TWI690894B - Automatic round-collecting for broadband customer and automatic round-collecting server - Google Patents

Abstract

An automatic round-collecting for broadband customer and an automatic round-collecting server are provided. Multiple users are grouped into multiple user group according to attribute thereof. The attribute of each user includes located area and rented rate. Sample size of each user group is determined according to corresponding rented rate. Then, the round-collecting operation for the user groups are performed according to the determined sample sizes. Accordingly, an easy and smart determination sampling method are provided, so as to achieve a best representation and collecting data without interruption.

Description

Automatic polling method and broadband polling server for broadband service

The invention relates to a telecommunication service spot check technology, and in particular to an automatic polling method and an automatic polling server for broadband services.

For telecom operators, users' traffic (traffic) data is the most basic application in big data analysis. However, taking the example of a large-scale industry, the number of users in the network is easily one million. If the industry wants to collect all users' communication information at the same time, it will often be limited by the performance of the access equipment and the collection equipment, which is not easy to achieve. In the prior art, although there is a re-sampling method, the traditional simple sampling usually cannot satisfy the characteristics of diversification and time-varying changes of telecommunication users.

In view of this, the present invention provides an automatic polling method and an automated polling server for broadband services, which are grouped according to user attributes, and determine the sampling ratio of each group according to the lease rate, and then perform batch polling.

The automatic polling method of the broadband service of the present invention includes the following steps. Group users according to their user attributes to generate user groups. The user attributes of each user include the location and lease rate. According to the corresponding lease rates of those user groups, the number of samples of each user group is determined separately. Carry out the polling operation based on the determined sample number of the user group.

The automatic polling server of the present invention includes a communication module, a storage and a processor. The communication module is used to send and receive data. The memory records several modules. The processor is coupled to the communication module and the memory, and accesses and loads those modules recorded in the memory. Those modules include attribute analysis module, grouping sample decision module and polling user sampling and setting module. The attribute analysis module groups users according to their user attributes to generate user groups. The user attributes of each user include the location and lease rate. The grouping sample decision module determines the sampling number of each user group according to the corresponding lease rate of those user groups. The polling user sampling and setting module performs the polling operation according to the determined number of samples of the user group.

Based on the above, the automatic polling method and the automatic polling server of the broadband service according to the embodiments of the present invention perform stratified sampling based on user attributes (for example, location and lease rate, etc.) from the obtained user data of the whole area. Among them, according to the number and importance of users at each lease rate, all users are automatically classified into several user groups (for example, divided into main rates (the first few rates with the largest number of users), important rates (new services or special Observed rate, or determined by the user) and secondary rate (not the aforementioned three types). Then, according to the different types of rate, assign the number of samples (for example, the main rate users use stratified simple sampling; the important rate users use comprehensive spot checks; the non-primary rate use the whole area sampling or ignore sampling according to the sampling situation). Summarize the total number of samples for each of the aforementioned rate categories and compare it with the total inspection allowance of the equipment. If the total inspection allowance is exceeded, the accuracy is reduced or the secondary items are not selected, and then the number of samples is determined and adjusted until the total inspection number meets the total inspection allowance. Arrange the list of each batch to be checked in order until all users are selected. Then, according to the batch, a fixed period of data collection and repeated polling are carried out to maintain the uninterrupted data acquisition. During the inspection period, if there are users withdrawing rent or other changes within a certain batch, new users will be added to maintain the total number of samples.

In order to make the above-mentioned features and advantages of the present invention more obvious and understandable, the embodiments are specifically described below in conjunction with the accompanying drawings for detailed description as follows.

FIG. 1 is a block diagram of components of an automated polling server 1 according to an embodiment of the invention. Referring to FIG. 1, the automatic polling server 1 may be various types of servers, desktop computers, notebook computers, workstations, back-end hosts, and other electronic devices. The automatic polling server 1 includes at least but not limited to the storage 120, the communication module 130 and the processor 150.

The storage 120 can be any type of fixed or removable random access memory (RAM), read only memory (Read Only Memory, ROM), flash memory (flash memory), traditional hard disk (Hard Disk Drive, HDD), Solid-State Drive (SSD) or similar components, and used to record code, software modules (for example, attribute analysis module 111, attribute weight management module 112, user office Module 113, group sampling number generator 114, decentralized sample decision module 115, polling user sampling and setting module 116, circuit transaction management module 117, network management module 118, etc.), user list 119. A matrix of grouping sample numbers, and other data or files, the details of which are to be detailed in subsequent embodiments.

