Submersible guest mining method and related equipment

Abstract

The embodiment of the invention provides a latent guest mining method and related equipment, and belongs to the technical field of latent guest identification. The method comprises the steps of obtaining platform behavior data of a first object; the platform behavior data is obtained by performing identity unification on original behavior data which are obtained from a plurality of different platforms and are associated with the first object; performing feature extraction according to the platform behavior data to obtain a car purchase behavior feature and a social behavior feature of the first object; the vehicle purchase behavior characteristics are related to individual vehicle purchase behaviors; the social behavior characteristics are related to group vehicle purchasing behaviors; determining an individual purchase intention of the first object based on the car purchase behavior characteristics; determining a group purchase impact of the first object based on the social behavior features; analyzing based on the individual purchase intention and the group purchase influence, and determining the customer state of the object; according to the embodiment of the invention, the automobile sales conversion rate can be effectively improved through the technical means of unification of multi-platform data identities and individual group double-aspect analysis.

Description

Technical Field

[0001] This application relates to the field of potential customer identification technology, and in particular to a method and related equipment for potential customer discovery. Background Technology

[0002] With the deepening of digital transformation in the automotive industry, car sales models have undergone tremendous changes. Traditional car sales mainly relied on offline showrooms and sales personnel, while modern car sales have expanded to multiple online channels such as live streaming sales, short video marketing, and social media interaction.

[0003] In the automotive sales sector, potential customer identification and mining are crucial for improving sales conversion rates. On one hand, existing technologies suffer from data silos, failing to effectively integrate behavioral data from different channels such as live streaming, short videos, and private messages. The data from each platform is independent, failing to form a complete customer behavior trajectory, thus hindering potential customer identification and mining. Furthermore, existing technologies cannot accurately identify whether customers are high-value potential customers—those who can effectively drive sales orders—resulting in low automotive sales conversion rates.

[0004] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the Invention

[0005] The main objective of this application is to propose a potential customer mining method and related equipment. By acquiring behavioral data associated with target customers from multiple different platforms and unifying their identities, and analyzing the individual purchasing intentions and group purchasing influence of target customers, it is possible to accurately assess whether a customer is a high-value potential customer and improve the car sales conversion rate.

[0006] To achieve the above objectives, one aspect of this application proposes a potential customer discovery method, the method comprising: Obtain platform behavior data of the first object; the platform behavior data is obtained by unifying the identity of the original behavior data associated with the first object obtained from several different platforms. Based on the platform behavior data, feature extraction is performed to obtain the car purchase behavior features and social behavior features of the first object; the car purchase behavior features are related to individual vehicle purchase behavior; the social behavior features are related to group vehicle purchase behavior. Determine the individual purchase intention of the first object based on the aforementioned car purchase behavior characteristics; The group purchasing influence of the first target is determined based on the aforementioned social behavior characteristics; Based on the analysis of individual purchasing intentions and the influence of group purchasing, the customer status of the object is determined; the customer status indicates whether the first object is a potential customer who meets the requirements.

[0007] In some embodiments, the platform data of the first object is obtained in the following ways: Raw behavioral data associated with the first object are obtained from at least two different platforms, each of which has its own independent user identification system; For each platform, based on the original behavioral data obtained from the platform, a set of identity features and a set of behavioral sequence features are extracted, and a user fingerprint vector representing the first object within the platform is generated based on the set of identity features and behavioral sequence features. The similarity between user fingerprint vectors from different platforms is calculated separately. If the similarity calculation result is higher than a preset threshold, it is determined that the user fingerprint vectors participating in the similarity calculation correspond to the same object. Generate a global identity identifier for each user fingerprint vector that is identified as the same object; Based on the global identity identifier, the original behavioral data from different platforms are aggregated in chronological order of the occurrence of the behaviors to obtain the platform behavioral data of the first object.

[0008] In some embodiments, determining the individual purchase intention of the first object based on the car purchase behavior characteristics includes: An initial car purchase intention score is calculated based on a preset multi-factor scoring model and the car purchase behavior characteristics. The initial car purchase intention score is then corrected based on preset special scoring rules. The car purchase intention score of the first object is obtained based on the correction result. The correction logic in the preset special scoring rules is based on one or more of the following: the market popularity of the car model of interest, the sensitivity to price, and the degree of coincidence between the time of car purchase-related behavior and the promotion cycle. The purchase intention status of the first object is determined by matching the car purchase behavior characteristics with the preset intention stage identification rules; the purchase intention status includes the initial interest stage, the in-depth understanding stage, and the purchase intention stage. Based on the car purchase intention score and car purchase intention status, the individual purchase intention of the first object is determined.

[0009] In some embodiments, the scoring factors of the preset multi-factor scoring model include basic behavioral factors, content depth factors, purchase intention factors, and time-sensitive factors; the individual vehicle purchase behavior includes browsing, interaction, and consultation; the vehicle purchase behavior characteristics indicate the content, frequency, and duration of individual vehicle purchase behavior; the calculation of the initial vehicle purchase intention score based on the preset multi-factor scoring model and the vehicle purchase behavior characteristics includes: The basic behavioral factor score is calculated based on the frequency and duration of individual vehicle purchases. A content depth factor score is calculated based on the depth of the content browsed and the quality of the interaction. Determine whether there are keywords related to the car purchase process in the consultation content, and calculate the purchase intention factor score based on the judgment result and the depth of the consultation content. The time sensitivity factor score is calculated based on the degree of overlap between the last active time in the distribution of vehicle purchase behavior and the car promotion period. The initial car purchase intention score is obtained by weighting the scores of the basic behavioral factors, the content depth factors, the purchase intention factors, and the time sensitivity factors.

[0010] In some embodiments, the platform data includes the first object's social relationships on various platforms; the social behavior characteristics of the first object are obtained through the following methods: Based on the social relationships of the first object on various platforms, identify several second objects that have explicit social relationships with the first object; Based on the analysis of platform behavior data of each object on each platform, several third objects with implicit relationships to the first object are identified; the similarity between the third objects and the first object on each platform in terms of browsing records, interaction records and consultation records meets the preset requirements. Feature extraction is performed based on the interaction records between the first object and each of the second objects and between each of the third objects to obtain the interaction depth feature and interaction frequency feature of the first object; the interaction depth feature indicates the text, sound and video content related to the interaction between objects; the interaction frequency feature indicates the frequency and time distribution of the interaction between objects. A time series is constructed based on the interaction depth features and interaction frequency features, and the social behavior features of the first object are obtained based on the construction results.

