A method for optimal configuration of SIM card traffic based on massive data analysis model

Abstract

The present invention discloses a method for optimally configuring SIM card traffic based on a massive data analysis model. The method comprises the following steps: S1, proactively checking and collecting abnormal terminals and SIM cards; S2, bidirectional data comparison and cleaning of SIM card relationships; and S3, establishing a feature library to configure an optimal SIM card traffic plan. By using power data from an electricity information collection system, a marketing system, and a PMS system, and SIM card traffic data provided by the three major operators, abnormal terminals and SIM cards are checked and collected, bidirectional data comparison and cleaning of SIM card relationships are performed, and a dedicated variable-based feature library for data traffic usage of collection devices is established to identify SIM card usage characteristics. A random forest algorithm is used to create SIM card traffic usage models for different collection devices, and optimal SIM card traffic plans are configured for each type of collection device. The configuration method of the present invention can select appropriate SIM card packages, saving a significant amount of money spent on SIM card traffic.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a method for optimizing SIM card traffic flow based on a massive data analysis model. Background Art

[0002] As the scale of electricity consumption data collection and access continues to expand, the number of access terminals is also increasing, leading to a growing number of SIM cards. Currently, the SIM card data requirements of various collection devices vary, with a wide variety of data plans. Monthly SIM card rental fees are substantial and increasing year by year. Many of these cards are "zombie" cards, often left unused for long periods of time. By improving daily SIM card management processes, applying big data analytics to monitor status, and optimizing SIM card package selection, we are increasing utilization rates and supporting lean management and improving quality and efficiency for the company.

[0003] A Chinese patent document titled "A Method and Apparatus for Utilizing Universal Data for Multiple SIM Cards," published as CN113747383A, discloses a method and apparatus for utilizing universal data for multiple SIM cards, including: binding a cross-operator data pool to at least two SIM cards, wherein the cross-operator data pool is bound to a preset package, and the type and data size of the package bound to the SIM cards are the same as the type and data size of the preset package; when one of the at least two SIM cards is used, receiving data usage data from the SIM card in real time and deducting the total data usage from the cross-operator data pool accordingly. However, the invention does not address the specific configuration method for SIM card data usage. Summary of the Invention

[0004] The present invention solves the problem of numerous and serious wasteful existing SIM card traffic packages and proposes a method for optimal SIM card traffic configuration based on a massive data analysis model. By using power data from an electricity information collection system, a marketing system, and a PMS system, and SIM card traffic data provided by the three major operators, terminals and SIM cards with abnormal data collection are checked, and two-way data comparison is performed to clarify SIM card relationships. Simultaneously, a data traffic usage feature library for collection devices based on a dedicated variable is established to identify SIM card usage characteristics. A random forest algorithm is used to create SIM card traffic usage models for different collection devices, and optimal SIM card traffic plans are configured for each type of collection device. The configuration method of the present invention can select appropriate SIM card packages and save a large amount of SIM card traffic costs.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for optimally configuring SIM card traffic based on a massive data analysis model, comprising the following steps:

[0006] S1, proactively investigates and collects abnormal terminals and SIM cards;

[0007] S2, two-way data comparison and SIM card relationship cleaning;

[0008] S3: Establish a feature library to configure the optimal SIM card traffic plan. In this invention, the daily work method of checking for unsuccessful terminals through the full data collection success rate module in the electricity collection system and arranging personnel for on-site repair and maintenance is relatively passive. On the one hand, the system prompts that the terminals with unsuccessful data collection have affected the line loss of the line on that day. On the other hand, the system displays the terminals that are not online, but does not count the terminals with poor data collection status. Therefore, RPA robots are used for active data collection; two-way data comparison is used to clean SIM card relationships. At the same time, a data traffic usage feature library of collection equipment based on dedicated variables is established to identify SIM card usage characteristics. The random forest algorithm is used to create SIM card traffic usage models for different collection equipment, and the optimal SIM card traffic plan is configured for each type of collection equipment, saving a large amount of SIM card fees.

