TWI721685B - System and method for predicting customer complaint hot zone of mobile communication network - Google Patents

Abstract

A system and a method for predicting customer complaint hot zone of a mobile communication network are provided. The method includes: obtaining a plurality of historical user data and a plurality of current user data, wherein each one of the plurality of historical user data includes a communication log, position information, and a customer complaint record; generating a reference area according to the position information; generating a plurality of training data corresponding to the reference area according to the plurality of historical user data; training at least one identification model according to the plurality of training data; and estimating the customer complaint hot zone according to the at least one identification model and the plurality of current user data.

Description

System and method for predicting customer complaint hot spots of mobile network

The present invention relates to a system and method, and more particularly to a system and method for predicting the hot spots of customer complaints in a mobile network.

With the development of network technology and smart phones, the number of mobile network users has increased dramatically. In order to improve the user experience of mobile network users, telecommunications companies deal with the user’s problems immediately after the user submits a complaint. However, in the case of a large number of users, telecommunications companies are likely to miss the time to solve the problem due to lack of manpower or slow analysis of customer complaints, which will cause users to feel dissatisfied. This will lead to a decline in the evaluation of telecommunications companies.

The present invention provides a system and method for predicting customer complaint hotspots of mobile networks, which can prompt network managers to predict the customer complaint hotspots of mobile networks, and the network managers can target customer complaint hotspots in advance Mobile network to adjust or debug.

The system of the present invention for predicting the hot spots of customer complaints in a mobile network includes a processor, a storage medium, and a transceiver. The storage medium stores multiple modules. The processor is coupled to the storage medium and the transceiver, and accesses and executes a plurality of modules. The plurality of modules include a data collection module, a data grouping module, a training module, and a prediction module. The data collection module obtains multiple pieces of historical user data and multiple pieces of current user data through the transceiver. Each of the multiple pieces of historical user data includes communication logs, location information, and customer complaint records. The data grouping module generates a reference area based on the location information. The training module generates multiple training data corresponding to the reference area based on multiple historical user data, and trains at least one recognition model based on the multiple training data. The prediction module predicts the hot spots of customer complaints based on at least one identification model and multiple pieces of current user data.

In an embodiment of the present invention, the above-mentioned multiple modules further include feature extraction modules. The feature extraction module determines the features associated with the customer complaint event according to the customer complaint record and the communication log, wherein multiple pieces of training data correspond to the feature.

In an embodiment of the present invention, the aforementioned prediction module predicts the hot spots of customer complaints based on characteristics.

In an embodiment of the present invention, the aforementioned at least one identification model includes at least one of the following: a random forest model, a gradient-enhanced decision tree, and a support vector machine.

In an embodiment of the present invention, the aforementioned prediction module determines the customer complaint probability of the area according to at least one identification model, and determines that the customer complaint hot area includes the area in response to the customer complaint probability exceeding a threshold.

In an embodiment of the present invention, the above-mentioned at least one recognition model includes a first recognition model and a second recognition model, wherein the prediction module determines the first customer complaint probability of the area according to the first recognition model, and judges according to the second recognition model The second customer complaint probability of the area, and the customer complaint probability is judged based on the first customer complaint probability and the second customer complaint probability.

In an embodiment of the present invention, the above-mentioned location information includes the coordinates of the user equipment.

In an embodiment of the present invention, the aforementioned customer complaint record indicates whether each of the multiple pieces of historical user data is associated with the customer complaint event.

A method of the present invention for predicting the hot spots of customer complaints in a mobile network includes: obtaining multiple pieces of historical user data and multiple pieces of current user data, wherein each of the multiple pieces of historical user data includes communication logs, location information, and Customer complaint records; generate a reference area based on location information; generate multiple training data corresponding to the reference area based on multiple historical user data; train at least one recognition model based on multiple training data; and based on at least one recognition model and multiple current users The data predicts the hot spots for customer complaints.

Based on the above, the system and method for predicting customer complaint hotspots of a mobile network of the present invention can predict customer complaint hotspots with a high probability of occurrence of customer complaint events. Network administrators can refer to customer complaint hotspots to adjust or debug the mobile network in advance, thereby reducing the probability of customer complaints and improving the user experience of users.

FIG. 1 illustrates a schematic diagram of a system 100 for predicting the hot spots of customer complaints in a mobile network according to an embodiment of the present invention. The system 100 may include a processor 110, a storage medium 120, and a transceiver 130.

The processor 110 is, for example, a central processing unit (CPU), or other programmable general-purpose or special-purpose micro control unit (MCU), microprocessor, or digital signal processing Digital signal processor (DSP), programmable controller, application specific integrated circuit (ASIC), graphics processing unit (GPU), arithmetic logic unit (ALU) , Complex programmable logic device (CPLD), field programmable gate array (FPGA) or other similar components or a combination of the above components. The processor 110 may be coupled to the storage medium 120 and the transceiver 130, and access and execute multiple modules and various application programs stored in the storage medium 120.