[0011] In some embodiments, determining the group purchasing influence of the first object based on the social behavior characteristics includes: Based on the social behavior characteristics, a number of interactive objects that have social relationships with the first object are determined, as well as the interaction between the first object and each of the interactive objects; Using the first object as the main point and several interactive objects as several secondary points, connection edges are constructed between the main point and each of the secondary points according to the interaction between the first object and each of the interactive objects, so as to construct the social relationship graph of the first object; the type of the connection edge is determined by the type of interaction channel between the first object and the interactive objects. The social relationship graph of the first object is input into a preset graph neural network model. Based on the information transmission mechanism of the graph neural network model, the interactive information features transmitted between the main point and several secondary points are aggregated. Based on the aggregation result, the group purchase influence of the first object is output.

[0012] In some embodiments, the method further includes: The system acquires the historical car purchase behavior characteristics, real-time car purchase intention status, and real-time external market condition characteristics of the first object; the historical car purchase behavior characteristics are obtained by feature extraction based on the historical platform behavior data of the first object within a certain period of time. The historical car purchase behavior characteristics, the real-time car purchase intention status, and the real-time external market condition characteristics are input into a preset state prediction model for calculation to obtain the predicted car purchase intention status of the first object. Based on the analysis of the first object's real-time car purchase intention status and the predicted car purchase intention status, a target marketing strategy is determined and delivered to the first object.

[0013] To achieve the above objectives, another aspect of this application proposes a potential customer discovery system, the system comprising: The data acquisition module is used to acquire platform behavior data of the first object; the platform behavior data is obtained by unifying the identity of the original behavior data associated with the first object obtained from several different platforms. The feature extraction module is used to extract features based on the platform behavior data to obtain the car purchase behavior features and social behavior features of the first object; the car purchase behavior features are related to individual vehicle purchase behavior; the social behavior features are related to group vehicle purchase behavior. The analysis module is used to determine the individual purchase intention of the first object based on the car purchase behavior characteristics; determine the group purchase influence of the first object based on the social behavior characteristics; and analyze the individual purchase intention and the group purchase influence to determine the customer status of the object; the customer status indicates whether the first object is a potential customer who meets the requirements.

[0014] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0015] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0016] The embodiments of this application include at least the following beneficial effects: This application provides a method, apparatus, electronic device, storage medium, and program product for identifying potential customers. This solution obtains the target customer's platform behavior data by acquiring initial behavioral data associated with the target customer from multiple different platforms and unifying their identities, thereby achieving data fusion of independent platform data and effectively solving the data silo problem. Furthermore, the method of this application extracts features based on the target customer's platform behavior data to obtain the target customer's car purchase behavior characteristics and social behavior characteristics, and further analyzes the target customer's individual purchase intention and group purchase influence to determine whether the target customer is a potential high-value customer. By analyzing the target customer from both individual and group perspectives, sales orders can be effectively facilitated. Compared with traditional methods that only consider the target customer themselves, potential customers can be identified more accurately, thereby improving the car sales conversion rate. Attached Figure Description

[0017] Figure 1 This is a flowchart of a potential customer discovery method provided in an embodiment of this application; Figure 2 This is a structural block diagram of a potential customer discovery system provided in an embodiment of this application; Figure 3 This is a structural block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0020] Before providing a detailed description of the embodiments of this application, some of the nouns and terms involved in the embodiments of this application will be explained first. The nouns and terms involved in the embodiments of this application are subject to the following interpretations.

[0021] 1) LSTM (Long Short-Term Memory Network): This is a special type of recurrent neural network (RNN) specifically designed to address the vanishing or exploding gradient problems that traditional RNNs often encounter during long sequence training. By introducing "memory units" and sophisticated "gating mechanisms" (including input gates, forget gates, and output gates), it can selectively remember, forget, and output information, thereby effectively capturing long-distance temporal dependencies.

[0022] 2) GNN (Graph Neural Network): A deep learning model specifically designed for processing graph-structured data. Its core idea is to use a "message passing" mechanism to allow nodes (entities) in the graph to exchange and aggregate information with their neighbors, thereby learning a powerful representation that includes graph topological relationships and node features.

[0023] 3) Similarity algorithms: These are methods used to quantify the degree of similarity or relevance between two objects (such as text, images, and vectors). Common algorithms include: literal-based (such as cosine similarity and Jaccard coefficient), semantic-based (such as those calculated through word embedding models), and structure-based (such as in graph data).

[0024] 4) CRM (Customer Relationship Management) System: This is a digital platform used by enterprises to systematically manage, analyze, and optimize interactions with customers throughout their entire lifecycle. Its core is to integrate sales, marketing, and service processes, consolidating customer data into a unified view to help enterprises improve customer satisfaction, increase sales efficiency, and drive continuous growth. It is a core tool for modern enterprises to conduct data-driven, refined customer operations.

[0025] With the deepening of digital transformation in the automotive industry, car sales models have undergone tremendous changes. Traditional car sales mainly relied on offline showrooms and sales personnel, while modern car sales have expanded to multiple online channels such as live streaming sales, short video marketing, and social media interaction.

[0026] In the automotive sales scenario, the behavioral patterns of potential customers exhibit characteristics of multi-platform and multi-scenario engagement: Live streaming platform: Users watch product introductions in the live streaming rooms of car brands and send bullet comments to inquire about prices, configurations, and other information. Short video platform: Users browse car-related short videos and express their purchase intentions or ask questions in the comment section. Social media: Users share car-related content and discuss car-buying plans with friends on platforms such as WeChat and Weibo. Private message consultation: Users can conduct in-depth consultations through private messages or the customer service system to learn about specific car models, preferential policies, etc.

