Channel complaint risk determination method, device, equipment, medium and product

Abstract

The application discloses a channel complaint risk determination method, device, equipment, medium and product. The channel complaint risk determination method comprises the following steps: obtaining target channel information corresponding to a target channel and target user information corresponding to a target user, wherein the target user is a user who handles a business in the target channel; performing feature extraction on the target channel information and the target user information to obtain target complaint risk features corresponding to the target channel; inputting the target complaint risk features into a complaint risk assessment model, performing complaint risk assessment on the target channel according to the target complaint risk features through the complaint risk assessment model, and obtaining a complaint risk assessment result corresponding to the target channel. According to the embodiment of the application, the accuracy of complaint risk assessment prediction can be improved.

Description

Technical Field

[0001] This application belongs to the field of information technology, and in particular relates to a method, apparatus, equipment, medium and product for determining channel complaint risks. Background Technology

[0002] As society increasingly focuses on protecting user rights, businesses need to safeguard customers' right to know and right to fair trade. In the telecommunications industry, complaints about irregularities in service processing account for a high percentage, representing a significant reason for customer dissatisfaction. Businesses need to effectively utilize complaint information for scientific analysis. For example, by conducting complaint risk assessments, they can promptly identify and rectify irregularities that are likely to trigger customer complaints, thereby reducing the complaint rate, enhancing brand image, and ensuring the healthy development of their business.

[0003] Currently, risk assessments for corporate complaints typically focus only on channel behavior characteristics. However, besides irregularities in channel operations, the vulnerability of complaining users is also a significant factor in their decision to file a complaint. Therefore, assessing risk solely based on channel behavior characteristics fails to accurately predict corporate complaint risks, thus reducing the accuracy of complaint risk prediction. Summary of the Invention

[0004] This application provides a method, apparatus, equipment, medium, and product for determining channel complaint risk, which can improve the accuracy of complaint risk assessment and prediction.

[0005] In a first aspect, embodiments of this application provide a method for determining channel complaint risk, the method comprising:

[0006] Obtain target channel information and target user information corresponding to the target channel, where the target user is the user who conducts business through the target channel;

[0007] Feature extraction is performed on target channel information and target user information to obtain target complaint risk characteristics corresponding to the target channel;

[0008] The target complaint risk characteristics are input into the complaint risk assessment model. The complaint risk assessment model assesses the complaint risk of the target channel based on the target complaint risk characteristics, and obtains the complaint risk assessment result corresponding to the target channel.

[0009] Secondly, embodiments of this application provide a channel complaint risk determination device, the device comprising:

[0010] The first acquisition module is used to acquire target channel information and target user information corresponding to the target channel and the target user, where the target user is the user who conducts business through the target channel;

[0011] The first extraction module is used to extract features from target channel information and target user information to obtain target complaint risk features corresponding to the target channel;

[0012] The input module is used to input the target complaint risk characteristics into the complaint risk assessment model. The complaint risk assessment model assesses the complaint risk of the target channel based on the target complaint risk characteristics and obtains the complaint risk assessment result corresponding to the target channel.

[0013] Thirdly, embodiments of this application provide an electronic device, which includes: a processor and a memory storing computer program instructions;

[0014] When the processor executes the computer program instructions, it implements the steps of the channel complaint risk determination method as described in any embodiment of the first aspect.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps of the channel complaint risk determination method as described in any embodiment of the first aspect.

[0016] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by a processor of an electronic device, cause the electronic device to perform the steps of the channel complaint risk determination method as described in any embodiment of the first aspect.

[0017] The channel complaint risk determination method, apparatus, equipment, medium, and product in this application embodiment extracts features from acquired target channel information and target user information to obtain target complaint risk characteristics corresponding to the target channel. Based on these target complaint risk characteristics, a complaint risk assessment model is used to assess the complaint risk of the target channel, yielding a complaint risk assessment result corresponding to the target channel. Therefore, when assessing complaint risk for enterprises, this application embodiment considers the impact of target user susceptibility on complaint risk assessment by including not only target channel information but also target user information corresponding to target users conducting business through the target channel, thus improving the accuracy of complaint risk assessment prediction. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a method for determining channel complaint risk provided in an embodiment of this application;

[0020] Figure 2 This is a flowchart illustrating another method for determining the risk of complaints through an embodiment of this application;

[0021] Figure 3 A flowchart for identifying responsible complaint information is provided in several embodiments of this application;

[0022] Figure 4 This is a model structure diagram provided in an embodiment of this application;

[0023] Figure 5 This is a schematic diagram of a channel complaint risk determination device provided in an embodiment of this application;

[0024] Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0025] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0026] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0027] Currently, when assessing enterprise complaint risks, the evaluation typically focuses only on channel behavior characteristics. However, besides irregularities in channel operations, the susceptibility of complaining users is also a crucial factor in determining the cause of complaints. Furthermore, current complaint risk assessments, failing to uncover deeper connections between user characteristics and focusing solely on channel behavior, cannot accurately predict enterprise complaint risks, thus reducing the accuracy of complaint risk prediction.

