Customer dissatisfaction predicting system and customer dissatisfaction predicting method

Abstract

To provide a customer dissatisfaction predicting system capable of predicting a future customer claim and a potential customer claim based on past customer claims.SOLUTION: A customer dissatisfaction predicting claim that predicts future and potential customer claims based on past customer claims includes: a data cleaning / preprocessing unit which collects data from multiple sources, and which maintains the uniformity of a time stamp so as to satisfy a missing value; a data encode unit that converts the data into a desired format for a machine learning model; a deep learning model that converts input data into necessary features; a first analyzing unit that searches the other case having a similar action to a customer's complaint case; a second analyzing unit that proposes a solution based on past complaints; and an action trigger block that prompts a user to take a necessary action based on the proposal by the second analyzing unit.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a system configuration and method for supporting business operations, and in particular to a technique that is effective when applied to a customer dissatisfaction prediction system that predicts future complaints from customers. [Background technology]

[0002] In recent years, IoT (Internet of Things) home appliances that have various functions added by connecting to the Internet have become increasingly popular and are becoming indispensable in daily life. IoT home appliances bring many benefits and improve the user experience many times over.

[0003] Therefore, if an IoT home appliance breaks down, it will disrupt daily life. To improve user experience, preventing potential future complaints has become a key challenge for many home appliance and electronics manufacturers.

[0004] Background art in this technical field includes, for example, technology such as that described in Patent Document 1. Patent Document 1 discloses "a server that generates the number and / or type of solutions to a new inquiry and / or a display mode thereof by utilizing ratings or an accumulation of ratings of solutions that have been useful to device users in the past."

[0005] Furthermore, Patent Document 2 discloses an "abnormality diagnosis and analysis device that finds the cause of erroneous or missed reports and supports improvement of the accuracy of machine abnormality diagnosis."

[0006] Furthermore, Patent Document 3 discloses "behavioral analysis for automating direct and / or indirect monitoring of the health of Internet of Things (IoT) devices." [Prior art documents] [Patent documents]

[0007] [Patent Document 1] JP 2019-175025 A [Patent Document 2] JP 2016-133944 A [Patent Document 3] Special Publication No. 2018-513457 Summary of the Invention [Problem to be solved by the invention]

[0008] However, if it is possible to detect appliances or parts that may fail in the future, future customer complaints can be prevented. Also, if it is possible to present suggested solutions to maintenance personnel based on past complaints, the target products can be repaired early, thereby improving customer satisfaction.

[0009] However, little has been developed in this field to date to prevent potential future complaints. Existing methods are primarily targeted at anomaly detection for faulty components.

[0010] For example, improper usage such as leaving the refrigerator door open for a long time or putting too much food in the refrigerator, or the environment in which the home appliance is installed can cause abnormal operation. These causes cannot be detected by conventional status monitoring or abnormality detection methods.

[0011] The above-mentioned Patent Documents 1 to 3 do not fully consider the above-mentioned problems, and there is room for improvement.

[0012] SUMMARY OF THE PRESENT EMBODIMENT An object of the present invention is to provide a customer dissatisfaction prediction system and method capable of predicting future and potential customer complaints based on past customer complaints. [Means for solving the problem]

[0013] In order to solve the above problems, the present invention provides a customer dissatisfaction prediction system that predicts future and potential customer complaints based on past customer complaints, comprising: a data cleaning / preprocessing unit that collects data from multiple sources, fills missing values ​​and maintains uniformity of timestamps; a data encoding unit that converts the data into a desired format for a machine learning model; a deep learning model that converts input data into required features; a first analysis unit that searches for other cases whose behavior is similar to the customer complaint case; a second analysis unit that suggests solutions based on past complaints; and an action trigger block that prompts necessary actions based on the suggestions of the second analysis unit.

[0014] The present invention also includes the steps of: (a) collecting data from multiple sources, satisfying missing values ​​and maintaining uniformity of timestamps; (b) converting the data processed in step (a) into a desired format for a machine learning model; (c) applying machine learning to the data converted into the desired format in step (b) to convert it into required features; (d) searching for other cases whose behavior is similar to the customer complaint case; (e) extracting solutions based on past complaints; and (f) encouraging necessary actions based on the solutions extracted in step (e). Effect of the Invention

[0015] According to the present invention, it is possible to realize a customer dissatisfaction prediction system and a customer dissatisfaction prediction method that can predict future and potential customer complaints based on past customer complaints.

[0016] This will enable the affected products to be repaired quickly, contributing to improved customer satisfaction.