[0027] This cross-platform behavioral trajectory contains a wealth of car purchase intention information, but traditional car sales CRM systems cannot effectively integrate this multi-source data, resulting in inaccurate identification of potential customers and a lack of targeted marketing strategies.

[0028] In view of this, this application provides a method, apparatus, electronic device, storage medium, and program product for identifying potential customers. This solution obtains initial behavioral data associated with target customers from multiple different platforms and unifies their identities to obtain platform behavioral data of the target customers, achieving data fusion of independent platform data and effectively solving the data silo problem. Furthermore, the method of this application extracts features based on the platform behavioral data of target customers to obtain the target customers' car purchase behavior characteristics and social behavior characteristics, and further analyzes the individual purchase intentions and group purchase influence of target customers to determine whether the target customers are potential high-value customers. By analyzing target customers from both individual and group perspectives, sales orders can be effectively facilitated. Compared with traditional methods that only consider the target customers themselves, potential customers can be identified more accurately, thereby improving the car sales conversion rate.

[0029] The potential customer discovery method provided in this application relates to the field of potential customer identification technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or in-vehicle terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the potential customer discovery method, but is not limited to the above forms.

[0030] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0031] It should be noted that in all specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirection to confirmation pages. Only after obtaining the user's separate permission or consent is the necessary user-related data required for the proper functioning of these embodiments acquired.

[0032] Figure 1 This is an optional flowchart of the potential customer mining method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S100 to S300.

[0033] Step S100: Obtain the platform behavior data of the first object; the platform behavior data is obtained by unifying the identity of the original behavior data associated with the first object obtained from several different platforms.

[0034] This embodiment adopts a cross-platform customer identity unified identification method in the automotive industry. It obtains the target customer's platform behavior data by acquiring the original behavioral data associated with the target customer from multiple platforms and unifying the identity. This step can realize the data fusion of multiple platform data and unify the identity identification, thereby effectively solving the data silo problem.

[0035] Step S200: Based on the platform behavior data, feature extraction is performed to obtain the car purchase behavior features and social behavior features of the first object; the car purchase behavior features are related to individual vehicle purchase behavior; the social behavior features are related to group vehicle purchase behavior.

[0036] After obtaining the platform behavior data of the target customers, feature extraction is performed to obtain behavioral characteristics at the individual and group levels. These characteristics include car purchase behavior related to individual vehicle purchase behavior and social behavior related to group vehicle purchase behavior. These serve as the data basis for subsequent potential customer identification.

[0037] Step S300: Determine the individual purchase intention of the first target based on car purchase behavior characteristics; determine the group purchase influence of the first target based on social behavior characteristics; analyze the individual purchase intention and group purchase influence to determine the customer status of the target; the customer status indicates whether the first target is a potential customer who meets the requirements.

[0038] Based on the feature extraction results obtained above, the individual purchasing intentions of target customers are analyzed from the individual perspective, and the group purchasing influence of target customers is analyzed from the group perspective. Thus, a comprehensive analysis is conducted to determine whether the target customers are potential high-value customers, that is, customers who can effectively facilitate car sales orders.

[0039] In some embodiments, in step S100, the platform data of the first object is obtained in the following manner: Step S110: Obtain raw behavioral data associated with the first object from at least two different platforms, each platform having its own independent user identification system.

[0040] Raw behavioral data related to target customers is obtained from different platforms as the data foundation. At the same time, each platform has an independent user identification system, which is the direct cause of data silos.

[0041] Step S120: For each platform, based on the original behavioral data obtained from the platform, extract a set of identity features and a set of behavioral sequence features, and generate a user fingerprint vector representing the first object within the platform based on the set of identity features and behavioral sequence features.

[0042] This invention uses an algorithm of "multi-dimensional identity matching + behavioral fingerprint comparison" to unify the same user appearing on different platforms into a single GlobalCustomerId. First, it determines the identity features corresponding to each user on each platform, i.e., the user fingerprint vector. Specifically, the identity features used may include: accountId (enterprise ID), openId (unique ID within the platform), deviceId (device fingerprint), geo (geographical approximation), and behavioral fingerprints (behavioral sequence features, such as viewing duration distribution, dwell time rhythm, and nighttime activity patterns).

[0043] The user fingerprint vector generated from data from each source can be calculated using this algorithm: UserFingerprint = f(openId, deviceId, behaviorPattern).

[0044] Step S130: Calculate the similarity between user fingerprint vectors from different platforms. If the similarity calculation result is higher than a preset threshold, it is determined that the user fingerprint vectors participating in the similarity calculation correspond to the same object.

[0045] After obtaining user identifiers from different platforms, similarity calculations are performed on each pair of identifiers to confirm that these identifiers belong to the same user. When the similarity exceeds a threshold, the two user identifiers involved in the similarity calculation are determined to belong to the same user.

[0046] For example, the similarity function can be used to calculate whether two users are the same person using the following algorithm: similarity = cosine(fingerprintA, fingerprintB).

[0047] Step S140: Generate a global identity identifier for each user fingerprint vector that is determined to be the same object.

[0048] To generate a unified global identity identifier GlobalCustomerId from user fingerprint vectors that belong to the same object, this can be achieved using the following algorithm: GlobalCustomerId = hash(openId + deviceId + platform).

[0049] Step S150: Based on the global identity identifier, the original behavioral data from different platforms are aggregated according to the time sequence of the behavior to obtain the platform behavioral data of the first object.

[0050] Finally, the raw data from each platform belonging to the target user are aggregated according to the time series to construct the complete behavioral trajectory of the target customer and obtain the platform behavior data of the target user, which serves as the data foundation for subsequent potential customer identification.

[0051] Specifically, the data sources for each platform include, but are not limited to: customer behavior data such as bullet comments, likes, and follows on live streaming platforms; behavior data such as comments, shares, and favorites on short video platforms; and data such as private message conversations and consultation records with car sales consultants.

