Customer journey feature learning method and device, electronic equipment and readable storage medium

Abstract

The embodiment of the invention provides a customer journey feature learning method and device, electronic equipment and a readable storage medium, and relates to the technical field of artificial intelligence. The method comprises the steps of determining a behavior sequence of customer behaviors based on behavior time of the customer behaviors; performing feature coding on the behavior sequence, and determining behavior sequence features; performing feature coding based on a behavior category corresponding to the customer behavior, and determining a behavior category feature; and performing feature fusion based on the behavior sequence features and the behavior category features to obtain customer journey features. The customer journey features obtained based on learning of the scheme are effective characterization of the customer journey, and a better basis can be provided for customer journey analysis.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology. Specifically, the present application relates to a customer journey feature learning method, device, electronic device and readable storage medium. Background Art

[0002] The customer journey refers to the entire process of customer interaction with a product or service, which reveals the user's complete end-to-end experience with the product or service.

[0003] Customer journey is of great significance in marketing analysis. Therefore, if we can effectively learn customer journey characteristics based on customer journey data, we can effectively characterize the customer journey and provide a basis for customer journey analysis.

[0004] If, after effectively learning the customer journey characteristics, we can effectively predict subsequent customer behavior based on the customer journey characteristics, we can provide better support for marketing.

[0005] At present, the customer journey data used for customer journey analysis generally only includes some structured data, such as customer behavior events, etc., and it is impossible to incorporate unstructured data into customer journey analysis, which affects the accuracy of customer journey analysis. Summary of the invention

[0006] The purpose of this application is to solve at least one of the above technical defects, and to provide a customer journey feature learning method, device, electronic device and readable storage medium. The technical solution adopted by this application is as follows:

[0007] In a first aspect, an embodiment of the present application provides a method for learning customer journey features, the method comprising:

[0008] Determine the behavioral sequence of customer behaviors based on the behavioral timing of customer behaviors;

[0009] Characterize the behavior sequence and determine the behavior sequence characteristics;

[0010] Perform feature coding based on the behavior category corresponding to the customer behavior to determine the behavior category characteristics;

[0011] Feature fusion is performed based on behavior sequence features and behavior category features to obtain customer journey features.

[0012] In a second aspect, an embodiment of the present application provides a customer journey feature learning device, the device comprising:

[0013] A behavior sequence determination module, used to determine the behavior sequence of customer behaviors based on the behavior time of the customer behaviors;

[0014] A behavior sequence feature determination module is used to perform feature encoding on the behavior sequence and determine the behavior sequence feature;

[0015] A behavior category feature determination module is used to perform feature coding based on the behavior category corresponding to the customer behavior and determine the behavior category feature;

[0016] The customer journey feature determination module is used to perform feature fusion based on behavior sequence features and behavior category features to obtain customer journey features.

[0017] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device comprising: a processor and a memory;

[0018] A memory, used for storing operation instructions;

[0019] A processor is used to execute the customer journey feature learning method as shown in any implementation of the first aspect of the present application by calling an operation instruction.

[0020] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the customer journey feature learning method shown in any implementation of the first aspect of the present application.

[0021] The beneficial effects of the technical solution provided by the embodiment of the present application are:

[0022] The solution provided in the embodiment of the present application determines the behavior sequence of customer behaviors based on the behavior time of customer behaviors; performs feature encoding on the behavior sequence to determine the behavior sequence feature; performs feature encoding based on the behavior category corresponding to the customer behavior to determine the behavior category feature; performs feature fusion based on the behavior sequence feature and the behavior category feature to obtain the customer journey feature. The customer journey feature learned based on this solution is an effective representation of the customer journey and can provide a better basis for customer journey analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in describing the embodiments of the present application are briefly introduced below.

[0024] Figure 1 A flowchart of a customer journey feature learning method provided in an embodiment of the present application;

[0025] Figure 2 1 is a flowchart of text processing of a recording text provided in an embodiment of the present application;

[0026] Figure 3 A schematic diagram of the process of collecting customer journey data provided in an embodiment of the present application;

[0027] Figure 4 A flowchart of a specific implementation method of the method provided in the embodiment of the present application;

[0028] Figure 5 An overall flow chart of the method is provided for the embodiment of the present application;

[0029] Figure 6 A schematic diagram of the structure of a customer journey feature learning device provided in an embodiment of the present application;

[0030] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as limiting the present invention.