The communication module 130 may be a communication transceiver supporting wired communication technologies such as Ethernet and optical fiber, or fourth-generation (4G) or later generation mobile communication, Wi-Fi and other wireless communication technologies. The communication module 130 is used to transmit and receive data with the outside world.

The processor 150 is coupled to the storage 120 and the communication module 130. The processor 150 may also be a central processing unit (CPU), a microcontroller, a programmable controller, an application specific integrated circuit (ASIC), Wafer or other similar components or a combination of the above components. In this embodiment, the processor 150 executes all operations of the deployment decision device 1, and the processor 150 can access and load those software modules recorded in the storage 120.

In order to facilitate the understanding of the operation process of the embodiment of the present invention, a number of embodiments will be given below to describe in detail the process of polling for users in the embodiment of the present invention. Hereinafter, the methods described in the embodiments of the present invention will be described with various devices, components, and modules in the automatic polling server 1. The various processes of the method can be adjusted according to the implementation situation, and it is not limited to this.

The user-side communication needs to collect all user's communication data at the same time for big data analysis, but it is limited by the large number of user access devices and the limited collection efficiency of the device, which makes it difficult to collect all of them. The embodiment of the present invention solves the aforementioned problem by using the resampling method. Although traditional stratified sampling is simple and easy to implement, rough or unstratified sampling may deviate from the representativeness and accuracy of the sampling list to the user's parent, and thus fail to meet the characteristics of telecommunications users that diversify and change over time. The embodiment of the present invention proposes a simple and intelligent decision-making sampling inspection method to achieve the purpose of representative and uninterrupted data collection. This will be explained in detail below.

2 is a flowchart of an automatic polling method according to an embodiment of the invention. Referring to FIG. 2, the user situation module 113 first obtains a list 119 of all users receiving broadband services. The attribute analysis module 111 groups all users according to the user attributes recorded in the user list 119 to generate one or more user groups (step S210). The user attributes of the user include the location (for example, branch office/operation office) and lease rate. Specifically, when the automated polling server 1 performs the survey, the user samples can be grouped and the survey users can be selected for each user group to avoid over-concentration of the survey in certain areas or certain lease rates Users, which in turn leads to a deviation of the representativeness and accuracy of the random test list from the user’s parent. In addition, the grouping of users according to attributes also maintains the flexibility to adjust the individual sampling samples of each user group, so as to facilitate the subsequent decision-making to optimize the grouping samples and the basis for automatically evaluating the individual minimum sampling numbers of different user groups.

For example, FIG. 3 is an example illustrating the initial clustering sample number matrix. Each column is distinguished according to user attributes such as location (for example, branch office/operation office) and lease rate. For example, the unit level 2 of user group 1 is A1B1 and the lease rate is SP1; while the same unit level 2 is A1B1, but the lease rate is SP2 different from SP1 of user group 1, so it is classified into user group 2; The rest can be deduced by analogy.

After grouping, the attribute weight management module 112 determines the importance of different user groups according to the lease rates of these user groups (step S220). In one embodiment, the attribute weight management module 112 sets three major levels: primary rate, important rate, and secondary rate. The attribute weight management module 112 uses at least one of the users with the largest number of users corresponding to the lease rate in the user group (for example, the top 3, top 5, or top 6 lease rates with the largest number of users) as the main rate . Taking Figure 3 as an example, the first four lease rates with the largest number of users are SP1~SP4. The attribute weight management module 112 regards at least one of those user groups that needs to pay attention to the lease rate (for example, a lease rate that has recently been newly added or needs special observation according to actual needs) as an important rate. Taking Figure 3 as an example, the lease rates recently added or need attention are SP5~SP8. In addition, the attribute weight management module 112 regards at least one of those user groups that are not the primary rate and the important rate as the secondary rate. That is, other lease rates that are not categorized as a primary rate or an important rate will be considered secondary rates.

Next, the attribute weight management module 112 sets the sampling rate of those user groups whose main rate is based on multiple sampling confidence levels (that is, using stratified sampling survey), and sets the sampling rate of those user groups whose important rate is One hundred percent (ie, all surveys are used), and the sampling rate of those user groups of the secondary rate is set to zero (ie, no surveys). In addition, the attribute weight management module 112 can further prioritize the lease rates in each important level. Taking the main rate in Figure 3 as an example, SP1 has the highest priority, SP2 is second, SP3 is third, and SP4 is last.