[0052] In some instances, this example employs an intelligent multimodal data fusion algorithm specific to the automotive industry to fuse acquired platform behavior data. This algorithm can fuse acquired text, images, and videos, thereby improving the accuracy of potential customer intent recognition. The specific steps are as follows: (1) Automotive industry-specific cross-modal alignment strategies: The method in this embodiment does not simply concatenate modalities such as text, images, and videos, but rather: Component recognition (headlights, grille, spoiler, etc.) from car images; Perform entity recognition (model, configuration, price) on vehicle descriptions in text. Then perform cross-modal alignment between the visual components and the text description. For example, the text "The champion version of the headlights looks pretty good" will be automatically matched with the "headlight area features" in the image.

[0053] Cross-modal alignment strategies can significantly improve intent recognition accuracy.

[0054] (2) The temporal synchronization and intention reinforcement mechanism of multi-source behavior: Users' behavior timestamps are inconsistent across different platforms (live streaming, short videos, private messages). The method in this embodiment achieves cross-platform behavior data synchronization and precise fusion by setting a time-series alignment model, behavior priority rules, and intent enhancement factors (such as high weight for inquiry behavior).

[0055] (3) Dynamic weighting strategy for inter-modal confidence: Different modalities contribute differently to intent recognition: Text modality is most sensitive to car purchase intentions; Video modality reflects the user's actual time spent and focus; The audio modality can identify emotions and urgency; The method in this embodiment automatically learns modal weights through a linkage training mechanism, rather than manually setting them.

[0056] (4) Automotive customer intent recognition algorithm This example method uses a multimodal deep learning model for automotive customer intent recognition, and can improve recognition accuracy by incorporating an attention mechanism.

[0057] In some embodiments, step S300 involves determining the individual purchase intention of the first object based on car purchase behavior characteristics, including but not limited to steps S311 to S313: Step S311: Calculate the initial car purchase intention score based on the preset multi-factor scoring model and car purchase behavior characteristics, and correct the initial car purchase intention score based on the preset special scoring rules. Obtain the car purchase intention score of the first object based on the correction result. The correction logic in the preset special scoring rules is based on one or more of the following: the market popularity of the car model of interest, the sensitivity to price, and the degree of coincidence between the time of car purchase-related behavior and the promotion cycle.

[0058] The method of this invention analyzes the individual purchase intention of target customers, including two parts: purchase intention score and purchase intention status assessment. The purchase intention score is obtained by a basic score plus a revised score. The basic score (initial purchase intention score) is calculated by a multi-factor scoring model, while the revised score is calculated by a preset special scoring rule.

[0059] Step S312: Match the car purchase intention status of the first object with the preset intention stage identification rules and car purchase behavior characteristics to determine the car purchase intention status; the car purchase intention status includes the initial interest period, the in-depth understanding period and the purchase intention period.

[0060] Furthermore, the target customer's purchase intention status is obtained by analyzing the results of matching the preset intention stage identification rules with purchase behavior characteristics. Among them, the intention stages (purchase intention status) provided by the method of this application include, but are not limited to, the initial interest period, the in-depth understanding period, and the purchase intention period, which correspond to low, medium, and high intentions, respectively.

[0061] In some embodiments, the analysis of a target customer's car purchase intention can be performed from the following aspects: (1) Feature item → Description; (2) Depth of behavior → Number of words in the comment, professionalism of the consultation content; (3) Frequency of behavior → Number of views, number of follow-up visits; (4) Price relevance → Whether prices, discounts, and inventory are mentioned; (5) Vehicle model focus → Whether to focus on a single vehicle model; (6) Behavioral time interval → Whether it occurs frequently in a short period of time.

[0062] In some embodiments, the various car purchase intention statuses are described as follows: (1) Initial interest period → low intention Behavioral characteristics: Browsing multiple brand contents in short videos or live streams, with behavior mainly consisting of superficial interactions such as "liking" and "pausing briefly".

[0063] Identification signals: Disorganized behavior, unfocused interactive content, and no clear brand or model preference has been revealed.

[0064] (2) In-depth understanding period → moderate interest Behavioral characteristics: The behavior shifted from "general viewing" to "focused viewing". They began to repeatedly watch videos of specific car models, inquire about specific configuration parameters in live broadcasts, and conduct preliminary professional consultations through private messages.

[0065] Identification signals: The depth of interactive content has increased significantly, with topics focusing on professional areas such as the performance and configuration of specific car models.

[0066] (3) Purchase intention period → High intention Behavioral characteristics: The interactive behavior revolves around the "transaction" itself, and the core manifestation is that in private messages or live consultations, they repeatedly ask about the final transaction conditions such as price, discounts, availability of vehicles, and delivery period.

[0067] Recognizing signals: The communication content shifts from "Is the product good or bad?" to "When to buy and how much?", releasing clear decision and purchase signals.

[0068] Step S313: Analyze the purchase intention score and purchase intention status to determine the individual purchase intention of the first target.

[0069] Finally, based on the target customers' car purchase intention scores and purchase intention status, an individual profile of the target customers is created to determine their individual purchase intentions, that is, the car purchase intentions starting from the customer themselves.

[0070] In some embodiments, step S311 involves a pre-defined multi-factor scoring model whose scoring factors include basic behavioral factors, content depth factors, purchase intention factors, and time-sensitive factors; individual vehicle purchase behavior includes browsing, interaction, and consultation; vehicle purchase behavior characteristics indicate the content, frequency, and duration of individual vehicle purchase behavior; and the process of calculating an initial vehicle purchase intention score based on the pre-defined multi-factor scoring model and vehicle purchase behavior characteristics includes, but is not limited to, the following steps: (1) The basic behavior factor score is calculated based on the frequency and duration of individual vehicle purchase behavior; the content depth factor score is calculated based on the depth of the content browsed and the quality of the interaction; the purchase intention factor score is calculated based on the judgment result and the depth of the content consulted, and the purchase intention factor score is calculated based on the judgment result and the depth of the content consulted; the time sensitivity factor score is calculated based on the degree of coincidence between the last active time in the duration of vehicle purchase behavior and the car promotion period.

[0071] The method in this embodiment calculates the target customer's car purchase intention score using a multi-factor scoring model. In one embodiment, the calculation method of the car purchase intention score and the weight of each factor can be as follows: Basic Behavioral Score (0-40 points): Calculated based on the frequency and activity of car customer behavior; Content Depth Score (0-30 points): Calculated based on the depth of content viewed and the quality of interaction by automotive customers; Purchase Intention Score (0-20 points): Calculated based on car customer inquiries and purchase-related behaviors; Time-sensitive score (0-10 points): Calculated based on the car customer's last active time and the urgency of the purchase.