[0028] To address the problems in the prior art, embodiments of this application provide a method, apparatus, equipment, medium, and product for determining channel complaint risks.

[0029] The method for determining channel complaint risks provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0030] Figure 1 This is a flowchart illustrating a method for determining channel complaint risk provided in an embodiment of this application. Figure 1 As shown, the method for determining the complaint risk of this channel may specifically include the following steps:

[0031] S110. Obtain the target channel information corresponding to the target channel and the target user information corresponding to the target user, wherein the target user is the user who conducts business through the target channel;

[0032] S120. Extract features from target channel information and target user information to obtain target complaint risk features corresponding to the target channel;

[0033] S130. Input the target complaint risk characteristics into the complaint risk assessment model, and use the complaint risk assessment model to assess the complaint risk of the target channel based on the target complaint risk characteristics to obtain the complaint risk assessment result corresponding to the target channel.

[0034] Therefore, by extracting features from the acquired target channel and target user information, target complaint risk characteristics corresponding to the target channel are obtained. Based on these target complaint risk characteristics, a complaint risk assessment model is used to assess the complaint risk of the target channel, yielding a complaint risk assessment result corresponding to the target channel. Thus, in this embodiment of the application, when assessing the complaint risk of an enterprise, the target complaint risk characteristics corresponding to the target channel include not only the target channel information but also the target user information corresponding to the target users conducting business through the target channel. This takes into account the impact of the target users' susceptibility on the complaint risk assessment, improving the accuracy of complaint risk assessment prediction.

[0035] The specific implementation methods for each of the above steps are described below.

[0036] In some implementations, in S110, the target channel involved in this application embodiment can be any channel subject to complaint risk assessment. The target channel information can include the basic channel information and business processing behavior information of the target channel in different risk scenarios. The basic channel information may include, for example, channel type, channel personnel size, channel business processing scale, channel city level, channel revenue level, and channel billing level. Various risk scenarios for complaints related to irregular business processing may include, for example, abnormal use of multiple phone numbers with one ID, abnormal ordering of older mobile phones, irregular outbound calls, batch business processing, high-frequency business processing, business processing outside of working hours, and card issuance and binding business.

[0037] In addition, target users can be users who conduct business through target channels. Target user information can be user identity information and business usage information related to the target channel business. For example, the target user information can be the identity information of channel subscribers, package details, tariff usage, call and data usage, recharge and payment, business inquiry, business binding and service satisfaction, etc.

[0038] In some implementations, during S120, the target complaint risk characteristics involved in this application embodiment may include channel-dimensional characteristics corresponding to the target channel and user-dimensional characteristics corresponding to the target user. The channel-dimensional characteristics corresponding to the target channel may include, for example, basic characteristics corresponding to the channel's basic information, and may also include business processing characteristics calculated based on the channel dimension for different risk scenarios. The user-dimensional characteristics corresponding to the target user may include identity characteristics corresponding to the target user's identity information, and may also include business usage characteristics corresponding to business usage information.

[0039] In some implementations, in S130, after the complaint risk assessment model assesses the complaint risk of the target channel based on the target complaint risk characteristics, the resulting complaint risk assessment result can be the degree of complaint risk corresponding to the target channel, such as high risk, medium risk, or low risk.

[0040] To obtain the complaint risk model, this application also provides another embodiment of the method for determining channel complaint risk. For example... Figure 2 As shown, the method may further include:

[0041] S210. Obtain historical user complaint information;

[0042] S220. Identify target user complaint information for which the channel is responsible from historical user complaint information;

[0043] S230. Obtain information on the responsible channels corresponding to the complaint information of the target user, and information on the complainant corresponding to the complaint information of the target user.

[0044] S240. Extract features from the information of responsible channels and complainant information to obtain the complaint risk features corresponding to the responsible channels;

[0045] S250. Determine the target complaint risk level corresponding to the responsible channels;

[0046] S260. Using the complaint risk characteristics corresponding to the responsible channels as training samples and the target complaint risk level as the sample label corresponding to the training samples, the target deep learning model is trained to obtain the complaint risk assessment model.