[0017] Problems, configurations and effects other than those described above will become apparent from the following description of the embodiments. [Brief description of the drawings]

[0018] [Figure 1] FIG. 1 is a diagram illustrating the basic concept of the present invention. [Diagram 2] FIG. 1 is a diagram showing a schematic configuration of a customer dissatisfaction prediction system according to a first embodiment of the present invention. [Diagram 3] FIG. 1 is a diagram illustrating a method for acquiring IoT data. [Figure 4] FIG. 1 is a diagram illustrating a method for acquiring IoT data. [Diagram 5] FIG. 1 is a diagram illustrating an example of a sensor of an IoT home appliance. [Figure 6] FIG. 1 illustrates a method for pre-processing and validation of IoT data. [Figure 7] FIG. 1 illustrates the conversion of IoT data into other formats. [Figure 8] FIG. 1 illustrates an example of the configuration of a remote device for data processing. [Figure 9] FIG. 1 is a diagram showing a deep learning network structure. [Figure 10] FIG. 1 illustrates a method for proposing solutions based on past complaints. [Figure 11] FIG. 1 illustrates a method for predicting future complaints from past complaints. [Figure 12] FIG. 13 is a diagram showing a method of notifying a customer (user) and a repair person. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In the drawings, the same components are designated by the same reference numerals, and detailed description of overlapping parts will be omitted. EXAMPLES

[0020] A customer dissatisfaction prediction system and a customer dissatisfaction prediction method according to a first embodiment of the present invention will be described with reference to FIGS.

[0021] First, the basic concept of the present invention will be explained with reference to Fig. 1. Fig. 1 is a diagram showing the basic concept of the present invention.

[0022] As shown in Fig. 1, when a complaint is received from a customer (user) 2 via a communication method such as telephone or the Internet, data is obtained from the faulty IoT home appliance 3 that is the subject of the complaint, among IoT home appliances such as refrigerators, air conditioners, and washing machines. This data includes information such as the model number and manufacturing number (serial number) of the IoT home appliance 3, its operation history, failure / repair history, installation environment, and current operation status.

[0023] Next, the data acquired from the IoT home appliance 3 is encoded in a specific format. At this time, time-series data such as operation history and failure / repair history is also encoded.

[0024] The encoded data is then input into a deep learning-based network and machine-learned using a deep learning model.

[0025] Next, we analyze the output of the deep learning model based on past complaints and also analyze other IoT home appliances based on the complaints by observing similar IoT home appliances for a few hours or days.

[0026] Finally, based on the output analysis of the deep learning model, a solution to the complaint is proposed to the customer (user) 2. This proposal is a solution proposal based on past complaints. Also, if a fault repair is required, the repair technician is instructed to perform the necessary action.

[0027] A specific configuration of the customer dissatisfaction prediction system will be described with reference to Fig. 2. Fig. 2 is a diagram showing a schematic configuration of the customer dissatisfaction prediction system 1 of this embodiment.

[0028] 2, when a customer dissatisfaction prediction system 1 of this embodiment receives a new complaint from a customer (user) 2, it acquires IoT data 4 from an IoT home appliance 3 that is the subject of the complaint. Then, it encodes the acquired IoT data 4 to generate encoded data 5.

[0029] The customer dissatisfaction prediction system 1 performs machine learning on the encoded data 5 using a deep learning model. During this process, based on the features of the new complaint, similar cases are searched for from the features of past complaints from customers. Then, solutions to similar past complaints are extracted and a list of solutions is generated. The list of generated solutions is presented to the repair technician 6 (or the customer 2).

[0030] Furthermore, based on the characteristics of the new complaint, the customer dissatisfaction prediction system 1 searches for similar cases from the characteristics of past complaints about other home appliances other than the IoT home appliance 3 that is the subject of the complaint, which are collected from all customers of the IoT home appliance 3. Then, the system notifies other customers (users) of the extracted similar past cases as dissatisfaction that may occur in the future, and notifies the repair technician 6 of a new repair request if repairs were required for the similar past complaint.

[0031] A method of acquiring IoT data 4 from an IoT home appliance 3 will be described with reference to Fig. 3 and Fig. 4. Both Fig. 3 and Fig. 4 are diagrams showing a method of acquiring IoT data 4, and Fig. 4 is a modified example of Fig. 3.

[0032] As shown in Figure 3, IoT home appliance (air conditioner) 3a, IoT home appliance (refrigerator) 3b, and IoT home appliance (washing machine) 3c are all connected to a local wireless network 7, and each IoT data 4 of the IoT home appliances 3a to 3c is input to an IoT server 9 via the Internet 8, and also to a remote device 10 for data processing.

[0033] The remote data processing device 10 is incorporated into a customer dissatisfaction prediction system 1, which uses data encoded by the remote data processing device 10 to predict future customer complaints (dissatisfaction).