[0072] (2) The initial car purchase intention score is obtained by weighting the scores of basic behavioral factors, content depth factors, purchase intention factors and time sensitivity factors.

[0073] By aggregating the scores of each factor, the initial car purchase intention score of the target user is obtained, in order to assess the user's car purchase intention, which is the potential customer value score of the individual user.

[0074] In some embodiments, the special scoring rules are described as follows: 1) Model popularity: Considering the market popularity of the models being viewed, customers who are interested in popular models receive higher scores.

[0075] The bonus points for selected car models are calculated by multiplying a base score by a popularity coefficient. This popularity coefficient is based on a dynamically updated car popularity ranking, with data sources including actual sales figures for each model (from the internal sales system), exposure and discussion volume on major content platforms (obtained through web scraping and NLP analysis), advertising spending, and search indexes. Each car model has a base popularity coefficient (e.g., 0.8-1.2). For example, the coefficient for the popular model A is 1.2, while the coefficient for a less popular model that is about to be discontinued is 0.8. If a customer shows interest in a popular model, they not only receive the base score for "interest in the model" but also an additional popularity bonus from the popularity coefficient.

[0076] 2) Price sensitivity: Consider the degree of price sensitivity. Customers who are price-sensitive will score higher.

[0077] In this embodiment, the method identifies pricing intent in customer interactions using natural language processing algorithms, including both content depth and behavioral frequency and timing. In some embodiments: In terms of content depth (different levels of sensitivity correspond to different price sensitivity scores): Superficial price inquiry (low sensitivity): For example, simply asking "How much?" during a live stream. In-depth price comparison (medium to high sensitivity): For example, "How much difference is there between the Champion Edition and the older model? What upgrades are included?" Seeking discounts (highly sensitive): For example, "Other stores offered me a price of XX, can you offer a lower price?" Regarding frequency and timing of actions: The price sensitivity score is calculated by combining the frequency of actions (multiple inquiries about prices within a short period) and the timing (following up on discounts at the end of the promotional season). Customers who are still eager to inquire about prices after the promotional period ends receive the highest scores.

[0078] 3) Timing of purchase: Consider the degree of coincidence between the time of car purchase-related behavior and the promotion period. Customers who take car purchase-related behaviors during the car promotion period have higher scores.

[0079] Identify whether customer behavior aligns with the company's sales rhythm and marketing activities. Customers who are active at key marketing moments have a higher conversion rate.

[0080] In some embodiments, the weights in the weighted calculation process of the car purchase intention score described above can be adaptively adjusted using a dynamic weight adjustment model to achieve a change from static configuration to dynamic optimization. The model is constructed based on logistic regression or gradient boosting decision trees and is trained and updated intermittently using phased data (various scores during the potential customer period and the final conversion status).

[0081] In some embodiments, the method of this embodiment can identify the potential customer's emotional state and emotional changes under different content, thereby uncovering the potential customer's implicit intentions and more accurately identifying the potential customer's car purchase intentions, including the following steps: (1) Multimodal acquisition and preprocessing of emotional signals Text emotion features originate from: comments, bullet comments, and private messages; Handling method: Use industry sentiment dictionaries (such as sentiment indicators like "in stock", "price increase", "discount" and "delivery"); Use the BERT sentiment classification model to identify sentiment polarity (positive, neutral, negative); Extract industry-specific emotional tags such as "anxiety," "hesitation," and "strong purchase intention" from semantics.

[0082] (2) Cross-modal emotion alignment and fusion This embodiment uses a multimodal Transformer + attention mechanism for cross-modal emotion fusion to solve the problem of "inconsistent expression across different modalities". By determining which modality contributes the most to emotion judgment, the complementary relationship between different modalities is determined (e.g., rational words, but urgent voice → overall judgment of "urgency"), making emotion judgment no longer dependent on a single modality, and more stable and accurate.

[0083] (3) Temporal Emotion Tracking In the context of car purchase, the method in this embodiment continuously tracks the emotional changes of the same potential customer over time.

[0084] (4) The mapping from emotion to car buying intention (Emotion → Buying Intent) This embodiment uses an Emotion-to-Intent Network to automatically determine the car purchase intention based on emotional patterns.

[0085] In response, the method in this embodiment transforms various emotional representation data into a deep and comprehensive understanding of users' car purchase intentions through a cross-modal alignment and fusion model, achieving a technological leap from "understanding words" to "understanding hearts".

[0086] In some embodiments, step S200 includes the platform data of the first object on various platforms; the social behavior characteristics of the first object are obtained in the following ways: Step S210: Based on the social relationships of the first object on various platforms, identify several second objects that have explicit social relationships with the first object.

[0087] Identify the target customer's explicit social relationships by analyzing their social connections on various platforms, such as friends / followers.

[0088] Step S220: Analyze the platform behavior data of each object on each platform to identify several third objects that have an implicit relationship with the first object; the similarity between the third objects and the first object on each platform in terms of browsing records, interaction records and consultation records meets the preset requirements.

[0089] On the other hand, by analyzing the platform behavior data of various users on different platforms, social contacts with implicit relationships to target customers can be identified. Specifically, this involves identifying common characteristics within groups, such as browsing history with the same objects or time periods, discussions about the same car model, or inquiries about the same car model.

[0090] Step S230: Based on the interaction records between the first object and each second object and each third object, feature extraction is performed to obtain the interaction depth feature and interaction frequency feature of the first object; the interaction depth feature indicates the text, sound and video content related to the interaction between objects; the interaction frequency feature indicates the frequency and time distribution of the interaction between objects.

[0091] Furthermore, features are extracted based on the content, frequency, or market distribution of interactions between target users and explicit or implicit interactive objects. The extracted features are then used as the data foundation for constructing the social behavior characteristics of target users.

[0092] Step S240: Construct a time series based on interaction depth features and interaction frequency features, and obtain the social behavior features of the first object based on the construction results.