[0032] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0033] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0034] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0035] Figure 1 A flow chart of a customer journey feature learning method provided in an embodiment of the present application is shown, such as Figure 1As shown, the method may mainly include:

[0036] Step S110: determining the behavior sequence of the customer's behavior based on the behavior time of the customer's behavior;

[0037] Step S120: Characterize the behavior sequence and determine the behavior sequence characteristics;

[0038] Step S130: performing feature coding based on the behavior category corresponding to the customer behavior to determine the behavior category feature;

[0039] Step S140: Perform feature coding based on the behavior category corresponding to the customer behavior to determine the behavior category feature.

[0040] In the embodiment of the present application, customer behavior data can be obtained, and customer journey features can be extracted based on the customer behavior data. Customer behavior data is mostly structured data, which includes behavior time.

[0041] As an example, taking the bank credit card scenario as an example, customer behavior data may include annual fee deduction records, transaction failures, redemption promotions, bill viewing, etc.

[0042] In the embodiment of the present application, customer behaviors may be sorted based on the behavior time of the customer behaviors to obtain a behavior sequence, and the specific form of the behavior sequence may be 1, 2...n.

[0043] Compared with behavior time, behavior sequence removes time-related information and is a clear expression of the order in which each customer's behavior occurs. By encoding the behavior sequence into behavior sequence features, the customer journey features extracted based on the behavior sequence features can also effectively express the order in which customer behaviors occur.

[0044] When predicting customer behavior based on customer journey characteristics, the order of customer behavior is of great significance to the accuracy of the prediction results.

[0045] As an example, a positional encoding method may be used to perform feature encoding on the behavior sequence to obtain an embedding vector of the behavior sequence, namely, a behavior sequence feature.

[0046] In the embodiment of the present application, different categories of customer behaviors may be represented by categories, and each customer behavior category may be mapped into a vector of a fixed length to capture the relationship and similarity between different categories.

[0047] After determining the behavior sequence characteristics and behavior category characteristics, feature fusion can be performed to determine the customer journey characteristics.

[0048] The customer journey features learned in the embodiments of the present application include the behavior sequence and behavior category information of customer behaviors, which is an effective representation of the customer journey and can provide a basis for subsequent customer journey analysis.

[0049] The method provided in the embodiment of the present application determines the behavior sequence of customer behaviors based on the behavior time of customer behaviors; performs feature encoding on the behavior sequence to determine the behavior sequence feature; performs feature encoding based on the behavior category corresponding to the customer behavior to determine the behavior category feature; performs feature fusion based on the behavior sequence feature and the behavior category feature to obtain the customer journey feature. The customer journey feature learned based on this solution is an effective representation of the customer journey and can provide a better basis for customer journey analysis.

[0050] In an optional manner of the embodiment of the present application, feature fusion is performed based on the behavior sequence feature and the behavior category feature to obtain the customer journey feature, including:

[0051] Determine the behavior time characteristics based on the behavior time of customer behavior;

[0052] Customer journey features are obtained by fusion based on behavior sequence features, behavior category features, and behavior time features.

[0053] In the embodiment of the present application, the occurrence time of the customer behavior can also be encoded as a behavior time feature, so as to perform feature fusion based on the behavior sequence feature, the behavior category feature, and the behavior time feature to obtain the customer journey feature.

[0054] As an example, the time when a customer behavior occurs can be in the form of a date, and the behavior time feature can be obtained by encoding the date.

[0055] Compared with behavior sequence features, behavior time features can express the length of time intervals between different customer behaviors. For example, it can indicate that certain customer behaviors occurred in the same period of time. Customer journey features extracted based on behavior time features can also effectively express the behavior time information of customer behaviors.

[0056] In an optional manner of the embodiment of the present application, feature fusion is performed based on behavior sequence features, behavior category features, and behavior time features to obtain customer journey features, including:

[0057] Extract topic text based on customer-related text data;

[0058] Determine the text category to which the subject text belongs, perform feature coding based on the text category, and determine the text category features;

[0059] Determine the text order based on the text time corresponding to the subject text;

[0060] Perform feature encoding based on the text order to determine the text order features;

[0061] Based on text category features and text sequence features, and based on behavior sequence features, behavior category features, and behavior time features, feature fusion is performed to obtain customer journey features.

[0062] In the embodiment of the present application, in addition to using structured customer behavior data, unstructured customer-related text data can also be used to extract customer journey features.

[0063] As an example, customer-related text data may include but is not limited to the text of customer call recordings, the text of customer chat conversations with customer service, the customer's search terms, etc.

[0064] Based on the customer-related text data, the subject text can be extracted, and the subject text is a thematic expression of the customer-related text data.