The storage medium 120 is, for example, any type of fixed or removable random access memory (RAM), read-only memory (ROM), or flash memory. , Hard disk drive (HDD), solid state drive (SSD) or similar components or a combination of the above components, which are used to store multiple modules or various application programs that can be executed by the processor 110. In this embodiment, the storage medium 120 can store multiple modules including a data collection module 121, a data grouping module 122, a training module 123, a prediction module 124, and a feature extraction module 125. Its functions will be Follow-up instructions.

The transceiver 130 transmits and receives signals in a wireless or wired manner. The transceiver 130 may also perform operations such as low noise amplification, impedance matching, frequency mixing, up or down frequency conversion, filtering, amplification, and the like.

The data collection module 121 can obtain multiple pieces of historical user data corresponding to users of the mobile network through the transceiver 130, and each piece of historical user data can include information such as communication logs, location information, and customer complaint records. The location information indicates the geographic location information or coordinates of the user equipment when recording the historical user data. For example, the location information is the latitude and longitude of the location when the user equipment uploads historical user data. The customer complaint record indicates whether the historical user data is related to the customer complaint event. If the customer complaint record of a historical user profile indicates that the historical user profile is associated with the customer complaint event, the communication log representing the historical user profile may contain features associated with the customer complaint event. The communication log may include, for example, reference signals received power (RSRP), client experience index (CEI), CEI/RSRP ratio, data transmission volume (thrpt), CEI/thrpt, CEI/thrpt ratio, Evolved radio access bearer (E-RAB) packet loss (drop), CEI E-RAB packet loss ratio, handover failure rate (ho fail), outdoor or indoor and other information.

The data grouping module 122 can generate a reference area based on the location information and customer complaint records of historical user data. Fig. 2A illustrates a schematic diagram of an upload area of historical user data according to an embodiment of the present invention. FIG. 2B shows a schematic diagram of a reference area according to an embodiment of the present invention. In FIG. 2A, black dots representing the location information of historical user data occur in areas 210, 220, 230, 240, 250, and 260. The data grouping module 122 can select reference areas from the areas 210, 220, 230, 240, 250, and 260. For example, the data grouping module 122 may respond to the occurrence density of black dots in the area 220 being greater than or equal to the density threshold and the occurrence density of black dots in other areas (ie areas 210, 230, 240, 250, and 260) is less than the density threshold. The area 220 is determined as a reference area. The historical user data corresponding to the reference area will be used to train the recognition model.

The reference area selected by the data grouping module 122 represents a place where the user often stays. For example, the reference area is, for example, the user's office or home. These areas are usually the most likely locations for customer complaints.

The training module 123 can generate multiple pieces of training data corresponding to the reference area based on multiple pieces of historical user data. Specifically, the feature extraction module 125 can determine the features associated with the customer complaint event based on the customer complaint records of historical user data and the communication log. For example, if the communication logs of multiple historical user data corresponding to the customer complaint event all record the "RSRP too low event", then the level of the "RSRP" may have a high correlation with the customer complaint event. Therefore, the feature extraction module 125 can determine that “RSRP” is one of the features associated with the customer complaint event. After judging the features associated with the customer complaint event, the training module 123 can select historical user data corresponding to the reference area (ie: historical user data with location information in the reference area), and select historical user data from the selected historical user data. The characteristics associated with the customer complaint event are selected from each of them as training materials.

After generating multiple training data, the training module 123 can train at least one recognition model based on the multiple training data. The model type and the number of models of the at least one identification model can be adjusted according to usage requirements, and the present invention is not limited thereto. For example, the at least one identification model may include two identification models such as a first identification model and a second identification model, and the at least one identification model is, for example, a random forest model (RF) and a gradient boosting decision tree (gradient boosting decision tree). tree, GBDT) or support vector machine (support vector machine, SVM) and other types of models.

After generating at least one identification model, the data collection module 121 can receive multiple pieces of current user data through the transceiver 130. The prediction module 124 can predict customer complaint hotspots based on at least one recognition model and the multiple pieces of current user data. Specifically, the prediction module 124 can input the characteristics of the current user data into at least one recognition model, so as to use the at least one recognition model to determine whether the current user data is judged to correspond to the customer complaint event. After determining whether each piece of previous user data corresponds to a customer complaint event, the prediction module 124 can determine the customer complaint probability corresponding to each area according to the location information of the previous user data. If the customer complaint probability of a specific area exceeds the probability threshold, the prediction module 124 can determine that the specific area is a hot area for customer complaints.