[0047] Therefore, by identifying the responsible target user complaint information from the acquired historical user complaint information, and obtaining the responsible channel information corresponding to the target user complaint information, as well as the complaint user information corresponding to the target complaint user information, feature extraction is performed on the responsible channel information and the complaint user information to obtain the complaint risk features corresponding to the responsible channel, and the target complaint risk level is determined. By using the complaint risk features as training samples and the target complaint risk level as the sample label of the training samples, the target deep learning model is trained to obtain the complaint risk assessment model. Thus, this embodiment of the application constructs a complaint risk prediction model from the two dimensions of channel and user features through a deep learning model. In this way, when enterprises conduct complaint risk assessment based on the complaint risk assessment model, the influence of the target user's susceptibility on the complaint risk assessment is avoided, and the accuracy of complaint risk assessment prediction is improved.

[0048] In some embodiments, in S210, the historical user complaint information can be historical complaint data of all users on the network. This historical complaint information can be, for example, the text content of a user's historical complaints. The text content can be the text record obtained after voice recognition of a user's complaint call, or it can be the complaint text content entered by the user through filling out a form or other means.

[0049] In some embodiments, in S220, the target user complaint information may be channel-responsible complaint data after excluding non-channel violation operation complaints such as mobile signal issues and system vulnerabilities based on historical user complaint information.

[0050] In some embodiments, in order to determine the target user complaint information from historical user complaint information, the above-mentioned S220 may specifically include:

[0051] Historical user complaint information is input into the responsible channel identification model. The responsible channel identification model is used to identify whether the channel corresponding to the historical user complaint information is responsible, and the identification result is obtained. The responsible channel identification model is constructed by the TRANSFORMER model. The identification result is used to characterize whether the historical user complaint information is information caused by the channel's violation of regulations.

[0052] Based on the identification results, target user complaint information for which the channel is responsible is determined from historical user complaint information.

[0053] As an example, such as Figure 3 The process for identifying responsible target user complaint information, as shown, is based on a responsible channel identification model built using the TRANSFORMER model. Before identifying whether a channel is responsible, a small number of samples from historical user complaint information need to be labeled. Complaints involving channel risk are labeled as 0, representing channel responsibility, while complaints not involving channel risk are labeled as 1, representing channel no responsibility. The labeled samples and the samples to be identified from historical user complaint information are then preprocessed. For example, all content in the complaint information content field of the sample dataset can be preprocessed, including removing punctuation, HTML tags, and lowercase conversion. A stop word list is then created and removed. Different word segmenters are used to segment the complaint content, ultimately balancing recall and precision. The word segmenter is determined based on prediction accuracy, and words are mapped to index representations, constructing a vocabulary-index mapping table and outputting the model dataset. The complaint content information processing results are shown in Table 1.

[0054] Table 1

[0055]

[0056] After preprocessing historical user complaint information, the responsible channel identification model built using the TRANSFORMER model identifies the responsible user complaints from the historical user complaint information. For example, the above model dataset can be input into the Inputs module of the TRANSFORMER model for Embedding and positional encoding. Pre-trained word vectors are used to initialize the word embedding matrix, converting the words in the input model data into word vectors. The resulting matrix is ​​then input into the Encoder-Decoder module to process sequence-to-sequence problems. The Encoder module consists of a Multi-Head Attention module and a Feed-Forward Network module. After passing through Multi-Head Attention, the input matrix is ​​concatenated into multiple matrices. It then enters the Add and Normalize modules. Add introduces residual blocks to prevent decay during neural network training, while Normalize uses LN to normalize the data in each dimension, improving training speed and stability. Finally, after entering the fully connected Feed-Forward Network module and the Add and Normalize modules, the output is sent to the next Encoder. After passing through six encoder modules, the data is input into the decoder module. The decoder module adds a Masked Multi-Head Attention module compared to the encoder module. During matrix transformation, due to varying sequence lengths, certain values ​​are masked to reduce the negative impact of parameter updates. Finally, the output module provides two classification results: information where the channel is responsible and information where the channel is not responsible. This means that the target user complaint information where the channel is responsible is determined from historical user complaint information.

[0057] Furthermore, to improve the accuracy of the responsible channel identification model during training, the training parameters, model parameters, number of categories, and dataset partitioning ratio can be adjusted. For example, the inner layer one-dimensional convolutional kernel filters can be set to 128 to ensure that the output dimension and input dimension of each layer are consistent. The number of outer convolutional kernels can be set to equal the embedding size. The number of multi-head attention heads can be set to 8 to improve the model's long-range information capture capability. Since the complaint results are a binary classification problem, numClasses can be set to 1. A training set ratio of 0.8 is selected for training, reserving 0.2 for accuracy confirmation during model training. After training, the model performance can be evaluated using performance metrics such as precision, recall, variance, and standard deviation. Training can be terminated when the precision and recall of the model output meet preset conditions.