[0065] By extracting subject text in a specified format from customer-related text data, the subject text can also be converted into structured data, thereby extracting text sequence features and text category features in the same way as customer behavior data.

[0066] In an embodiment of the present application, the subject texts may be sorted based on the text time corresponding to the subject texts to obtain a text order, and the specific form of the text order may be 1, 2...n.

[0067] As an example, the positional encoding method can be used to perform feature encoding on the text sequence to obtain an embedding vector of the text sequence, that is, a text sequence feature.

[0068] In the embodiment of the present application, the subject text can be classified to determine the text category, and the category representation can be performed for different categories of the subject text, and each text category is mapped to a vector of a fixed length.

[0069] After determining the text sequence features and text category features, they can be fused with the behavior sequence features, behavior category features, and behavior time features to obtain customer journey features.

[0070] In an optional manner of the embodiment of the present application, based on the text category feature and the text sequence feature, and based on the behavior sequence feature, the behavior category feature and the behavior time feature, feature fusion is performed to obtain the customer journey feature, including:

[0071] Perform feature encoding based on the text time corresponding to the subject text to determine the text time feature;

[0072] Based on text category features, text sequence features, and text time features, and based on behavior sequence features, behavior category features, and behavior time features, feature fusion is performed to obtain customer journey features.

[0073] In an embodiment of the present application, the text time corresponding to the subject text can also be time-encoded into a text time feature, so as to obtain a customer journey feature based on text category features, text sequence features, and text time features, and based on behavior sequence features, behavior category features, and behavior time features.

[0074] As an example, the text time may be in the form of a date, and the behavior time feature may be obtained by encoding the date.

[0075] In an optional manner of the embodiment of the present application, based on the text category feature, the text sequence feature and the text time feature, and based on the behavior sequence feature, the behavior category feature and the behavior time feature, feature fusion is performed to obtain the customer journey feature, including:

[0076] The text category feature, the text sequence feature and the text time feature are concatenated to obtain the text concatenation feature;

[0077] The behavior sequence feature, the behavior category feature, and the behavior time feature are spliced ​​together to obtain the behavior splicing feature;

[0078] The text concatenation features and behavior concatenation features are input into the Transformer network to obtain the customer journey features output by the Transformer network.

[0079] In an embodiment of the present application, when performing feature fusion, the text category features, text sequence features, and text time features corresponding to each subject text can be first spliced ​​together to obtain a text splicing feature, and the behavior sequence features, behavior category features, and behavior time features corresponding to each customer behavior can be spliced ​​together to obtain a behavior splicing feature.

[0080] The text concatenation features and behavior concatenation features are used as input vectors and input into the Transformer network. The Transformer network is composed of multiple layers of self-attention layers and has strong time series data processing capabilities. It can achieve effective feature fusion and output customer journey features.

[0081] As an example, the vectors starting with [CLS] extracted from the output of the Transformer encoder can be used as customer journey features for subsequent tasks.

[0082] In processing the characteristics of customer unstructured text, this solution proposes a multi-task model to classify and probabilistically predict customer behavior and emotion categories, integrate customer text data with customer behavior characteristics, and build the customer journey through feature aggregation. It effectively supplements the deficiencies and shortcomings of existing technologies in constructing customer journeys based on unstructured data, and realizes effective analysis and application of the entire customer journey.

[0083] In an optional manner of the embodiment of the present application, the subject text includes a text tag, and the text tag includes at least one of the following:

[0084] Customer sentiment labels;

[0085] Business category label;

[0086] Customer intent labeling.

[0087] In an embodiment of the present application, the subject text may include text tags, and the text tags may include customer emotion tags, business category tags, and customer intention tags, that is, emotion recognition is performed on customer-related text data to obtain customer emotion tags, intention recognition is performed on customer-related text data to obtain customer emotion tags, and business classification is performed on customer-related text data to obtain business category tags.

[0088] In an optional manner of the embodiment of the present application, extracting the subject text based on the text data related to the customer includes:

[0089] In response to the character quantity of the text data related to the customer being greater than a preset value, dividing the text data into paragraphs to obtain a plurality of paragraph texts;

[0090] Extracting paragraph text features of each paragraph text;

[0091] Determine text labels based on paragraph text features.

[0092] In the embodiment of the present application, when the number of characters in the text data related to the customer is greater than a preset value, it indicates that the amount of text is large and text tags can be extracted from it.

[0093] Specifically, the text data may be divided into paragraphs to obtain a plurality of paragraph texts, and text tags may be extracted for each paragraph.