[0009] Preferably, step S1 includes the following steps:

[0010] S11, export the load data of the dedicated transformer users in the current month from the collection system and calculate the collection rate;

[0011] S12, by comparing the information in the marketing system to inquire whether the dedicated transformer user has reported suspension, the terminals with a collection rate lower than 95% are listed as terminals with collection abnormalities and the SIM cards are recorded in the first database;

[0012] S13, using the SIM card traffic usage data provided by the operator, identify SIM cards whose traffic exceeds twice their package quota. The collection system then compares the data to identify the corresponding terminals, and these terminals and SIM cards are also included in the second database as having abnormal data collection. In the present invention, a 100% collection rate means 97 data items per day, of which 96 are load data items recorded every 15 minutes per day, and 1 is meter reading data. The operators are specifically the three major operators.

[0013] Preferably, step S2 includes the following steps:

[0014] S21, the operator data and the power system data are self-checked respectively. On the operator data side, the monthly traffic data of the SIM cards provided by the operator is used to filter out the SIM cards with traffic data of 0 and record them as the third database;

[0015] S22, on the power data side, query and derive from the marketing system the correspondence between the terminal SIM cards whose IP addresses do not match the terminals and whose binding is incorrect, and record it as a fourth database;

[0016] S23, combining the above-mentioned two-way data, finds the corresponding terminal exchange number and account number in the collection system for the SIM card number in the third database. Then, in the marketing system, use the account number to check whether the user is a user who has suspended transformer use. The remaining SIM cards and terminals of suspended users are removed and considered abnormal cards and recorded in the fifth database. In the present invention, the binding of SIM cards to dedicated transformer terminals is implemented through manual binding, which is a weak verification relationship. Two common situations occur in daily life: incorrect binding of an inactive SIM card number with no traffic data to an active terminal, and incorrect binding of two active SIM cards. This leads to a mismatch between the device logical address and the SIM card number. The above steps can effectively solve this problem.

[0017] Preferably, step S3 includes the following steps:

[0018] S31, remove abnormal data, create a sample, take all terminals and SIM cards with special changes as objects, and remove the terminals in the first to fifth databases;

[0019] S32, compare the random forest and vector machine models, and select the random forest model as the optimal model;

[0020] S33, calculate the maximum traffic plan for different terminals. In the present invention, all terminals and SIM cards with dedicated transformers are taken as objects, and terminals in the first to fifth databases are eliminated. The data is sampled with information on one terminal and one SIM card per day. This information includes the SIM card operator, card type (IoT card or communication card), dedicated transformer capacity, terminal manufacturer, terminal production year, daily traffic usage, daily average humidity, and daily average temperature. In the sample data, dedicated transformer capacity, terminal manufacturer, and terminal production year are obtained through the acquisition system; daily traffic usage, card type, and operator are provided by the three major operators; and daily average temperature and average humidity are obtained from the meteorological website. Some elements of the sample data are labeled, with the three major operators (China Mobile, China Telecom, and China Unicom) marked as numbers respectively; other data (terminal manufacturer and card type) are marked as numbers respectively. The traffic data is used as the sample output, and the rest are the sample's characteristic values. The traffic data, average temperature, and humidity are rounded, totaling 8 characteristic values.

[0021] Preferably, step S32 specifically selects several samples as a sample data set D, randomly extracted from a data center in one year, and the random forest and vector machine use the same sample data. Using the ten-fold cross-validation method, the data set D is randomly divided into 10 subsets D1, D2, D3... and D10 of equal capacity; one of the data Di is taken as the test data set TestDATA i, and the remaining 9 points are used as the training data set TrainDATA i to form the i-th training and test set (TestDATAi, TrainDATA i) (i=1,2,3...9,10). The 10 sets of training sets and test sets are used for training and testing. If the training and test accuracy requirements are met, the next step of calculation can be carried out. If not, it is necessary to set parameters or resample, and select the optimal model based on the test accuracy and training accuracy. In the present invention, when constructing the random forest model and the vector machine model, the corresponding model parameters need to be set. The optimal parameters are obtained after multiple debugging and parameter sensitivity analysis.