[0034] The IoT server 9 is connected to a remote data processing device 10 via the Internet 8, a local area network (LAN), etc., and the IoT data 4 stored in the IoT server 9 is read out as needed, encoded by the remote data processing device 10, and used to predict future customer complaints (dissatisfaction).

[0035] Here, the customer dissatisfaction prediction system 1 is described as including a remote device 10 for data processing, but the customer dissatisfaction prediction system 1 may also include an IoT server 9 together with the remote device 10 for data processing.

[0036] In addition to the method of automatically collecting IoT data 4 as shown in Fig. 3, a repairman 6 may physically collect various data on IoT home appliances 3a to 3c and use the data to predict future customer complaints (dissatisfaction) as shown in Fig. 4. Alternatively, without automatically collecting IoT data 4, only the data physically collected by repairman 6 may be used to predict future customer complaints (dissatisfaction).

[0037] A sensor for acquiring the IoT data 4 will be described with reference to Fig. 5. Fig. 5 is a diagram showing an example of a sensor of an IoT home appliance (refrigerator) 3b.

[0038] Taking a refrigerator as an example, IoT home appliance (refrigerator) 3b is equipped with sensors such as sensor 11a related to the opening and closing of the door of each compartment, sensor 11b related to the temperature of each compartment, sensor 11c related to the operation of the compressor, sensor 11d related to the operating state such as rapid freezing and thawing, and sensor 11n related to the external environment such as temperature and humidity, and can acquire IoT data 4 related to the operating state of IoT home appliance (refrigerator) 3b, the presence or absence of a malfunction, the type of malfunction, the installation environment, etc.

[0039] The IoT data 4 acquired by each sensor is temporarily stored in a local data storage 12 in the home appliance and is then transmitted to an IoT server 9 or the like for data processing.

[0040] A method for pre-processing and verifying the IoT data 4 will be described with reference to Fig. 6. Fig. 6 is a diagram showing a method for pre-processing and verifying the IoT data.

[0041] As shown in Fig. 6, IoT data 4 can be collected from IoT home appliances 3 in various types of file formats. The raw data collected from IoT home appliances 3 is pre-processed and verified in the remote device 10 for data processing before being encoded.

[0042] The raw data entered into the remote data processing device 10 has blanks and missing values ​​replaced, then converted from text data to numeric values. If necessary, other processing options are applied and sorting based on timestamps (time data) is performed. The processed data in the remote data processing device 10 is output as cleaned and verified data.

[0043] Encoding of the IoT data 4 will be described with reference to Fig. 7. Fig. 7 is a diagram showing conversion of the IoT data 4 into another format.

[0044] As shown in Fig. 7, the encoding block 13 of the remote device 10 for data processing encodes the input verified IoT data 4 based on the information input by the system operator about the system design such as (1) the type and number of sensors to be used, and (2) the type of encoding to be used. The encoding format can be graphic images, pixel-by-pixel images, time series data, etc.

[0045] A specific configuration of the remote device 10 for data processing will be described with reference to Fig. 8. Fig. 8 is a diagram showing an example of the configuration of the remote device 10 for data processing.

[0046] When the remote data processing device 10 receives IoT data 4 collected from IoT home appliances 3, the data processing and verification block cleans and pre-processes the raw data, and performs data verification to verify the continuity, including correlation, of each piece of data.

[0047] The remote data processing device 10 determines the encoding method based on the system information (information on system design) input by the system manager. Then, the cleaned up and verified data is encoded and output in the form of a graphic image, a pixel-by-pixel image, time series data, etc. The encoded data is subjected to machine learning using a deep learning model.

[0048] The deep learning model will be described with reference to Fig. 9. Fig. 9 is a diagram showing a network structure of deep learning.

[0049] This block consists of a deep learning model that satisfies the following input / output conditions:

[0050] Input: The input can be either encoded data or directly entered data.

[0051] Output: Consists of a latent feature vector that, when fed to a decoder, recreates the input data with reasonable accuracy.

[0052] Note that the encoder / decoder can be part of a larger network that satisfies input / output requirements.

[0053] The data trained by the deep learning model 14 is output as an output feature vector.

[0054] The analysis method for proposing solutions to customer complaints will be described with reference to Fig. 10. Fig. 10 is a diagram showing a method for proposing solutions based on past complaints.

[0055] Based on new complaints learned by machine learning as IoT data4, we search for similar / approximate cases in the feature space based on the characteristics of customers' past complaints.

[0056] Based on similar / close past complaints extracted from the characteristics of the customer's past complaints, previous solutions are extracted and presented to the repair technician 6.

[0057] An analysis method for proposing the possibility of a future complaint will be described with reference to Fig. 11. Fig. 11 is a diagram showing a method for predicting a future complaint from a past complaint.