[0093] By aggregating relevant characteristic data of target users in a time series manner, we can obtain the social behavior characteristics of users, which serves as the data basis for subsequent analysis of their group purchasing influence.

[0094] In some embodiments, step S300, the process of determining the group purchasing influence of the first object based on social behavioral characteristics, includes, but is not limited to, steps S321 to S323: Step S321: Based on social behavior characteristics, determine several interactive objects that have social relationships with the first object, as well as the interaction between the first object and each interactive object.

[0095] Analyzing the social and interactive relationships among car customers serves as the data foundation for subsequently identifying the purchasing influence of target user groups.

[0096] Step S322: Taking the first object as the main point and several interactive objects as several secondary points, construct connecting edges between the main point and each secondary point according to the interaction between the first object and each interactive object, so as to construct the social relationship graph of the first object; the type of connecting edge is determined by the type of interaction channel between the first object and the interactive objects.

[0097] This example method analyzes the impact on group purchasing by constructing a social relationship graph of the target customers. Specifically, the steps for constructing a social relationship graph (or customer relationship network graph) include: (1) Node construction: Each potential car customer is represented as a graph node, and its node features include: user interest vector, historical behavior features, intention score, and browsing car model preference encoding.

[0098] (2) Construction of edges: Create different types of edges based on interaction behavior across different channels, for example: Comment interaction side: Two users interact multiple times under the same car model content; Q&A session: A answers B's car buying questions; Model co-occurrence: Two users are continuously monitoring the same car model; Price discussion side: Users discuss the same price range.

[0099] Each edge has a weight: Weight = Frequency of behavior × Intent intensity × Channel confidence.

[0100] (3) Graph construction This ultimately forms a heterogeneous graph with multiple edge types.

[0101] In response, by constructing a social relationship graph of the target users, we can achieve the following: 1) automatically discover "hidden relationship chains" (such as those who have never interacted directly but have similar behaviors); 2) build a more accurate user relationship network based on the behavior types in the automotive industry; and 3) provide structured input for subsequent GNN propagation, significantly improving prediction accuracy.

[0102] Step S323: Input the social relationship graph of the first object into the preset graph neural network model, aggregate the interactive information features transmitted between the main point and several secondary points based on the information transmission mechanism of the graph neural network model, and output the group purchase influence of the first object based on the aggregation result.

[0103] Finally, the social relationship graph of the target user is analyzed using a pre-defined graph neural network model (GNN model can be used) to analyze its influence in the social network.

[0104] In some embodiments, traditional centrality metrics are static. GNNs, however, can dynamically learn the influence of nodes through a message-passing mechanism. The steps for analyzing the group purchasing influence of target users using a GNN model are as follows: 1) Message passing: Each node (customer) receives information from its neighboring nodes. For example, a node that owns car model A will "pass on" positive experience information about "car model A" to all its fan nodes.

[0105] 2) Aggregation and Update: The GNN model aggregates all incoming messages (e.g., by weighted summation) and combines them with the node's own attributes to update the node's hidden state (i.e., a richer feature vector). The more neighbors a node has, the greater the influence of its neighbors, and the more frequent the interactions, the more important the information it aggregates becomes.

[0106] 3) Influence Score Calculation: After multiple iterations of the GNN, each node learns a final feature vector. This vector is input into a prediction layer, which outputs a scalar, representing the customer's influence score. This score considers not only the number of followers (degree centrality) but also deeply integrates the customer's network position, interaction quality, and the influence of their followers.

[0107] Furthermore, based on the group purchasing influence of individual users within the group, it is possible to predict group car purchasing behavior and identify key influencers: 1) Community Discovery: Using graph clustering algorithms (such as the Louvain algorithm), closely connected "small groups" or "communities" are automatically identified in a large customer network. These communities are usually formed based on real social circles (such as family, colleagues, car clubs).

[0108] 2) Group Behavior Prediction: When a certain percentage of members in a community (e.g., predicted by GNN) show a high intention for a particular car model, the system determines that the entire community has a high risk / high opportunity for group car purchases. The model analyzes the influence flow path within the community and predicts how car purchase intentions spread within the community through key figures.

[0109] 3) Identify key influencers: Within each community, the system ranks individuals based on their influence scores calculated by the GNN. The individuals with the highest scores are identified as key influencers or opinion leaders within that community. Marketing resources should be prioritized for these individuals.

[0110] In response, the method in this embodiment constructs a customer relationship graph and applies GNN, which not only sees the "trees" (individual customers) but also the structure and dynamics of the "forest" (customer relationship network). This enables precise positioning, leverage amplification, and group prediction in potential customer mining, which is a fundamental leap forward from traditional CRM systems.

[0111] In some embodiments, the method of this embodiment further includes: Step S410: Obtain the historical car purchase behavior characteristics, real-time car purchase intention status, and real-time external market condition characteristics of the first object; the historical car purchase behavior characteristics are obtained by feature extraction based on the historical platform behavior data of the first object within a certain period of time.

[0112] We acquire the target user's historical behavior sequence, real-time car purchase intention status, and external market conditions as the data basis for subsequent prediction of their car purchase intention status conversion.

[0113] Step S420: Input the historical car purchase behavior characteristics, real-time car purchase intention status and real-time external market condition characteristics into the preset state prediction model for calculation to obtain the predicted car purchase intention status of the first object.

[0114] The aforementioned data is input into a preset state prediction model for calculation, thereby obtaining the prediction results of the target customer's car purchase intention status, which serves as a data reference for determining their marketing strategy.

[0115] Step S430: Analyze the real-time and predicted car purchase intention status of the first target to determine the target marketing strategy to be delivered to the first target.

[0116] Based on the target customer's real-time car purchase intention status and the potential change in car purchase intention status, targeted marketing strategies are selected and delivered to the customer, realizing the recommendation of the best marketing strategy based on the current status of the potential car customer and the prediction of the best marketing time for the potential customer.

[0117] In some embodiments, the process of determining the appropriate marketing strategy based on the current purchase intention status of potential car buyers is as follows: (1) Strategy knowledge base construction: First, a structured strategy knowledge base is established. Each strategy includes: strategy content (e.g., pushing limited-time coupons, inviting test drives, pushing model comparison reports), applicable potential customer status (initial interest period, in-depth understanding period, purchase intention period, etc.), trigger conditions, and priority and cooldown time (to avoid repeated disturbances).