[0058] Finally, in order to obtain more accurate recognition results, the model parameters can be continuously optimized based on the recognition results of the labeled samples. As a result, the user complaint information recognition results shown in Table 2 can be obtained.

[0059] Table 2

[0060]

[0061] In some embodiments, in S230, after identifying historical user complaint information based on the responsible channel identification model, target user complaint information can be obtained, and responsible channel information and complainant information can be obtained based on the target user complaint information.

[0062] In some embodiments, in S240, based on the responsible channel information and the complainant user information, features can be extracted from two aspects: the channel dimension and the user dimension involved in the channel. For example, the complaint risk features corresponding to the target channel and the target user respectively involved in S120 above are used to obtain the complaint risk features corresponding to the responsible channel.

[0063] In some embodiments, in S250, the target complaint risk level involved in the embodiments of this application may be determined by an experienced person or by calculation and analysis based on a comprehensive data sample.

[0064] Based on this, in order to determine the target complaint risk level corresponding to the responsible channel, the aforementioned S250 may specifically include:

[0065] Obtain the complaint rates corresponding to multiple historical responsible channels within a preset historical time period;

[0066] Based on the complaint rates corresponding to multiple historical responsible channels, the complaint risk level is divided into different levels, and the complaint rate thresholds corresponding to each of the multiple complaint risk levels are obtained.

[0067] Obtain the target complaint rate corresponding to the responsible channel, and determine the complaint risk level of the target complaint rate based on the complaint rate thresholds corresponding to multiple complaint risk levels. This is the target complaint risk level corresponding to the responsible channel.

[0068] In some embodiments, the preset historical time period involved in this application can be, for example, the complaint rate of multiple responsible channels over the past year. The complaint rate can be obtained based on the total number of complaints and the total business volume handled by the responsible channels within the preset historical time period, or it can be obtained by weighting the complaint percentage of each month within the preset historical time period of one year, for example. The complaint rate thresholds corresponding to multiple complaint risk levels can be determined, for example, based on the overall complaint rate distribution.

[0069] As an example, obtaining the complaint rates corresponding to multiple historical responsible channels within a preset historical time period can specifically include:

[0070] For each of the multiple historically accountable channels within a preset historical time period, perform the following steps to obtain the complaint rate corresponding to each of the multiple historically accountable channels:

[0071] Determine the proportion of complaints received by the historical accountability channel in each time period of the preset historical time period to the total business volume handled by the channel, and obtain the complaint proportions corresponding to the multiple time periods included in the preset historical time period.

[0072] Based on the weight corresponding to each time period in multiple time periods, the complaint percentages corresponding to each of the multiple time periods are weighted and summed to obtain the complaint rate corresponding to the historical responsible channels.

[0073] Thus, for example, by determining the proportion of complaints received by the historically responsible channel in each month within a one-year period relative to the channel's total business volume, the complaint percentage for each month within a one-year period can be determined according to the following formula (1). Then, according to the following formula (2), the complaint percentages for each month are weighted and summed according to their respective weights. The weights can be calculated by increasing the weight of the most recent month, ultimately yielding the complaint rate corresponding to the historically responsible channel. Here, month T represents the current month, and a1, a2, and a3 are the weights for month T, month T-1, and month T-2, respectively, satisfying a1 > a2 > a3 and a1 + a2 + a3 = 1.

[0074] Complaint percentage calculation = Number of complaints / Number of transactions processed through the channel × 100% (1)

[0075] Overall complaint rate = Complaint percentage (month T) × a1 + Complaint percentage (month T-1) × a2 +

[0076] Complaint percentage (T-2 months) × a3 + …(2)

[0077] In some embodiments, the aforementioned multiple complaint risk levels may include low risk, medium risk, and high risk.

[0078] Based on this, the above-mentioned classification of complaint risk levels according to the complaint rates corresponding to the multiple historical responsible channels can specifically include:

[0079] The complaint rates corresponding to multiple historical accountability channels were statistically analyzed to determine the distribution of complaint rates corresponding to multiple historical accountability channels;

[0080] The complaint rate value corresponding to the target quantile in the complaint rate distribution is used as the complaint rate threshold between medium risk and high risk.

[0081] As an example, in determining the complaint rate threshold between medium and high risk, the distribution of the complaint rate can be determined based on the complaint rates corresponding to multiple historical responsible channels. According to the overall distribution trend of the complaint rate, the target quantile in the complaint distribution, such as the 70th percentile of the inflection point of the complaint rate distribution, can be used as the complaint rate threshold between medium and high risk, for example, the corresponding complaint rate could be 0.00875%.