[0094] In an optional manner of the embodiment of the present application, extracting paragraph text features of each paragraph text includes:

[0095] The paragraph text is input into the Bidirectional Encoder Representations from Transformers (BERT) model to obtain the paragraph text features output by the BERT model.

[0096] As an example, text label extraction can be performed on customer audio recordings.

[0097] The recorded text is preprocessed by segmentation and divided into multiple text paragraphs. Each paragraph represents a part of the entire call content. The model can understand the text content and efficiently segment the recorded text. Compared with single sentence segmentation as the input text of the model, multi-sentence merged paragraphs are more conducive to the model's semantic understanding of the scene based on the context content.

[0098] Each text paragraph is processed through a word embedding layer to convert the words in the text into word vectors. The model obtains the semantics of the word vectors through training. Position encoding is used to represent the position of each word in the text paragraph to consider the order information of the words, which helps the model distinguish words in different positions.

[0099] BERT is a powerful natural language processing model that can capture vocabulary and context information. In this example, we can encode based on the BERT model.

[0100] The word embeddings and position encodings of each text paragraph are input into the BERT model to obtain a more advanced text representation.

[0101] In the BERT output of each text paragraph, the [CLS] tag at the beginning is usually used as the encoding header of the text paragraph. The [CLS] symbol is used as a classification flag. After the self-attention mechanism, it integrates the sentence and classification-related information. It is used as the semantic representation of the input text in the subsequent task classification. Compared with other words in the text, [CLS], as a symbol without obvious semantic information, can more "fairly" integrate the semantic information of each word in the text.

[0102] In this example, a multi-task prediction model can be connected. By performing different tasks on each text paragraph, the model can handle multiple tasks at the same time, such as sentiment analysis, entity recognition, text classification, etc.

[0103] Specifically, the sentiment analysis model can be used to label each text paragraph as positive, negative, or neutral, and the corresponding probability value can be calculated for each label category. The text classification model can be used to classify each text paragraph into categories related to credit card business, such as customer card cancellation, consultation bill, etc., and the corresponding probability value can be calculated for each label category. The text classification model can be used to classify each text paragraph into categories related to customer intent, such as complaints, complaints, reasons for card cancellation, etc., and the corresponding probability value can be calculated for each label category.

[0104] As an example, Figure 2It is a flowchart of text processing of recorded text provided in an embodiment of the present application.

[0105] like Figure 2 As shown in , the audio recording text is first segmented into paragraphs to obtain multiple audio recording text paragraphs.

[0106] After word embedding and position encoding, multiple audio recording text paragraphs are input into the BERT encoder, and then the output vector of the BERT encoder is input into the multi-task prediction model. The prediction model of each task can include a fully connected layer, which is activated by the Sigmoid function and finally outputs text labels of different categories.

[0107] In an optional manner of the embodiment of the present application, the subject text also includes a text summary.

[0108] In the embodiment of the present application, the subject text may also include a text summary. The text summary may be understood as a global text tag. Combining the global text tag with the paragraph text tag can better extract the information contained in the text data.

[0109] As an example, Figure 3 It is a schematic diagram of the process of collecting customer journey data provided by an embodiment of the present application.

[0110] like Figure 3 As shown in , customer touchpoint data, that is, customer behavior data, in the bank credit card scenario, customer behavior data may include annual fee deduction records, transaction failures, redemption promotions, bill viewing, etc.

[0111] Customer behavior data is mostly structured data, and in this case, customer unstructured data can also be obtained. Specifically, the obtained customer unstructured data can include customer recorded call text data and online customer service text data.

[0112] The recorded text data is segmented into paragraphs, and each call recording is segmented into multiple text paragraphs for subsequent label classification through the text multi-task model. The recorded text is summarized by the generative model to extract key information. Then, the multi-task learning text model is used for label classification. Specific labels include but are not limited to emotion labels, business category labels, and customer intent labels:

[0113] The above text labels and customer behavior data are used together as customer journey data to extract customer journey features.

[0114] In an optional manner of the embodiment of the present application, after obtaining the customer journey feature, the method further includes:

[0115] Predict customer behavior based on customer journey characteristics.

[0116] In the embodiment of the present application, customer behavior prediction may specifically include three tasks: customer experience insight, customer churn warning, and customer marketing recommendation.

[0117] Among them, customer experience insights are used to predict the next possible behavior of customers. This task can be a classification task, for example, predicting whether a customer will buy or churn.

[0118] Customer churn warning is to predict the churn of customers. This task can be a binary classification task, for example, predicting whether a customer will churn.