[0118] (2) The potential customer status and conversion probability are calculated in real time by the model. Real-time monitoring of user behavior, and invocation of the state prediction model preset by the method in this embodiment: The Transformer / LSTM algorithm is used to analyze the latest behavioral sequence and the current potential customer state; it also outputs the predicted conversion probability of the potential customer state in the next stage.

[0119] (3) Strategy matching and trigger determination (core implementation) The strategy engine compares the input against the strategy knowledge base item by item, and the matching logic is as follows: For each strategy in StrategyBase: If the user's current state is equal to strategy.requiredState; And all triggerRules are satisfied; → Add this strategy to the list of executable strategies.

[0120] (4) Best Timing Prediction The system monitors trends in user behavior to determine if there is a "leap signal" in the conversion probability.

[0121] If any one of the conditions is met, it is determined to be the "best time to sell": The conversion probability increases significantly in a short period of time (e.g., ≥40% increase within 1 hour). The user performs multiple decision-making actions (multiple inquiries about prices, requests for quotations, etc.) within 10 minutes. The timing of the action is close to key points in the marketing calendar (such as the end of the month or holidays).

[0122] (5) Final strategy selection (with priority) If multiple strategies simultaneously meet the triggering conditions, the system will automatically select the appropriate strategy in the following order: Strategy priority (high > medium > low); User history preferences (such as a greater willingness to accept "test drive invitations"); Predicting maximum return (the expected return value given by the reinforcement learning model). Ultimately, it produces one or more optimal strategies.

[0123] In some embodiments, the method of this embodiment provides a federated learning framework and related technologies for protecting the privacy of automotive customers, enabling data mining and model training while protecting the privacy of automotive customers; its specific technical implementation includes: (1) Automotive Customer Federated Learning Framework Step 1: Training on local data (without leaving the domain) Each car dealership or regional node (such as a dealership's local system) will complete model training locally, without including any external data, and the original customer data will not leave the local area.

[0124] Step 2: Upload model parameters (not raw data) After local training is completed, only the model gradients or parameters are uploaded. All parameters have been privacy-protected before uploading (see Differential Privacy Technology below for details).

[0125] Step 3: The federated server performs parameter aggregation (Federated Averaging algorithm). The central server uses Federated Averaging to aggregate parameters uploaded by all nodes. After aggregation, the model is updated to a global model. Federated Averaging (FedAvg) is the core collaborative training algorithm of federated learning. It allows multiple clients (such as mobile phones and enterprises) to jointly train a global model without sharing their original local data.

[0126] Step 4: Deploy the global model, and the nodes continue training (iteration). The aggregated model is then distributed back to all dealer nodes, which continue to train based on their local data, thus continuously improving the global model.

[0127] (2) Differential privacy technology for car customers Step 1: Gradient Clipping, which limits the gradients obtained in each training session to a fixed range to avoid exposing user behavior features with excessively large gradients; Step 2: Add noise (add privacy protection) by adding random noise, such as Gaussian noise or Laplacian noise, to the clipped gradient; Step 3: Upload the noisy gradient so that even if a hacker intercepts the gradient, they cannot infer user behavior.

[0128] (3) Homomorphic encryption for car customers Step 1: Local gradient encryption: Each node encrypts its own gradients; Step 2: The federated server aggregates parameters in an encrypted state, and the server can perform addition without decryption; Step 3: The updated encryption model can only be decrypted by authorized nodes.

[0129] In summary, the embodiments of this application include, but are not limited to, the following beneficial effects: 1. Unified Customer Identification Method Across Platforms in the Automotive Industry: Achieve unified identification of automotive customers across different platforms through multi-dimensional identification, thus solving the problem of siloed automotive sales data.

[0130] 2. Intelligent fusion algorithm for multimodal data in the automotive industry: Integrates multimodal data such as text, images, and videos for in-depth analysis to improve the accuracy of identifying potential customers for automobiles.

[0131] 3. Real-time prediction method for potential customer status changes in the automotive industry: Based on machine learning models, predict changes in the status of potential car customers in real time to provide the best sales opportunity.

[0132] 4. Multi-dimensional potential customer value scoring algorithm for the automotive industry: Combines multiple factors such as behavior frequency, content depth, purchase intention, and time sensitivity to conduct a comprehensive score.

[0133] 5. Automotive customer social relationship mining based on graph neural networks: Construct automotive customer relationship graphs to analyze influence spread and predict group car purchase behavior.

[0134] 6. Optimization of automotive industry marketing strategies using reinforcement learning: Use deep reinforcement learning to optimize automotive marketing strategies in real time and improve conversion rates.

[0135] 7. Federated Learning Framework for Car Customer Privacy Protection: Data mining and model training are conducted while protecting the privacy of car customers.

[0136] Please see Figure 2 This application also provides a potential customer discovery system that can implement the above-described method. The system includes: The data acquisition module is used to acquire the platform behavior data of the first object; the platform behavior data is obtained by unifying the identity of the original behavior data associated with the first object obtained from several different platforms. The feature extraction module is used to extract features based on platform behavior data to obtain the car purchase behavior features and social behavior features of the first object; the car purchase behavior features are related to individual vehicle purchase behavior; the social behavior features are related to group vehicle purchase behavior. The analysis module is used to determine the individual purchase intention of the first target based on car purchase behavior characteristics; determine the group purchase influence of the first target based on social behavior characteristics; analyze the individual purchase intention and group purchase influence to determine the customer status of the target; and the customer status indicates whether the first target is a potential customer who meets the requirements.

[0137] It is understood that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0138] Please see Figure 3 This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0139] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0140] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0141] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0142] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0143] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0144] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0145] This application provides a method, apparatus, electronic device, storage medium, and program product for identifying potential customers. This solution obtains initial behavioral data associated with target customers from multiple different platforms and unifies their identities, thereby acquiring platform behavioral data of the target customers. This achieves data fusion of independent platform data, effectively solving the data silo problem. Furthermore, the method extracts features from the target customers' platform behavioral data to obtain their car-buying behavior characteristics and social behavior characteristics. It further analyzes the individual purchase intentions and group purchase influence of the target customers to determine whether they are potential high-value customers. By analyzing target customers from both individual and group perspectives, sales orders can be effectively facilitated. Compared to traditional methods that only consider the target customer themselves, this approach more accurately identifies potential customers, thereby improving car sales conversion rates.

