095] Corresponding weights d, e and f are assigned to the first primary prediction result, the second primary prediction result and the third primary prediction result respectively, and substituted into the prediction formula. The prediction formula is as follows:
[0096] Final prediction result = (first primary prediction result * d) + (second primary prediction result * e) + (third primary prediction result * f);
[0097] Among them, d+e+f=1, and d, e and f are all greater than 0.
[0098] The beneficial effects of this embodiment are as follows: by performing trend factor decomposition, seasonal factor decomposition and cycle factor decomposition on communication traffic data, it is possible to have a deeper understanding of the inherent laws of communication traffic; the characteristics of trend factors, seasonal factors and cycle factors are correlated to screen out characteristic variables with strong correlation, thereby improving the accuracy and reliability of the prediction model; before building the prediction model, the feature set is first screened to screen out the optimal feature set, which can exclude features with weak correlation with the prediction target and avoid overfitting problems, while reducing the complexity of model training and improving the operating efficiency of the prediction model; using the first primary prediction model, the second primary prediction model and the third primary prediction model for parallel prediction can fully utilize the advantages of different models and improve the performance of the overall prediction model through the complementary effect between models.
[0099] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.
[0100] Those of ordinary skill in the art will appreciate that the algorithmic steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0101] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0102] In the several embodiments provided in this application, it should be understood that the disclosed methods can be implemented in other ways.
[0103] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for 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 methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.
[0104] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. A communication traffic intelligent prediction method based on big data analysis, characterized in that: The process is as follows: Step 1: Obtain the communication traffic data of the area to be predicted, perform trend decomposition, seasonal decomposition, and periodic decomposition on the communication traffic data to obtain trend factors, seasonal factors, and periodic factors respectively; Step 2: Extract the characteristics of trend factors, including long-term growth trend characteristics and short-term fluctuation characteristics; The features of extracted periodic factors include daily periodic features, weekly periodic features, and monthly periodic features; the features of extracted seasonal factors include holiday effect features and seasonal change features; Step 3: Correlation processing is performed on the features of trend factors, seasonal factors, and periodic factors, and feature variables with correlation greater than a set threshold are screened out and integrated into an available feature set, where correlation refers to the correlation between the feature and the target feature of communication traffic prediction; the available feature set refers to the feature set required for communication traffic prediction; Step 4: Screen the available feature sets to select the optimal feature set; Step 5: input the optimal feature subset into the pre-trained communication traffic first primary prediction model, second primary prediction model and third primary prediction model, and output the first primary prediction result, the second primary prediction result and the third primary prediction result respectively; Step 6: Integrate the first primary prediction result, the second primary prediction result and the third primary prediction result to obtain the ultimate prediction result of communication traffic.
2. The communication traffic intelligent prediction method based on big data analysis according to claim 1 is characterized in that: The trend factor reflects the overall upward or downward trend of the communication traffic data over a long period of time; the periodic factor reflects the pattern or trend of the communication traffic data that repeats within a fixed time interval; and the seasonal factor reflects the specific pattern or trend of the communication traffic data as it changes with the seasons within a year.
3. The communication traffic intelligent prediction method based on big data analysis according to claim 1 is characterized in that: The trend decomposition, seasonal decomposition, and periodic decomposition are specifically as follows: The trend factor is calculated using the moving average method. The calculation formula is as follows: Wherein, t=T, t=T+1, ..., M, M represents the last time point of the communication flow sequence; At represents the cumulative displacement value at time t in the communication flow sequence; It represents the trend item flow corresponding to time t in the communication flow sequence; T represents the period, and its value is selected according to different time modes; Then, separate the period factor C, period factor = communication flow sequence divided by trend factor T; After decomposing the trend factor T and the cycle factor C, the trend factor T and the cycle factor C are combined into a sequence TC, which represents the long-term trend and periodic fluctuation in the original time series; Finally, the time series data are grouped by month, and the average value of each seasonal unit is calculated. These seasonal average values are divided by the average value of the entire communication traffic series to obtain the seasonal index S.