[0119] Customer marketing recommendation is to predict personalized marketing recommendations for customers. This task is usually a multi-class classification task, for example, predicting the customer's interest level or possible response to different types of marketing activities.

[0120] In the embodiment of the present application, each prediction task includes an independent neural network structure, which is usually composed of fully connected layers. These fully connected layers take the vector of the output coding header ([CLS]) as input and generate corresponding prediction results through the Sigmoid function.

[0121] In an embodiment of the present application, for each prediction task, the model outputs the predicted category label and the corresponding probability score. These probability scores represent the probabilities of different categories of customer behavior prediction, churn warning, and marketing recommendation tasks. The key idea of ​​the algorithm structure is to use the comprehensive features of customer behavior, location, and date information, pass through the Transformer encoder, and then perform different prediction tasks through a task-specific fully connected layer network. Such a multi-task model can simultaneously handle multiple customer behavior-related prediction tasks in a comprehensive framework, thereby improving the efficiency and performance of the model.

[0122] As an example, Figure 4 A flowchart of a specific implementation method of the method provided in the embodiment of the present application is provided.

[0123] like Figure 4 As shown in the figure, date embedding encodes the date information of customer behavior into a vector form so that the model can capture time-related features. Position embedding, that is, behavior sequence representation, encodes the order of customer behavior into a vector form. Customer behavior category representation (Label Embedding) maps customer behavior categories into a fixed-length vector.

[0124] Add the behavior category features, sequence features, and time features element by element to form an input feature vector. This vector will contain comprehensive features of customer behavior, behavior sequence, and date information. Input this input feature vector to the Transformer network. The output vector of the Transformer network is used as the customer journey feature.

[0125] The customer journey features are input into the multi-task prediction model. Each prediction model contains an independent neural network structure, usually composed of fully connected layers. These fully connected layers take the vector of the output coding head [CLS] as input and generate the corresponding prediction results through the Sigmoid function. Specific prediction tasks can include customer experience insights, customer churn warning, and customer marketing recommendations.

[0126] As an example, Figure 5 The present invention provides an overall flow chart of the method according to the embodiment of the present application.

[0127] like Figure 5 As shown in , data integration and acquisition are first performed. This method involves collecting and integrating data from various data sources. These data can be structured (such as database records, tabular data) or unstructured (such as text, images, audio, video, etc.). These data may come from different channels, such as social media, websites, application logs, customer feedback, etc.

[0128] Customer journey construction: Based on the integrated data, the system constructs the customer's journey during their interaction process. This includes various customer interactions, such as visiting the website, purchasing products, communicating with customer service, etc. By analyzing the data, key events and nodes of the customer journey can be identified.

[0129] Customer Journey Feature Extraction: Extract relevant features from structured and unstructured data for each event and node in the customer journey. These features can include customer behavior patterns, sentiment analysis, product preferences, geographic location, etc. This helps to quantify various aspects of the customer journey for analysis.

[0130] Data modeling: Use the extracted customer journey features to build prediction models or classification models to achieve different applications. This solution can be applied to customer journey applications and strategy delivery scenarios in various fields, including but not limited to: customer experience insights, customer behavior prediction, customer churn warning, and customer marketing recommendations.

[0131] After applying the above solution, you can monitor the application effect of the customer journey, collect and analyze data backflow, implement closed-loop management of customer data, and continuously optimize models and processes to improve understanding of customer needs and feedback and better meet customer expectations.

[0132] Through the above process, detailed customer journey characteristics are constructed to better understand customer needs and emotions, improve customer satisfaction, optimize business processes, and develop more targeted marketing and customer service strategies.

[0133] Based on Figure 1 The same principle as shown in the method, Figure 6 A schematic diagram of the structure of a customer journey feature learning device provided in an embodiment of the present application is shown. Figure 6 As shown, the customer journey feature learning device 60 may include:

[0134] A behavior sequence determination module 610, for determining the behavior sequence of the customer's behavior based on the behavior time of the customer's behavior;

[0135] The behavior sequence feature determination module 620 is used to perform feature encoding on the behavior sequence and determine the behavior sequence feature;

[0136] The behavior category feature determination module 630 is used to perform feature coding based on the behavior category corresponding to the customer behavior and determine the behavior category feature;

[0137] The customer journey feature determination module 640 is used to perform feature fusion based on the behavior sequence feature and the behavior category feature to obtain the customer journey feature.