[0082] Furthermore, after determining the complaint rate threshold between medium and high risk, the target complaint rate corresponding to the responsible channel can be used to determine the complaint risk level of the target complaint rate based on the complaint rate threshold, thus determining the target complaint risk level corresponding to the responsible channel.

[0083] In some embodiments, in S260, after extracting complaint risk features and determining the target risk level of the responsible channel, the complaint risk features are used as training samples, and the target complaint risk level is used as the sample label corresponding to the training samples to train the target deep learning model, thereby obtaining the complaint risk assessment model. The target deep learning model can be trained using the PyTorch deep learning framework.

[0084] Additionally, in some embodiments, the method may further include the following to obtain a low-risk complaint marker:

[0085] Obtain information on complaint-free channels that have not received any complaints within a preset historical time period, as well as information on complaint-free users who have conducted business through complaint-free channels.

[0086] Feature extraction is performed on information from channels without complaints and information from users without complaints to obtain complaint risk features corresponding to channels without complaints.

[0087] The complaint risk level corresponding to the lack of a complaint channel is marked as low risk.

[0088] As an example, by obtaining information on complaint-free channels and complaint-free users within a preset historical time period, such as one year, and after extracting features from the information on complaint-free channels and complaint-free users, the complaint risk characteristics corresponding to the complaint-free channels can be obtained, and the complaint risk level corresponding to the complaint-free channels can be marked as low risk.

[0089] Based on this, in order to train the target deep learning model to obtain the complaint risk assessment model, the above-mentioned S260 may specifically include:

[0090] The complaint risk characteristics corresponding to the responsible channels are used as negative samples in the training sample, and the target complaint risk level is used as the sample label corresponding to the negative sample.

[0091] The complaint risk characteristics corresponding to the non-complaint channels are used as positive samples in the training samples, and the low-risk complaint risk level is used as the sample label corresponding to the positive samples.

[0092] The target deep learning model is trained based on negative samples and their corresponding sample labels, and positive samples and their corresponding sample labels, to obtain the complaint risk assessment model.

[0093] In some embodiments, when training the target deep learning model, the target deep learning model is trained based on negative samples and their corresponding sample labels and positive samples and their corresponding sample labels.

[0094] In addition, it is necessary to supplement the channel data with a small sample size by considering the number of channels corresponding to the target complaint risk level and the number of channels for low-risk complaint risk levels in the sample. This results in the complaint sample shown in Table 3. This balances the number of channels corresponding to medium-risk, high-risk, and low-risk complaint risk levels, ensuring that the amount of training data for different categories is relatively equal, thus making the model training more accurate.

[0095] Table 3

[0096] 102xxx731 high 2 328xxx351 middle 1 231xxx122 Low 0 … … …

[0097] In some embodiments, the target deep learning model described above may include an input layer, two hidden layers, two BN (Batch Normalization) layers, and an output layer.

[0098] As an example, such as Figure 4As shown, the target deep learning model can be trained using the PyTorch deep learning framework until the model converges. The model structure can include an input layer, two hidden layers, two batch normalization (BN) layers, and an output layer. Given 60 target complaint risk features, an input layer with 60 neurons can be created. For the hidden layer network settings, depending on the scale of the target channel data, three different hidden layer networks with 10, 60, and 100 network nodes can be tried. More network nodes result in greater network capacity, and a neural network with more neurons can express more complex functions and better fit complex complaint prediction information. However, more neurons may also lead to overfitting of the training data.

[0099] Further testing revealed that while a 10-neuron network converged quickly, it resulted in a higher loss value and low accuracy in predicting channel complaint samples. Networks with 60 and 100 neurons showed no significant difference in convergence and could encompass the data characteristics of the sample. However, larger neurons led to slower convergence and operational efficiency issues. Therefore, 60 hidden neurons were chosen as the target number of network nodes for the deep learning model. A batch normalization (BN) layer was added to prevent overfitting. Three neurons were used in the output layer to assess complaint risk. The activation function for each neuron in the output layer could be the ReLU function. ReLU causes some neurons to output zero, reducing parameter interdependencies and mitigating overfitting.

[0100] Additionally, setting `learning_rate_init=0.003` during training improves learning efficiency by a factor of 3 compared to the default learning rate of 0.001. Observations showed that learning convergence became faster, and gradients converged normally. Setting the number of iterations to 2000, it was found that the loss rate stabilized after nearly 400 batch iterations during training. After a certain number of training iterations, the loss rate decreased, and ultimately, testing showed that the accuracy of complaint risk assessment prediction was improved.

[0101] It should be noted that the application scenarios described in the above embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0102] Based on the same inventive concept, this application also provides a device for determining the risk of channel complaints. Specifically, in conjunction with... Figure 5 Please provide a detailed explanation.