It should be noted that, in other embodiments, if importance is not considered, the processor 150 may also adopt stratified sampling inspection for all user groups.

The grouped sample generator 114 designs different sampling confidence levels for each user group, and plans three sampling confidence levels (for example, 99%, 95% and 95%) under the condition that the margin of error (percentage %) is 3 90%), so as to obtain the minimum sampling number that meets the sampling of each user group for the decision basis of the main rate grouping stratified sampling number (step S230). Take the grouped sampling matrix of Figure 3 as an example. The number of users in user group 1 (A1/A1B1/SP1) is 12323, and the grouped sample generator 114 can then calculate the sampling confidence level based on 99%, 95%, and 90% for the error range percentage of 3, respectively The required number of samples is 1608, 983 and 705. Next, the initial matrix of sample numbers for clustering as shown in FIG. 3 can be established (step S240), where the important rates are all tested, so the number of samples corresponding to each sampling confidence level and the number of users in this user group the same.

The grouping sample decision module 115 then determines the number of samples of each user group according to the corresponding lease rates of those user groups. Specifically, the group sampling decision module 115 sets the initial sampling confidence level to 95% based on the group sampling matrix. The grouping sample decision module 115 calculates the sum of the minimum sampling number of all user groups (SP1~SP4) at the sampling rate at the sampling confidence level of 95% of the main rate, plus the total number of full inspections required for the important rate (ie, all Number of users), that is, the initial total number of samples (that is, the sum of the number of samples of all user groups) can be obtained (step S250).

The grouped sample decision module 115 determines whether the current total sampling number matches the total search capacity (step S260). The total inspection capacity is determined in advance and can be adjusted according to actual needs. If the total search capacity is not exceeded, the grouped sample decision module 115 records the results (step S265) and sequentially increases the sampling confidence level corresponding to those user groups of the main rate, so that the total sample number approaches the total search capacity , So as to maximize the efficiency of communication collection.

Specifically, the grouped sample decision module 115 can improve the sampling confidence level of different user groups. For example, the grouped sample decision module 115 first increases the sampling confidence level of the main rate SP1 from 95% to 99% for calculation, while the main rates SP2~SP4 still maintain the sampling confidence level of 95%. Next, the grouped sample decision module 115 adds the sample number after adjusting the sampling confidence level of the main rate and the total inspection number required by the important rate (SP5~SP8) to obtain the sample required by the new user group total. If the total number of samples still does not exceed the total search capacity, the cluster sample decision module 115 sets the importance of the attribute weight management module 112 for each lease rate, and increases the sampling confidence level of the main rate SP2 from 95% to 99 %, so that the main speeds SP1~SP2 will adopt 99% sampling. The grouped sample decision module 115 then adds the adjusted sample numbers of the main rates SP1~SP2, the number of samples maintained at 95% of the main rate SP3~SP4, and the total number of inspections required for the important rate (SP5~SP8). If the total number of samples still does not exceed the total inspection capacity, the grouped sample decision module 115 will continue to increase to the next major rate SP3 (next importance) until the total inspection capacity is exceeded, then the previous round of sampling The total number is the best solution for the number of samples of each group in each batch.

On the other hand, if the total number of samples exceeds the total inspection allowance, the grouped sample decision module 115 needs to reduce the total number of samples of the user group in a timely manner until the difference from the total inspection capacity is within the acceptable range, that is, the maximum information can be retained Collection efficiency. The grouping sample decision module 115 will also appropriately reduce the sampling confidence level of different user groups. For example, the grouping sample decision module 115 first reduces the sampling confidence level of the SP4 with the lowest main rate importance order from 95% to 90%, while the main rates SP1~SP3 still maintain the sampling rate of 95%. The grouped sample decision module 115 adds the total number of samples for the main rate and the total number of tests required for the important rate (SP5~SP8) to obtain the total number of samples required for the new user group. If the total number of total surveys is still exceeded, the cluster sample decision module 115 continues to reduce the sampling confidence level of the main rate SP3 from 95% to 90%, that is, the main rate SP3~SP4 will take 90% of the samples. The grouped sample decision module 115 adds the total number of samples of the main rate and the number of samples SP1~SP2 that maintain 95% of the main rate, and then adds up the total number of inspections (SP5~SP8) required for the important rate. If it is still exceeded, continue to decrease the SP2 to the next main rate until it is less than the total equipment inspection tolerance, then the total number of samples in this round is the best solution for the number of samples of each group in each round of inspection.