[0022] Preferably, step S33 specifically obtains a month's worth of weather data, including temperature Ti and humidity RHi. The user's basic information is then combined to form a sample Xi (i = 1, 2, 3, ..., 29, 30). This sample is then placed into a random forest model for calculation to obtain the daily traffic volume yi used by the user's terminal SIM card for the next month. The daily traffic volume is then summed and multiplied by the allowable error coefficient to obtain the maximum monthly traffic volume Ymax. If Ymax < 15, the first category of IoT card packages is selected. If Ymax > 15 and Ymax < 30, the second category of IoT card packages is selected. If Ymax > 30, the terminal SIM card is manually inspected. In the present invention, average temperature and humidity are used as daily variable factors in the data sample, and weather data for a month can be obtained through a weather website.

[0023] As a preference, manual on-site inspections of the terminals of the fourth and fifth databases are required every month to ensure the timeliness of the databases.

[0024] The beneficial effects of the present invention are as follows: the solution of the present invention uses SIM card traffic data to check terminals and SIM cards with abnormal collection, performs two-way data comparison to clean up SIM card relationships, and simultaneously establishes a data traffic usage feature library for collection devices based on a dedicated variable to find SIM card usage characteristics. The random forest algorithm is used to create SIM card traffic usage models for different collection devices, configure optimal SIM card traffic plans for various types of collection devices, select appropriate SIM card packages according to different situations, prevent waste, and save a lot of SIM card fees. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0026] Example:

[0027] This embodiment proposes a method for optimally configuring SIM card traffic based on a massive data analysis model. Figure 1 , mainly includes the following steps: Step S1, actively check and collect abnormal terminals and SIM cards; specifically, in this step, it also includes multiple sub-steps, among which, Step S11, exports the load data of the dedicated transformer users of the current month from the collection system and calculates the collection rate (a collection rate of 100% is 97 data per day); specifically, RPA robots are used for collection, and all power data collection is done by RPA robots; Step S12, by comparing with the information in the marketing system to inquire whether the dedicated transformer users have reported suspension, the terminals with a collection rate lower than 95% are listed as terminals with collection abnormalities and the SIM cards are recorded as the first database; Step S13, through the SIM card traffic usage data provided by the operator, the SIM cards whose traffic exceeds twice the package usage are checked, and the corresponding terminals are found through comparison in the collection system, and such terminals and SIM cards that are also included in the collection abnormalities are recorded as the second database, and manual on-site active inspections are carried out. In the present invention, a collection rate of 100% is 97 data per day, of which 96 are load data every 15 minutes a day, and 1 is meter reading data; the operators are specifically the three major operators;

[0028] The details of terminals and SIM cards that are actively troubleshooting data collection anomalies can be found in Table 1 below:

[0029] Table 1: Details of terminals and SIM cards that were proactively checked for data collection anomalies

[0030]

[0031] Step S2, two-way data comparison and cleaning of SIM card relationships; specifically, it includes the following multiple sub-steps: Step S21, the operator data and the power system data are self-checked respectively. On the operator data side, the SIM card with a flow data of 0 is filtered out through the monthly flow data of the SIM card provided by the operator and recorded as the third database; Step S22, on the power data side, the corresponding relationship of the terminal SIM card whose IP address does not match the terminal and has an incorrect binding is queried and exported from the marketing system and recorded as the fourth database; Step S23, combined with the above two-way data, the SIM card number of the third database is used to find the corresponding terminal office number and account number in the acquisition system, and then the marketing system uses the account number to find out whether the user belongs to a user who has suspended the use of the transformer. The remaining SIM cards and terminals of the suspended users are removed and regarded as abnormal cards and recorded as the fifth database. In the present invention, the binding of the SIM card and the dedicated transformer terminal is implemented through manual binding, which is a weak verification relationship. There are mainly two situations in daily life: the SIM card number that is not in use and has no flow data is incorrectly bound to the operating terminal and the SIM card number in operation is incorrectly bound to the SIM card number. This leads to the situation that the logical address of the device does not correspond to the SIM card number. The above steps can effectively solve this problem;

[0032] Among them, the details of the two-way data comparison and SIM card cleaning relationship are shown in Table 2 below:

[0033] Table 2: Details of the relationship between two-way data comparison and cleaning of SIM cards

[0034]