[0138] The device provided in the embodiment of the present application determines the behavior sequence of customer behaviors based on the behavior time of customer behaviors; performs feature encoding on the behavior sequence to determine the behavior sequence feature; performs feature encoding based on the behavior category corresponding to the customer behavior to determine the behavior category feature; performs feature fusion based on the behavior sequence feature and the behavior category feature to obtain the customer journey feature. The customer journey feature obtained by learning based on this solution is an effective representation of the customer journey and can provide a better basis for customer journey analysis.

[0139] Optionally, the customer journey feature determination module is specifically used to:

[0140] Determine the behavior time characteristics based on the behavior time of customer behavior;

[0141] Customer journey features are obtained by fusion based on behavior sequence features, behavior category features, and behavior time features.

[0142] Optionally, when the customer journey feature determination module performs feature fusion based on the behavior sequence feature, the behavior category feature, and the behavior time feature to obtain the customer journey feature, it is specifically used to:

[0143] Extract topic text based on customer-related text data;

[0144] Determine the text category to which the subject text belongs, perform feature coding based on the text category, and determine the text category features;

[0145] Determine the text order based on the text time corresponding to the subject text;

[0146] Perform feature encoding based on the text order to determine the text order features;

[0147] Based on text category features and text sequence features, and based on behavior sequence features, behavior category features, and behavior time features, feature fusion is performed to obtain customer journey features.

[0148] Optionally, when the customer journey feature determination module performs feature fusion based on the text category feature and the text sequence feature, and based on the behavior sequence feature, the behavior category feature and the behavior time feature to obtain the customer journey feature, it is specifically used to:

[0149] Perform feature encoding based on the text time corresponding to the subject text to determine the text time feature;

[0150] Based on text category features, text sequence features, and text time features, and based on behavior sequence features, behavior category features, and behavior time features, feature fusion is performed to obtain customer journey features.

[0151] Optionally, when the customer journey feature determination module performs feature fusion based on the text category feature, the text sequence feature, and the text time feature, and based on the behavior sequence feature, the behavior category feature, and the behavior time feature to obtain the customer journey feature, it is specifically used to:

[0152] The text category feature, the text sequence feature and the text time feature are concatenated to obtain the text concatenation feature;

[0153] The behavior sequence feature, the behavior category feature, and the behavior time feature are spliced ​​together to obtain the behavior splicing feature;

[0154] The text concatenation features and behavior concatenation features are input into the Transformer network to obtain the customer journey features output by the Transformer network.

[0155] Optionally, the subject text includes a text tag, and the text tag includes at least one of the following:

[0156] Customer sentiment labels;

[0157] Business category label;

[0158] Customer intent labeling.

[0159] Optionally, when extracting the subject text based on the text data related to the customer, the customer journey feature determination module is specifically used to:

[0160] In response to the character quantity of the text data related to the customer being greater than a preset value, dividing the text data into paragraphs to obtain a plurality of paragraph texts;

[0161] Extracting paragraph text features of each paragraph text;

[0162] Determine text labels based on paragraph text features.

[0163] Optionally, when extracting the paragraph text features of each paragraph text, the customer journey feature determination module is specifically used to:

[0164] Input the paragraph text into the BERT model and obtain the paragraph text features output by the BERT model.

[0165] Optionally, the subject text also includes a text summary.

[0166] Optionally, the above device further includes:

[0167] The customer behavior prediction module is used to predict customer behavior based on the customer journey characteristics after obtaining the customer journey characteristics.

[0168] It can be understood that the above modules of the customer journey feature learning device in this embodiment have the function of realizing Figure 1 The functions of the corresponding steps of the customer journey feature learning method in the embodiment shown in . The function can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions. The above modules can be software and / or hardware, and the above modules can be implemented separately or integrated with multiple modules. For the functional description of each module of the above customer journey feature learning device, please refer to Figure 1 The corresponding description of the customer journey feature learning method in the embodiment shown in will not be repeated here.

[0169] An embodiment of the present application provides an electronic device, including a processor and a memory;

[0170] A memory, used for storing operation instructions;

[0171] A processor is used to execute the customer journey feature learning method provided in any embodiment of the present application by calling an operation instruction.

[0172] As an example, Figure 7 A schematic diagram of the structure of an electronic device applicable to the embodiment of the present application is shown. Figure 7As shown, the electronic device 2000 includes: a processor 2001 and a memory 2003. The processor 2001 and the memory 2003 are connected, such as through a bus 2002. Optionally, the electronic device 2000 may also include a transceiver 2004. It should be noted that in actual applications, the transceiver 2004 is not limited to one, and the structure of the electronic device 2000 does not constitute a limitation on the embodiments of the present application.