Taking Figure 3 as an example, if the initial sampling confidence level is 95% of the samples, it is assumed that the sum of all users is 97534 (less than the total check allowable value of 120000). Since the total number of samples does not exceed the allowable value of the total inspection, the group sample decision module 115 can appropriately increase the sampling confidence level of different groups. The cluster sample decision module 115 first increases the sampling confidence level of the SP1 with the highest main rate importance from 95% to 99%, then the sampling rate of the main rate will be taken from the SP1 with 99% of the main rate plus 95% with the main rate The total number of samples of SP2~SP4. The number of samples for the main rate will be added to the total number of inspections required for the important rate, that is, the total number of samples is 109198, but it is still less than the allowable value of the total inspection.

Since the total inspection tolerance value has not been exceeded, the cluster sample decision module 115 continues to increase the sampling confidence level of the main rate SP2 from 95% to 99%, then the sampling rate of the main speed is 99% of the main speed. The total number of samples of SP1~SP2 plus 95% of SP3~SP4 at the main rate. The number of samples for the main rate will be added to the total number of inspections required for the important rate, and the total number of samples is 120657, which exceeds the allowable value of the total inspection. The grouping sample decision module 115 can add 99% of the SP1 sampling rate of the main rate in the previous round plus 95% of the SP2~SP4 sampling rate of the main rate, and then add up the total number of inspections required for the important rate. That is, the total number of samples is 109198 (as shown in the sum of bold in Figure 4). The difference between the total number of samples and the total allowable value of the search is less than the threshold value (for example, 500, 1000, or 5000, etc.), which can be regarded as conforming to the allowable value of the total search, so it is also the best solution.

The grouping sample decision module 115 can obtain the number of samples for each user group based on this optimal solution. Taking user group 1 as an example, it can be seen that the data under the confidence level of 99% sampling is to be taken, and the number of samples taken by this group is 1608. The user groups 2~4 all need to use the data with 95% sampling confidence level. Taking user group 4 as an example, the number of samples taken by this group is 915. User groups 5 to 8 are fully investigated because they are important rates. Taking user group 6 as an example, the number of samples that this user group must take is 144 (equivalent to the total number of users).

Next, the polling user sampling and setting module 116 obtains the optimized sampling number of each user group (that is, the sampling number corresponding to each user group in the optimal solution), as each subsequent polling batch The number of sampling samples required by the user group. Next, the network management module 118 uses the communication module 130 to conduct a hierarchical random check on the user groups at the main rate, and to check all the user groups at the important rate in a non-repetitive sampling manner. List of secondary inspections (step S270). The round-robin user sampling and setting module 116 then performs batch-by-batch round-by-block round-robin inspection (step S285) until all the user lists in the planned main rate/important rate have been spot-checked.

For example, the polling user sampling and setting module 116 is based on the optimal number of samples for each sub-group, wherein the user groups of the main rate are sampled in layers, and the user groups of the important rate are fully checked. The polling user sampling and setting module 116 adopts non-repetitive sampling, arranges the check list for each batch, and then performs the batch-by-batch polling until all users are selected. Taking Figure 4 as an example, if the user group 1 adopts a 99% confidence level, if all users (12323) in user group 1 are scheduled according to the best sampling number 1608, they can be divided into 8 batches, which can cover users Group 1 all major rate/important rate users.

It is worth noting that the circuit change management module 117 will add/leave (circuit change) and service change users between two adjacent batches of polls in each batch of polls (may not conform to the original user group plan or Is a new user), included in the batch inspection list (ie, directly as a random inspection object), to reflect the immediacy and representativeness of each round of inspection (step S290).

Since each batch of continuous polling is measured by the best clustering and attribute importance, and the maximum sampling tolerance of the total inspection of the receiving equipment is used to maximize the sampling sample for testing, the automatic polling server 1 The obtained user information can instantly reflect the best representative of all users in proximity. At the same time, as the continuous polling progresses, by gradually expanding the coverage of the user matrix, it is possible to gradually improve the accuracy of collecting all user communication data and the quality of basic information required for big data prediction/analysis.

In summary, the automatic polling method and the automatic polling server of the broadband service according to the embodiments of the present invention have the following characteristics.

The embodiments of the present invention can automatically perform sampling inspection based on the performance of the access equipment and the collection equipment, and use a simple decision analysis method, thereby solving the problem that it is impossible to conduct an inspection of all users' communications in a short time.

In the embodiment of the present invention, groups can be grouped according to user attributes, and the individual minimum sampling numbers of different groups can be automatically evaluated to determine the number of samples in each batch of random inspections, thereby ensuring that each batch of rounds has the best representativeness.