[0058] Based on new complaints learned by machine learning as IoT data4, we search for similar / approximate cases in the feature space from the characteristics of currently used devices (IoT home appliances).

[0059] Based on similar / approximate complaints extracted from the characteristics of currently used devices (IoT home appliances), the devices are observed for a certain period of time, and are classified into devices that are still similar and devices that are no longer similar. Based on past complaints about devices that are still similar, actions such as notifying a repair technician 6 are performed.

[0060] A method of notifying a customer (user) and a repair person will be described with reference to Fig. 12. Fig. 12 is a diagram showing a method of notifying a customer (user) 2 and a repair person 6.

[0061] If a similar device is still found after a certain period of observation, this may be due to negligence such as incorrect use by the customer (user) 2, so a notification is sent to the customer (user) 2 regarding the correct operation. Also, this may be due to a malfunction of the device (IoT home appliance), so after notifying the customer (user) 2 about the malfunction of the device (IoT home appliance), a repair request is sent to the repair person 6.

[0062] As described above, the customer dissatisfaction prediction system 1 of this embodiment is a customer dissatisfaction prediction system that predicts future and potential customer complaints based on past customer complaints, and includes a data cleaning / preprocessing unit that collects IoT data 4 from multiple sources (IoT home appliances 3) and satisfies missing values ​​and maintains uniformity of timestamps, a data encoding unit (encoding block 13) that converts the IoT data 4 into a desired format for a machine learning model, a deep learning model 14 that converts the input data into required features, a first analysis unit that searches for other cases whose behavior is similar to the customer complaint case, a second analysis unit that proposes solutions based on past complaints, and an action trigger block that prompts necessary actions based on the proposals of the second analysis unit.

[0063] This makes it possible to predict future and potential customer complaints based on past customer complaints, enabling early repair of affected products and contributing to improved customer satisfaction.

[0064] The present invention is not limited to the above-described embodiments, and includes various modified examples. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those including all of the configurations described. It is also possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. It is also possible to add, delete, or replace a part of the configuration of each embodiment with another configuration. [Explanation of symbols]

[0065] 1. Customer dissatisfaction prediction system 2...Customer (user) 3,3a,3b,3c…IoT home appliances 4. IoT data 5. Encoded data 6. Repairman 7. Local wireless network 8. Internet 9. IoT server 10...Remote device for data processing 11a, 11b, 11c, 11d, 11n...Sensors 12. Local Data Storage 13...Encoding block 14…Deep learning model.

Claims

1. A customer dissatisfaction prediction system for predicting future and potential customer complaints based on past customer complaints, comprising: a data cleaning / pre-processing unit that collects data from multiple sources, fills in missing values ​​and maintains uniformity of timestamps; a data encoding unit for converting the data into a desired format for a machine learning model; A deep learning model that converts input data into the required features, A first analysis unit for searching other cases whose behavior is similar to that of the customer complaint case; a second analytical unit that proposes solutions based on past complaints; an action trigger block for prompting a necessary action based on the proposal of the second analysis unit; A customer dissatisfaction prediction system comprising:

2. 2. The customer dissatisfaction prediction system according to claim 1, The data encoding unit converts the data into a desired format based on the input type and number of sensors to be used and the encoding format to be used.

3. 2. The customer dissatisfaction prediction system according to claim 1, The action trigger block notifies the customer of a correct operating method.

4. 2. The customer dissatisfaction prediction system according to claim 1, The action trigger block notifies a repair person of a repair request and presents a solution based on past complaints.

5. A customer dissatisfaction prediction method that includes the following steps: (a) collecting data from multiple sources, filling in missing values ​​and maintaining uniformity of timestamps; (b) converting the data processed in step (a) into a desired format for a machine learning model; (c) subjecting the data converted into a desired format in the step (b) to machine learning to convert it into required features; (d) searching for other cases whose behavior is similar to the customer complaint case; (e) extracting solutions based on past complaints; (f) prompting a necessary action based on the solution extracted in the (e) step.

6. 6. A method for predicting customer dissatisfaction according to claim 5, comprising: A method for predicting customer dissatisfaction, characterized in that in the step (b), the data is converted into a desired format based on the input type and number of sensors to be used and the encoding format to be used.

7. 6. A method for predicting customer dissatisfaction according to claim 5, comprising: A method for predicting customer dissatisfaction, characterized in that in the step (f), the customer is notified of a correct operating method.

8. 6. A method for predicting customer dissatisfaction according to claim 5, comprising: A method for predicting customer dissatisfaction, characterized in that in the step (f), a repair person is notified of the repair request and a solution based on past complaints is presented.

Images