[0103] Figure 5 This is a schematic diagram of a channel complaint risk determination device provided in an embodiment of this application.

[0104] like Figure 5 As shown, the complaint risk assessment device 500 for this channel may include:

[0105] The first acquisition module 501 is used to acquire target channel information corresponding to the target channel and target user information corresponding to the target user, wherein the target user is the user who conducts business in the target channel;

[0106] The first extraction module 502 is used to extract features from target channel information and target user information to obtain target complaint risk features corresponding to the target channel;

[0107] The input module 503 is used to input the target complaint risk characteristics into the complaint risk assessment model. The complaint risk assessment model assesses the complaint risk of the target channel based on the target complaint risk characteristics and obtains the complaint risk assessment result corresponding to the target channel.

[0108] The following is a detailed description of the complaint risk assessment device 500 for the aforementioned channels:

[0109] In some embodiments, in order to obtain a complaint risk model, the aforementioned channel complaint risk determination device 500 may further include:

[0110] The second acquisition module is used to acquire historical user complaint information;

[0111] The identification module is used to identify target user complaints from historical user complaint information where the channel is responsible.

[0112] The third acquisition module is used to acquire information on the responsible channels corresponding to the complaint information of the target user, as well as information on the complainant corresponding to the complaint information of the target user.

[0113] The second extraction module is used to extract features from the information of responsible channels and complainant information to obtain the complaint risk features corresponding to the responsible channels.

[0114] The determination module is used to determine the target complaint risk level corresponding to the responsible channel;

[0115] The training module is used to train the target deep learning model by using the complaint risk characteristics corresponding to the responsible channels as training samples and the target complaint risk level as the sample label corresponding to the training samples, so as to obtain the complaint risk assessment model.

[0116] In some embodiments, in order to determine target user complaint information from historical user complaint information, the aforementioned identification module may specifically include:

[0117] The input submodule is used to input historical user complaint information into the responsible channel identification model. The responsible channel identification model is used to identify whether the channel corresponding to the historical user complaint information is responsible and obtain the identification result. The responsible channel identification model is constructed by the TRANSFORMER model. The identification result is used to characterize whether the historical user complaint information is information caused by channel violations.

[0118] The determination submodule is used to identify target user complaint information for which the channel is responsible from historical user complaint information based on the identification results.

[0119] In some embodiments, in order to determine the target complaint risk level corresponding to the responsible channel, the aforementioned determination module may specifically include:

[0120] The first acquisition submodule is used to acquire the complaint rates corresponding to multiple historical responsible channels within a preset historical time period;

[0121] The sub-module is used to classify the complaint risk level based on the complaint rate corresponding to multiple historical responsible channels, and obtain the complaint rate threshold corresponding to each of the multiple complaint risk levels.

[0122] The second acquisition submodule is used to acquire the target complaint rate corresponding to the responsible channel, and determine the complaint risk level of the target complaint rate based on the complaint rate thresholds corresponding to multiple complaint risk levels, which is the target complaint risk level corresponding to the responsible channel.

[0123] In some embodiments, for each historically responsible channel among multiple historically responsible channels within a preset historical time period, the aforementioned first acquisition submodule may specifically include:

[0124] The first determining sub-unit is used to determine the proportion of complaints received by the historically responsible channel in each time period of the preset historical time period to the total business volume handled by the channel, so as to obtain the complaint proportions corresponding to the multiple time periods included in the preset historical time period.

[0125] The weighted summation subunit is used to sum the complaint percentages corresponding to each of the multiple time periods according to the weights of each time period, so as to obtain the complaint rate corresponding to the historical responsible channels.

[0126] In some embodiments, the aforementioned multiple complaint risk levels may include low risk, medium risk, and high risk.

[0127] Based on this, the above-mentioned sub-module division can specifically include:

[0128] The second determination subunit is used to statistically analyze the complaint rates corresponding to multiple historical responsible channels and determine the complaint rate distribution corresponding to multiple historical responsible channels.

[0129] The complaint rate threshold determination subunit is used to determine the complaint rate value corresponding to the target quantile in the complaint rate distribution as the complaint rate threshold between medium risk and high risk.

[0130] In some embodiments, in order to obtain a low-risk complaint marker, the aforementioned channel complaint risk determination device 500 may further include:

[0131] The fourth acquisition module is used to acquire information on complaint-free channels that have not been complained about within a preset historical time period, as well as information on complaint-free users who have handled business through complaint-free channels.

[0132] The third extraction module is used to extract features from information on channels without complaints and information on users without complaints, so as to obtain complaint risk features corresponding to channels without complaints.

[0133] The tagging module is used to mark the complaint risk level corresponding to channels without complaints as low risk.