The embodiments of the present invention can determine the accuracy and error level of each type of user group according to the importance of the user's rate, and determine whether to increase or decrease the number of samples.

The embodiment of the present invention is a pre-planned sampling inspection. If a user change (addition or cancellation of rent) occurs during the inspection of each batch, the next batch may be automatically supplemented and adjusted.

Although the present invention has been disclosed as above with examples, it is not intended to limit the present invention. Any person with ordinary knowledge in the technical field can make some changes and modifications without departing from the spirit and scope of the present invention. The scope of protection of the present invention shall be subject to the scope defined in the appended patent application.

1: Automated polling server 120: storage 111: attribute analysis module 112: attribute weight management module 113: user situation module 114: cluster sample number generator 115: cluster sample decision module 116: for polling Road sampling and setting module 117: circuit change management module 118: network management module 119: user list 130: communication module 150: processor S210~S290: steps

FIG. 1 is a block diagram of components of an automated polling server according to an embodiment of the invention. 2 is a flowchart of an automatic polling method according to an embodiment of the invention. FIG. 3 is an example illustrating the initial clustering sample number matrix. FIG. 4 is an example illustrating the optimized clustering sample number matrix.

S210~S290: Step 119: User list

Claims (8)

An automatic polling method for broadband service includes: grouping multiple users according to their user attributes to generate multiple user groups, wherein each user's user attributes include location and lease rate; based on the user groups Corresponding to the rental rate, determine the sampling number of each user group separately, including: use at least one of the users with the largest number of users corresponding to the rental rate in these user groups as the main rate; At least one of the rates that needs attention is regarded as an important rate; and at least one of the user groups that is not the primary rate and the important rate is regarded as a secondary rate; and the execution is based on the determined number of samples of the user groups Polling homework. The automatic polling method as described in item 1 of the patent application scope, after the step of using at least one of the primary rate and the important rate in these user groups as the secondary rate, further includes: Confidence level to set the sampling rate of the user groups of the primary rate; the sampling rate of the user groups of the important rate to 100%; and the sampling of the user groups of the secondary rate The number is set to zero. The automatic polling method as described in item 1 of the patent application scope, wherein after the step of setting the sampling rate of the user groups of the secondary rate to zero, it further includes: Determine whether the sum of the sampling numbers of the user groups conforms to a total search capacity; if the total search capacity is not exceeded, increase the sampling confidence level corresponding to the user groups of the main rate in sequence; and if If the total search capacity is exceeded, the sampling confidence level corresponding to the user groups of the main rate is sequentially decreased. The automatic polling method as described in item 1 of the scope of the patent application, wherein the step of performing the polling operation based on the determined number of samples of these user groups includes: changing the circuit between the two polling operations and the users with service changes As a random inspection object. An automatic polling server includes: a communication module to transmit and receive data; a storage to record multiple modules; and a processor to couple the communication module and the storage and access and load The modules recorded in the storage, and the modules include: an attribute analysis module, grouping multiple users according to their user attributes to generate multiple user groups, each of which is a user of the user Attributes include location and lease rate; a sub-group sample decision module, which determines the number of samples for each user group according to the corresponding lease rates of these user groups; a round of user sampling and configuration modules, based on the determined The number of samples of these user groups is used to perform the polling operation through the communication module; and an attribute weight management module to correspond the lease rate in these user groups At least one of the largest number of users is regarded as the main rate, and at least one of the user groups corresponding to the lease rate needs to be regarded as the important rate, and the user group is not the main rate and the important rate. At least one is used as the secondary rate. The automatic polling server as described in item 5 of the patent application scope, wherein the attribute weight management module sets the sampling rate of the user groups of the main rate based on multiple sampling confidence levels, The number of samples of these user groups is set to 100%, and the number of samples of these user groups of the secondary rate is set to zero. The automatic polling server as described in item 5 of the patent application scope, wherein the grouped sample decision module determines whether the sum of the sampling numbers of the user groups conforms to a total search capacity; and if the total search is not exceeded The measurement capacity increases the sampling confidence level corresponding to the user groups of the main rate in sequence; and if the total search capacity is exceeded, it sequentially decreases the sampling confidence level corresponding to the user groups of the main rate. The automatic polling server as described in item 5 of the patent application scope, wherein the polling user sampling and setting module selects users for circuit changes and service changes between the two polling operations as spot checks.

Images