[0173] The processor 2001 is applied in the embodiment of the present application to implement the method shown in the above method embodiment. The transceiver 2004 may include a receiver and a transmitter. The transceiver 2004 is applied in the embodiment of the present application to implement the function of the electronic device of the embodiment of the present application communicating with other devices when executed.

[0174] Processor 2001 may be a CPU (Central Processing Unit), a general purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 2001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0175] The bus 2002 may include a path to transmit information between the above components. The bus 2002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 2002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0176] The memory 2003 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0177] Optionally, the memory 2003 is used to store application code for executing the solution of the present application, and the execution is controlled by the processor 2001. The processor 2001 is used to execute the application code stored in the memory 2003 to implement the customer journey feature learning method provided in any embodiment of the present application.

[0178] The electronic device provided in the embodiment of the present application is applicable to any embodiment of the above method and will not be described in detail here.

[0179] The embodiment of the present application provides an electronic device, which, compared with the prior art, determines the behavior sequence of customer behaviors based on the behavior time of customer behaviors; performs feature encoding on the behavior sequence to determine the behavior sequence feature; performs feature encoding based on the behavior category corresponding to the customer behavior to determine the behavior category feature; performs feature fusion based on the behavior sequence feature and the behavior category feature to obtain the customer journey feature. The customer journey feature learned based on this solution is an effective representation of the customer journey and can provide a better basis for customer journey analysis.

[0180] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the customer journey feature learning method shown in the above method embodiment is implemented.

[0181] The computer-readable storage medium provided in the embodiments of the present application is applicable to any embodiment of the above method and will not be described in detail here.

[0182] The embodiment of the present application provides a computer-readable storage medium. Compared with the prior art, the behavior sequence of customer behaviors is determined based on the behavior time of customer behaviors; the behavior sequence is feature-encoded to determine the behavior sequence feature; the behavior category feature is feature-encoded based on the behavior category corresponding to the customer behavior; and the customer journey feature is obtained by feature fusion based on the behavior sequence feature and the behavior category feature. The customer journey feature obtained by learning based on this solution is an effective representation of the customer journey and can provide a better basis for customer journey analysis.

[0183] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0184] The above descriptions are only some embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A customer journey feature learning method, characterized in that: include: Determining a behavior sequence of the customer's behavior based on the behavior time of the customer's behavior; Characterizing the behavior sequence and determining the behavior sequence characteristics; Perform feature coding based on the behavior category corresponding to the customer behavior to determine the behavior category feature; Feature fusion is performed based on the behavior sequence feature and the behavior category feature to obtain a customer journey feature.

2. The method according to claim 1, characterized in that The feature fusion based on the behavior sequence feature and the behavior category feature to obtain the customer journey feature includes: Determining a behavior time characteristic based on the behavior time of the customer behavior; Feature fusion is performed based on the behavior sequence feature, the behavior category feature, and the behavior time feature to obtain a customer journey feature.

3. The method according to claim 2, characterized in that The feature fusion based on the behavior sequence feature, the behavior category feature and the behavior time feature to obtain the customer journey feature includes: Extract topic text based on customer-related text data; Determine the text category to which the subject text belongs, and perform feature coding based on the text category to determine text category features; Determining the text order based on the text time corresponding to the subject text; Perform feature encoding based on the text sequence to determine text sequence features; Based on the text category feature and the text sequence feature, and based on the behavior sequence feature, the behavior category feature and the behavior time feature, feature fusion is performed to obtain a customer journey feature.

4. The method according to claim 3, characterized in that The feature fusion based on the text category feature and the text sequence feature, and based on the behavior sequence feature, the behavior category feature and the behavior time feature to obtain the customer journey feature includes: Perform feature coding based on the text time corresponding to the subject text to determine the text time feature; Based on the text category feature, the text sequence feature and the text time feature, and based on the behavior sequence feature, the behavior category feature and the behavior time feature, feature fusion is performed to obtain a customer journey feature.

5. The method according to claim 4, characterized in that The obtaining of customer journey features based on the text category feature, the text sequence feature, and the text time feature, and based on the behavior sequence feature, the behavior category feature, and the behavior time feature, comprises: Splicing the text category feature, the text sequence feature and the text time feature to obtain a text splicing feature; splicing the behavior sequence feature, the behavior category feature, and the behavior time feature to obtain a behavior splicing feature; The text concatenation features and the behavior concatenation features are input into a Transformer network to obtain customer journey features output by the Transformer network.

6. The method according to claim 3, characterized in that The subject text includes a text tag, and the text tag includes at least one of the following: Customer sentiment labels; Business category label; Customer intent labeling.