[0035] Step S3: Establishing a feature library to configure the optimal SIM card data plan. Specifically, the method includes the following sub-steps: First, step S31 is performed to eliminate abnormal data and establish a sample. All terminals and SIM cards with dedicated transformers are used as the sample, excluding terminals in the first through fifth databases. Step S32: Comparing operations using a random forest and vector machine model, the random forest model is selected as the optimal model. Step S33: Calculating the maximum data plan for different terminals. In this embodiment of the present invention, all terminals and SIM cards with dedicated transformers are used as the sample, excluding terminals in the first through fifth databases. The data sample is based on one terminal and one SIM card per day. This information includes the SIM card operator, card type (IoT card or communication card), dedicated transformer capacity, terminal manufacturer, terminal production year, daily data usage, daily average humidity, and daily average temperature. In the sample data, dedicated transformer capacity, terminal manufacturer, and terminal production year are obtained through the data collection system; daily data usage, card type, and operator are provided by the three major operators; and daily average temperature and average humidity are obtained from the meteorological website. We also labeled some elements of the sample data, marking the three major operators (China Mobile, China Telecom, and China Unicom) with numbers; other data terminal manufacturers and card types were also labeled with numbers. Traffic flow data was used as the sample output, and the rest were the sample's feature values. Traffic flow data, average temperature, and humidity were rounded, resulting in a total of eight feature values.

[0036] In step S32, specifically, several samples are selected as a sample dataset D. In this embodiment, 200,000 samples are randomly sampled from a data center over a one-year period. The random forest and vector machine models use the same sample data. Using a ten-fold cross-validation method, dataset D is randomly divided into 10 subsets of equal capacity: D1, D2, D3, ..., and D10. One of the data, Di, is used as the test dataset TestDATA i, and the remaining 9 are used as the training dataset TrainDATA i, forming the i-th training and test set (TestDATA i, TrainDATA i) (i = 1, 2, 3, ..., 9, 10). When constructing the random forest model and vector machine model, the corresponding model parameters need to be set. After multiple debugging and parameter sensitivity analysis, the optimal parameters are obtained.

[0037] Subsequently, 10 training and test sets were used for training and testing. If the training and test accuracy requirements were met, the next step of calculation could be performed. If not, parameters needed to be set or resampled, and the optimal model was selected based on the test and training accuracy. As shown in Table 3 below, the random forest model had higher accuracy in terms of test and training accuracy, so it was chosen:

[0038] Table 3: Accuracy comparison between random forest and vector machine

[0039]

[0040] The specific process of step S33 includes: in the data sample, the average temperature and humidity are used as daily variable factors. The weather data temperature Ti and humidity RHi within a month can be obtained through the weather website. The user's basic information is spliced ​​to form a sample Xi (i=1,2,3....29,30.), which is put into the random forest model for calculation to obtain the traffic yi used by the user's terminal SIM card every day in the next month. The traffic is added and multiplied by the allowable error coefficient to obtain the maximum monthly traffic Ymax. If Ymax<15, the first type of IoT card package is selected. If Ymax>15 and Ymax<30, the second type of IoT card package is selected. If Ymax>30, the terminal SIM card is manually checked. In this embodiment, the allowable error coefficient is 1.1. The SIM cards currently used by the dedicated transformer are the 2.2 yuan 15M IoT card and the 4.7 yuan 30M IoT card packages. The excess traffic needs to be calculated separately. In this embodiment, the 2.2 yuan 15M IoT card corresponds to the first type of IoT card package, and the 4.7 yuan 30M IoT card package corresponds to the second type of IoT card package.

[0041] During step S2, manual on-site inspections of the terminals of the fourth and fifth databases are required every month to ensure the timeliness of the databases.

[0042] In the present invention, the daily work method of checking for terminals that failed to collect data through the full collection success rate module in the electricity collection system and arranging personnel to perform on-site repair and maintenance is relatively passive. On the one hand, the system prompts that the terminals that failed to collect data have affected the line loss on that day. On the other hand, the system displays terminals that are not online and does not count the terminals with poor collection status. Therefore, RPA robots are used for active collection; and two-way data comparison is used to clean SIM card relationships. At the same time, a data traffic usage feature library of collection equipment based on dedicated variables is established to find SIM card usage characteristics, and a random forest algorithm is used to create SIM card traffic usage models for different collection equipment, and the best SIM card traffic solutions are configured for various types of collection equipment, saving a lot of SIM card fees.