[0134] In some embodiments, in order to train the target deep learning model to obtain the complaint risk assessment model, the training module may specifically include:

[0135] The negative sample correspondence submodule is used to take the complaint risk characteristics corresponding to the responsible channel as negative samples in the training sample, and to take the target complaint risk level as the sample label corresponding to the negative sample.

[0136] The positive sample correspondence submodule is used to take the complaint risk characteristics corresponding to the non-complaint channels as positive samples in the training samples, and to take the low-risk complaint risk level as the sample label corresponding to the positive sample;

[0137] The training submodule is used to train the target deep learning model based on negative samples and their corresponding sample labels, and positive samples and their corresponding sample labels, to obtain the complaint risk assessment model.

[0138] Therefore, by extracting features from the acquired target channel and target user information, target complaint risk characteristics corresponding to the target channel are obtained. Based on these target complaint risk characteristics, a complaint risk assessment model is used to assess the complaint risk of the target channel, yielding a complaint risk assessment result corresponding to the target channel. Thus, in this embodiment of the application, when assessing the complaint risk of an enterprise, the target complaint risk characteristics corresponding to the target channel include not only the target channel information but also the target user information corresponding to the target users conducting business through the target channel. This takes into account the impact of the target users' susceptibility on the complaint risk assessment, improving the accuracy of complaint risk assessment prediction.

[0139] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0140] The electronic device 600 may include a processor 601 and a memory 602 storing computer program instructions.

[0141] Specifically, the processor 601 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0142] Memory 602 may include mass storage for data or instructions. For example, and not limitingly, memory 602 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 602 may include removable or non-removable (or fixed) media. Where appropriate, memory 602 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 602 is non-volatile solid-state memory.

[0143] In certain embodiments, the memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Thus, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this application.

[0144] The processor 601 reads and executes computer program instructions stored in the memory 602 to implement any of the channel complaint risk determination methods in the above embodiments.

[0145] In some examples, electronic device 600 may also include communication interface 603 and bus 610. For example, Figure 6 As shown, the processor 601, memory 602, and communication interface 603 are connected through bus 610 and complete communication with each other.

[0146] The communication interface 603 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0147] Bus 610 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not as a limitation, bus 610 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 610 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0148] For example, the electronic device 600 can be a mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc.

[0149] The electronic device 600 can execute the channel complaint risk determination method in the embodiments of this application, thereby achieving a combination of Figure 1 and Figure 5 The described method and apparatus for determining channel complaint risk.

[0150] Furthermore, in conjunction with the channel complaint risk determination method in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the channel complaint risk determination methods in the above embodiments. Examples of computer-readable storage media include non-transitory computer-readable storage media, such as portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, etc.

[0151] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0152] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0153] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0154] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0155] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for determining channel complaint risk, characterized in that, include: Obtain target channel information corresponding to the target channel and target user information corresponding to the target user, wherein the target user is a user who conducts business on the target channel; Feature extraction is performed on the target channel information and the target user information to obtain the target complaint risk features corresponding to the target channel; The target complaint risk characteristics are input into the complaint risk assessment model, and the complaint risk assessment model is used to assess the complaint risk of the target channel based on the target complaint risk characteristics to obtain the complaint risk assessment result corresponding to the target channel; The method further includes: Obtain historical user complaint information; Identify target user complaints for which the channel is responsible from the historical user complaint information; Obtain the responsible channel information corresponding to the target user's complaint information, and the complaint user information corresponding to the complainant's complaint information; Feature extraction is performed on the responsible channel information and the complainant user information to obtain the complaint risk features corresponding to the responsible channel; Determine the target complaint risk level corresponding to the aforementioned responsible channels; The complaint risk characteristics corresponding to the responsible channels are used as training samples, and the target complaint risk level is used as the sample label corresponding to the training samples. The target deep learning model is trained to obtain the complaint risk assessment model. The determination of the target complaint risk level corresponding to the responsible channel includes: Obtain the complaint rates corresponding to multiple historical responsible channels within a preset historical time period; Based on the complaint rates corresponding to the multiple historical responsible channels, the complaint risk levels are divided, and the complaint rate thresholds corresponding to the multiple complaint risk levels are obtained. Obtain the target complaint rate corresponding to the responsible channel, and determine the complaint risk level of the target complaint rate based on the complaint rate thresholds corresponding to the multiple complaint risk levels, which is the target complaint risk level corresponding to the responsible channel.

2. The method according to claim 1, characterized in that, The step of identifying target user complaint information from the historical user complaint information where the channel is responsible includes: The historical user complaint information is input into the responsible channel identification model. The responsible channel identification model is used to identify whether the channel corresponding to the historical user complaint information is responsible, and the identification result is obtained. The responsible channel identification model is constructed by the TRANSFORMER model. The identification result is used to characterize whether the historical user complaint information is information caused by channel violations. Based on the identification results, target user complaint information for which the channel is responsible is determined from the historical user complaint information.