7. The method according to claim 6, characterized in that The extracting of subject text based on the text data related to the customer includes: In response to the character quantity of the text data related to the customer being greater than a preset value, dividing the text data into paragraphs to obtain a plurality of paragraph texts; Extracting paragraph text features of each paragraph text; The text tag is determined based on the paragraph text feature.

8. The method according to claim 7, characterized in that The extracting of paragraph text features of each paragraph text comprises: The paragraph text is input into the transformer-based bidirectional encoding representation BERT model to obtain the paragraph text features output by the BERT model.

9. The method according to claim 3, characterized in that: The subject text also includes a text abstract.

10. The method according to any one of claims 1 to 9, characterized in that After obtaining the customer journey feature, the method further includes: Customer behavior prediction is performed based on the customer journey characteristics.

11. A customer journey feature learning device, characterized in that: include: A behavior sequence determination module, used to determine the behavior sequence of the customer's behavior based on the behavior time of the customer's behavior; A behavior sequence feature determination module, used to perform feature encoding on the behavior sequence and determine the behavior sequence feature; A behavior category feature determination module, used to perform feature coding based on the behavior category corresponding to the customer behavior and determine the behavior category feature; The customer journey feature determination module is used to perform feature fusion based on the behavior sequence feature and the behavior category feature to obtain the customer journey feature.

12. The device according to claim 11, characterized in that The customer journey feature determination module is specifically used for: Determining a behavior time characteristic based on the behavior time of the customer behavior; Feature fusion is performed based on the behavior sequence feature, the behavior category feature, and the behavior time feature to obtain a customer journey feature.

13. The device according to claim 12, characterized in that When the customer journey feature determination module performs feature fusion based on the behavior sequence feature, the behavior category feature, and the behavior time feature to obtain the customer journey feature, the customer journey feature determination module is specifically used to: Extract topic text based on customer-related text data; Determine the text category to which the subject text belongs, and perform feature coding based on the text category to determine text category features; Determining the text order based on the text time corresponding to the subject text; Perform feature encoding based on the text sequence to determine text sequence features; Based on the text category feature and the text sequence feature, and based on the behavior sequence feature, the behavior category feature and the behavior time feature, feature fusion is performed to obtain a customer journey feature.

14. The device according to claim 13, characterized in that When the customer journey feature determination module performs feature fusion based on the text category feature and the text sequence feature, and based on the behavior sequence feature, the behavior category feature, and the behavior time feature to obtain the customer journey feature, it is specifically used to: Perform feature coding based on the text time corresponding to the subject text to determine the text time feature; Based on the text category feature, the text sequence feature and the text time feature, and based on the behavior sequence feature, the behavior category feature and the behavior time feature, feature fusion is performed to obtain a customer journey feature.

15. The device according to claim 14, characterized in that When the customer journey feature determination module performs feature fusion based on the text category feature, the text sequence feature, and the text time feature, and based on the behavior sequence feature, the behavior category feature, and the behavior time feature to obtain the customer journey feature, it is specifically used to: Splicing the text category feature, the text sequence feature and the text time feature to obtain a text splicing feature; splicing the behavior sequence feature, the behavior category feature, and the behavior time feature to obtain a behavior splicing feature; The text concatenation features and the behavior concatenation features are input into a Transformer network to obtain customer journey features output by the Transformer network.

16. The device according to claim 13, characterized in that The subject text includes a text tag, and the text tag includes at least one of the following: Customer sentiment labels; Business category label; Customer intent labeling.

17. The device according to claim 16, characterized in that The customer journey feature determination module is specifically used to extract the subject text based on the customer-related text data: In response to the character quantity of the text data related to the customer being greater than a preset value, dividing the text data into paragraphs to obtain a plurality of paragraph texts; Extracting paragraph text features of each paragraph text; The text tag is determined based on the paragraph text feature.

18. The device according to claim 17, characterized in that When extracting the paragraph text features of each paragraph text, the customer journey feature determination module is specifically used to: The paragraph text is input into the BERT model to obtain the paragraph text features output by the BERT model.

19. The device according to claim 13, characterized in that The subject text also includes a text abstract.

20. The device according to any one of claims 11 to 19, characterized in that Also includes: A customer behavior prediction module is used to predict customer behavior based on the customer journey characteristics after obtaining the customer journey characteristics.

21. An electronic device, characterized in that: including a processor and a memory; The memory is used to store operation instructions; The processor is used to execute the method according to any one of claims 1 to 10 by calling the operation instruction.

22. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

Images