[0043] The above embodiments are further elaborations and illustrations of the present invention for ease of understanding, and are not intended to limit the present invention in any way. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for optimally configuring SIM card traffic based on a massive data analysis model, characterized in that: include: S1, proactively investigates and collects abnormal terminals and SIM cards; this includes: exporting the load data of dedicated transformer users for the current month from the collection system and calculating the collection rate; comparing this with the information in the marketing system regarding whether dedicated transformer users have reported suspension, and listing terminals with a collection rate below 95% as abnormally collected terminals and SIM cards in the first database; using SIM card traffic usage data provided by the operator, identifying SIM cards with traffic exceeding twice the package amount, and matching these terminals through the collection system, recording these abnormally collected terminals and SIM cards in the second database; S2, two-way data comparison and cleaning of SIM card relationships; including: self-checking of operator data and power system data respectively. On the operator data side, SIM cards with zero traffic data are screened out through the monthly traffic data of SIM cards provided by the operator and recorded as the third database; on the power data side, the corresponding relationships of SIM cards of terminals with mismatched IP addresses and terminals and incorrect bindings are queried and exported from the marketing system and recorded as the fourth database; combined with the above two-way data, the SIM card number in the third database is found in the collection system with the corresponding terminal office number and account number, and then the marketing system uses the account number to find out whether the user belongs to a user who has suspended the use of the transformer. The remaining SIM cards and terminals of the suspended users are removed and regarded as abnormal cards and recorded as the fifth database; S3, establish a feature library to configure the best SIM card traffic plan; eliminate the terminals in the first to fifth databases, select the random forest model as the optimal model, and calculate the maximum traffic plan for different terminals.

2. The method for optimally configuring SIM card traffic based on a massive data analysis model according to claim 1, characterized in that: The step S3 comprises the following steps: S31, remove abnormal data, create a sample, take all terminals and SIM cards with special changes as objects, and remove the terminals in the first to fifth databases; S32, compare the random forest and vector machine models, and select the random forest model as the optimal model; S33, calculating the maximum flow rate solutions of different terminals.

3. The method for optimally configuring SIM card traffic based on a massive data analysis model according to claim 2, characterized in that: The step S32 specifically selects several samples as the sample data set D, which are randomly extracted from a data center in one year. The random forest and vector machine use the same sample data, and adopt the ten-fold cross-validation method to randomly divide the data set D into 10 subsets D1, D2, D3... and D10 of equal capacity; take one of the data Di as the test data set TestDATAi, and the remaining 9 points as the training data set TrainDATAi to form the i-th training and test set (TestDATAi, TrainDATAi) (i=1,2,3...9,10), and use 10 groups of training sets and test sets for training and testing. If the training and test accuracy requirements are met, the next step of calculation can be carried out. If not, it is necessary to set parameters or re-sampling, and select the optimal model according to the test accuracy and training accuracy.

4. The method for optimizing SIM card traffic flow based on a massive data analysis model according to claim 2, characterized in that: The step S33 specifically obtains the weather data temperature Ti and humidity RHi for one month, splices the user's basic information to form a sample Xi (i=1, 2, 3....29, 30), puts it into the random forest model for calculation, and obtains the traffic yi used by the user's terminal SIM card every day in the next month. The daily traffic is added and multiplied by the allowable error coefficient to obtain the maximum monthly traffic Ymax. If Ymax<15, the first type of IoT card package is selected.

5. The method for optimally configuring SIM card traffic based on a massive data analysis model according to claim 4, characterized in that: If Ymax>15 and Ymax<30, select the second type of IoT card package. If Ymax>30, manually check the terminal SIM card.

6. The method for optimally configuring SIM card traffic based on a massive data analysis model according to claim 1, characterized in that: Manual on-site inspections of the terminals of the fourth and fifth databases are required every month.

7. The method for optimally configuring SIM card traffic based on a massive data analysis model according to claim 1, characterized in that: In step S1, an RPA robot is used for collection, and all power data is collected using an RPA robot.

Images