3. The method according to claim 1, characterized in that, The acquisition of complaint rates corresponding to multiple historical responsible channels within a preset historical time period includes: For each historically responsible channel within a preset historical time period, the following steps are performed to obtain the complaint rate corresponding to each of the multiple historically responsible channels: Determine the proportion of complaints received by the historically responsible channel in each time period of the preset historical time period to the total business volume handled by the channel, and obtain the complaint proportions corresponding to the multiple time periods included in the preset historical time period. Based on the weight corresponding to each of the multiple time periods, the complaint percentages corresponding to each of the multiple time periods are weighted and summed to obtain the complaint rate corresponding to the historical responsible channels.

4. The method according to claim 1, characterized in that, The multiple complaint risk levels include low risk, medium risk, and high risk; The classification of complaint risk levels based on the complaint rates corresponding to the multiple historical responsible channels includes: The complaint rates corresponding to the various historically responsible channels are statistically analyzed to determine the complaint rate distribution corresponding to the various historically responsible channels; The complaint rate value corresponding to the target quantile in the complaint rate distribution is used as the complaint rate threshold between the medium risk and the high risk.

5. The method according to claim 1, characterized in that, The method further includes: Obtain information on complaint-free channels that have not been complained about within a preset historical time period, as well as information on complaint-free users who have conducted business through those complaint-free channels. Feature extraction is performed on the information of the complaint-free channels and the information of the complaint-free users to obtain the complaint risk features corresponding to the complaint-free channels; The complaint risk level corresponding to the aforementioned non-complaint channels is marked as low risk; The step of using the complaint risk characteristics corresponding to the responsible channel as training samples and the target complaint risk level as the sample label corresponding to the training samples to train the target deep learning model to obtain the complaint risk assessment model includes: The complaint risk characteristics corresponding to the responsible channels are used as negative samples in the training samples, and the target complaint risk level is used as the sample label corresponding to the negative samples. The complaint risk characteristics corresponding to the non-complaint channels are used as positive samples in the training samples, and the low-risk complaint risk level is used as the sample label corresponding to the positive samples. The target deep learning model is trained based on the negative samples and their corresponding sample labels, and the positive samples and their corresponding sample labels, to obtain the complaint risk assessment model.

6. The method according to claim 1 or 5, characterized in that, The target deep learning model includes an input layer, two hidden layers, two batch normalized (BN) layers, and an output layer.

7. A device for determining the risk of channel complaints, characterized in that, include: The first acquisition module is used to acquire target channel information corresponding to the target channel and target user information corresponding to the target user, wherein the target user is a user who conducts business on the target channel; The first extraction module is used to extract features from the target channel information and the target user information to obtain target complaint risk features corresponding to the target channel; The input module is used to input the target complaint risk characteristics into the complaint risk assessment model, and to conduct a complaint risk assessment of the target channel based on the target complaint risk characteristics through the complaint risk assessment model to obtain a complaint risk assessment result corresponding to the target channel. The second acquisition module is used to acquire historical user complaint information; The identification module is used to identify target user complaint information for which the channel is responsible from the historical user complaint information; The third acquisition module is used to acquire the responsible channel information corresponding to the target user's complaint information, and the complaint user information corresponding to the complainant's complaint information. The extraction module is used to extract features from the responsible channel information and the complainant user information to obtain the complaint risk features corresponding to the responsible channel; The determination module is used to determine the target complaint risk level corresponding to the responsible channel; specifically, it is used to: obtain the complaint rate corresponding to multiple historical responsible channels within a preset historical time period; classify the complaint risk level according to the complaint rate corresponding to the multiple historical responsible channels, and obtain the complaint rate threshold corresponding to the multiple complaint risk levels; obtain the target complaint rate corresponding to the responsible channel, and determine the complaint risk level of the target complaint rate based on the complaint rate threshold corresponding to the multiple complaint risk levels, which is the target complaint risk level corresponding to the responsible channel; The training module is used to train the target deep learning model by using the complaint risk characteristics corresponding to the responsible channel as training samples and the target complaint risk level as the sample label corresponding to the training samples, so as to obtain the complaint risk assessment model.

8. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the steps of the channel complaint risk determination method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the steps of the channel complaint risk determination method as described in any one of claims 1-6.

10. A computer program product comprising computer instructions, characterized in that, When the instructions in the computer program product are executed by the processor, they implement the steps of the channel complaint risk determination method as described in any one of claims 1-6.

Images