In an embodiment, the at least one identification model includes two identification models, a first identification model and a second identification model. The prediction module 124 can determine the first customer complaint probability of a specific area according to the first recognition model, and determine the second customer complaint probability of the specific area according to the second recognition model. Then, the prediction module 124 can calculate the customer complaint probability corresponding to the specific area according to the first customer complaint probability and the second customer complaint probability. For example, the prediction module 124 may average the first customer complaint probability and the second customer complaint probability to calculate the customer complaint probability of the specific area.

FIG. 3 shows a flowchart of a method for predicting the hot spots of customer complaints in a mobile network according to an embodiment of the present invention, wherein the method may be implemented by the system 100 shown in FIG. 1. In step S301, multiple pieces of historical user data and multiple pieces of current user data are obtained, where each of the multiple pieces of historical user data includes communication logs, location information, and customer complaint records. In step S302, a reference area is generated based on the location information. In step S303, multiple pieces of training data corresponding to the reference area are generated based on multiple pieces of historical user data. In step S304, at least one recognition model is trained based on multiple pieces of training data. In step S305, the hot spots for customer complaints are predicted based on at least one identification model and multiple pieces of current user data.

In summary, the system and method for predicting customer complaint hotspots of mobile networks of the present invention can determine features associated with customer complaint events based on historical user data, thereby generating identification models, such as random forest models, Gradient enhancement decision tree or support vector machine, etc. The present invention can use the identification model to analyze the probability of customer complaints in each area, thereby predicting customer complaint hot spots where there will be extremely high chances of customer complaint events. Network administrators can refer to customer complaint hotspots to adjust or debug the mobile network in advance, thereby reducing the probability of customer complaints and improving the user experience of users.

100: A system for predicting the hot spots of customer complaints on mobile networks 110: processor 120: storage media 121: Data Collection Module 122: Data grouping module 123: Training Module 124: Prediction Module 125: Feature Extraction Module 130: Transceiver S301, S302, S303, S304, S305: steps

FIG. 1 illustrates a schematic diagram of a system for predicting the hot spots of customer complaints in a mobile network according to an embodiment of the present invention. Fig. 2A illustrates a schematic diagram of an upload area of historical user data according to an embodiment of the present invention. FIG. 2B shows a schematic diagram of a reference area according to an embodiment of the present invention. Fig. 3 shows a flow chart of a method for predicting the hot spots of customer complaints in a mobile network according to an embodiment of the present invention.

S301, S302, S303, S304, S305: steps

Claims (8)

A system for predicting customer complaint hotspots of a mobile network, comprising: a transceiver; a storage medium that stores a plurality of modules; and a processor, which is coupled to the storage medium and the transceiver, and accesses and executes The plurality of modules, wherein the plurality of modules include: a data collection module, through the transceiver to obtain a plurality of historical user data and a plurality of current user data, wherein each of the plurality of historical user data These include communication logs, location information, and customer complaint records; a data grouping module, which generates a reference area based on the location information; a feature extraction module, which determines the events related to the customer complaint event based on the customer complaint records and the communication log Features; a training module that generates multiple training data corresponding to the reference area based on the multiple historical user data, and trains at least one recognition model based on the multiple training data, wherein the multiple training data corresponds to The characteristics; and a prediction module, which predicts the hot spots of customer complaints based on the at least one identification model and the multiple pieces of current user data. The system according to item 1 of the scope of patent application, wherein the prediction module predicts the hot spot of customer complaints based on the characteristics. The system according to claim 1, wherein the at least one identification model includes at least one of the following: a random forest model, a gradient enhancement decision tree, and a support vector machine. The system according to item 1 of the scope of patent application, wherein the prediction module determines the customer complaint probability of the area according to the at least one identification model, and determines the customer complaint hot area in response to the customer complaint probability exceeding a threshold Including the area. The system according to item 4 of the scope of patent application, wherein the at least one identification model includes a first identification model and a second identification model, and the prediction module determines the first identification model of the region according to the first identification model. The customer complaint probability is to determine the second customer complaint probability of the area according to the second identification model, and the passenger complaint probability is determined according to the first customer complaint probability and the second customer complaint probability. The system described in the first item of the scope of patent application, wherein the location information includes the coordinates of the user equipment. The system according to claim 1, wherein the customer complaint record indicates whether each of the multiple pieces of historical user data is related to the customer complaint event. A method for predicting the hot spots of customer complaints on a mobile network includes: obtaining multiple pieces of historical user data and multiple pieces of current user data, wherein each of the multiple pieces of historical user data includes communication logs, location information, and customer information. Complaint record; generate a reference area based on the location information; determine the characteristics associated with the customer complaint event based on the customer complaint record and the communication log; Generate multiple training data corresponding to the reference area according to the multiple historical user data, wherein the multiple training data correspond to the feature; train at least one recognition model based on the multiple training data; and The at least one identification model and the multiple pieces of current user data predict the hot spots of customer complaints.

Images