4. The communication traffic intelligent prediction method based on big data analysis according to claim 1 is characterized in that: The step three is as follows: Substitute the feature variable A and the target feature variable B into the association formula, the formula is as follows: in, and are the means of feature variable A and target feature variable B respectively, α is the correlation coefficient, and the value range of α is [-1,1]. The larger the |α| is, the stronger the correlation between feature variable A and target feature variable B is. When the |α| value is between 0.8-1.0, the feature variable A and target feature variable B are judged to have an extremely strong correlation; when the |α| value is between 0.6-0.8, the feature variable A and target feature variable B are judged to have a strong correlation; when the |α| value is between 0.4-0.6, the feature variable A and target feature variable B are judged to have a medium correlation; when the |α| value is between 0.2-0.4, the feature variable A and target feature variable B are judged to have a weak correlation; when the |α| value is between 0.0-0.2, the feature variable A and target feature variable B are judged to have a weak correlation or no correlation; Features with absolute values of correlation coefficients |α| between 0.6 and 1.0 were selected and integrated to generate a usable feature set.
5. The method for intelligent prediction of communication traffic based on big data analysis according to claim 1 is characterized in that: The step 4 is as follows: When there are two or more features in the available feature set that carry almost the same information, the features are marked as redundant features, and the redundant features are removed to screen out the optimal features. The available feature variable set C = (C1, C2, C3, ..., C m ), target feature variable set B = (B1, B2, B3, ..., B m ); The calculation formula of mutual information I(B;C) is as follows: Among them, P(B m ,C m ) is the joint distribution probability of the target feature variable set B and the available feature variable set C, P(B m ) and P(C m ) is the marginal distribution probability; mutual information I(B;C) is the joint distribution P(B m , C m ) and marginal distribution P(B m ) and P(C m )'s relative entropy; S401, filter out the feature with the largest mutual information value with the target feature in the available feature set through the correlation calculation formula, and obtain the available feature subset. The calculation formula is as follows: S402, filter out the feature with the smallest mutual information value with the target feature variable in the available feature subset through the redundancy calculation formula, the calculation formula is as follows: S403. Combine maxD(S,C) and minR(S) by a combination formula to obtain the correlation between the available features and the target features. The combination formula is as follows: maxΦ(D,R),Φ=DR; S404, sorting the features according to the calculated relevance, ensuring that the features most relevant to the target variable and with the lowest redundancy are placed in front; S405, from the sorted feature list, select the feature with the greatest correlation and the lowest redundancy with other selected features, and add it to the optimal feature set; S406. Repeat steps S401 to S405 until a preset number of features are selected to form an optimal feature set.
6. The communication traffic intelligent prediction method based on big data analysis according to claim 1 is characterized in that: The first primary prediction model is a periodic factor prediction model, which is trained based on multiple historical periodic factor data and is used to output a communication traffic prediction periodic factor; the second primary prediction model is a seasonal factor prediction model, which is trained based on multiple historical seasonal factor data and is used to output a communication traffic prediction seasonal factor; the third primary prediction model is a trend factor prediction model, which is trained based on multiple historical trend factor data and is used to output a communication traffic prediction trend factor.
7. The communication traffic intelligent prediction method based on big data analysis according to claim 1 is characterized in that: The integration process is as follows: Corresponding weights d, e and f are assigned to the first primary prediction result, the second primary prediction result and the third primary prediction result respectively, and substituted into the prediction formula. The prediction formula is as follows: Final prediction result = (first primary prediction result * d) + (second primary prediction result * e) + (third primary prediction result * f); Among them, d+e+f=1, and d, e and f are all greater than 0.
8. The method for intelligent prediction of communication traffic based on big data analysis according to claim 1 is characterized in that: The daily periodicity feature reflects the daily changes in the communication traffic data; the weekly periodicity feature reflects the weekly changes in the communication traffic data; the monthly periodicity feature reflects the monthly changes in the communication traffic data; the long-term growth trend feature reflects that the communication traffic data generally presents a long-term growth trend; the short-term fluctuation feature reflects the short-term fluctuations in the communication traffic data; the holiday effect feature reflects that the communication traffic data shows a significant increase during holidays; The seasonal variation characteristics reflect the seasonal distribution of communication traffic data.