[0146] The embodiments described in 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 evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0147] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0148] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0149] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0150] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0151] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0152] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0153] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0154] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0155] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for identifying potential customers, characterized in that, The method includes the following steps: Obtain platform behavior data of the first object; the platform behavior data is obtained by unifying the identity of the original behavior data associated with the first object obtained from several different platforms. Based on the platform behavior data, feature extraction is performed to obtain the car purchase behavior features and social behavior features of the first object; the car purchase behavior features are related to individual vehicle purchase behavior; the social behavior features are related to group vehicle purchase behavior. Determine the individual purchase intention of the first object based on the aforementioned car purchase behavior characteristics; The group purchasing influence of the first target is determined based on the aforementioned social behavior characteristics; Based on the analysis of individual purchasing intentions and the influence of group purchasing, the customer status of the object is determined; the customer status indicates whether the first object is a potential customer who meets the requirements.

2. The method according to claim 1, characterized in that, The platform data of the first object is obtained in the following way: Raw behavioral data associated with the first object are obtained from at least two different platforms, each of which has its own independent user identification system; For each platform, based on the original behavioral data obtained from the platform, a set of identity features and a set of behavioral sequence features are extracted, and a user fingerprint vector representing the first object within the platform is generated based on the set of identity features and behavioral sequence features. The similarity between user fingerprint vectors from different platforms is calculated separately. If the similarity calculation result is higher than a preset threshold, it is determined that the user fingerprint vectors participating in the similarity calculation correspond to the same object. Generate a global identity identifier for each user fingerprint vector that is identified as the same object; Based on the global identity identifier, the original behavioral data from different platforms are aggregated in chronological order of the occurrence of the behaviors to obtain the platform behavioral data of the first object.

3. The method according to claim 1, characterized in that, Determining the individual purchase intention of the first object based on the car purchase behavior characteristics includes: An initial car purchase intention score is calculated based on a preset multi-factor scoring model and the car purchase behavior characteristics. The initial car purchase intention score is then corrected based on preset special scoring rules. The car purchase intention score of the first object is obtained based on the correction result. The correction logic in the preset special scoring rules is based on one or more of the following: the market popularity of the car model of interest, the sensitivity to price, and the degree of coincidence between the time of car purchase-related behavior and the promotion cycle. The purchase intention status of the first object is determined by matching the car purchase behavior characteristics with the preset intention stage identification rules; the purchase intention status includes the initial interest stage, the in-depth understanding stage, and the purchase intention stage. Based on the car purchase intention score and car purchase intention status, the individual purchase intention of the first object is determined.

4. The method according to claim 3, characterized in that, The pre-defined multi-factor scoring model includes basic behavioral factors, content depth factors, purchase intention factors, and time-sensitive factors. The individual vehicle purchase behavior includes browsing, interaction, and consultation; the vehicle purchase behavior characteristics indicate the content, frequency, and duration of the individual vehicle purchase behavior. The initial car purchase intention score is calculated based on a preset multi-factor scoring model and the car purchase behavior characteristics, including: The basic behavioral factor score is calculated based on the frequency and duration of individual vehicle purchases. A content depth factor score is calculated based on the depth of the content browsed and the quality of the interaction. Determine whether there are keywords related to the car purchase process in the consultation content, and calculate the purchase intention factor score based on the judgment result and the depth of the consultation content. The time sensitivity factor score is calculated based on the degree of overlap between the last active time in the distribution of vehicle purchase behavior and the car promotion period. The initial car purchase intention score is obtained by weighting the scores of the basic behavioral factors, the content depth factors, the purchase intention factors, and the time sensitivity factors.

5. The method according to claim 1, characterized in that, The platform data includes the first object's social relationships on various platforms; the first object's social behavior characteristics are obtained through the following methods: Based on the social relationship data of the first object on various platforms, identify several second objects that have explicit social relationships with the first object; Based on the analysis of platform behavior data of each object on each platform, several third objects with implicit relationships to the first object are identified; the similarity between the third objects and the first object on each platform in terms of browsing records, interaction records and consultation records meets the preset requirements. Based on the interaction records between the first object and each of the second objects and between each of the third objects, feature extraction is performed to obtain the interaction depth feature and interaction frequency feature of the first object; The interaction depth feature indicates the text, audio, and video content related to the interaction between objects; the interaction frequency feature indicates the frequency and time distribution of the interaction between objects. A time series is constructed based on the interaction depth features and interaction frequency features, and the social behavior features of the first object are obtained based on the construction results.

6. The method according to claim 1, characterized in that, The determination of the group purchasing influence of the first object based on the social behavior characteristics includes: Based on the social behavior characteristics, a number of interactive objects that have social relationships with the first object are determined, as well as the interaction between the first object and each of the interactive objects; Using the first object as the main point and several interactive objects as several secondary points, several different types of connection edges are constructed between the main point and each of the secondary points according to the interaction between the first object and each of the interactive objects, so as to construct the social relationship graph of the first object; the type of the connection edge is determined by the type of interaction channel between the first object and the interactive objects. The social relationship graph of the first object is input into a preset graph neural network model. Based on the information transmission mechanism of the graph neural network model, the interactive information features transmitted between the main point and several secondary points are aggregated. Based on the aggregation result, the group purchase influence of the first object is output.

7. The method according to claim 3, characterized in that, The method further includes: The system acquires the historical car purchase behavior characteristics, real-time car purchase intention status, and real-time external market condition characteristics of the first object; the historical car purchase behavior characteristics are obtained by feature extraction based on the historical platform behavior data of the first object within a certain period of time. The historical car purchase behavior characteristics, the real-time car purchase intention status, and the real-time external market condition characteristics are input into a preset state prediction model for calculation to obtain the predicted car purchase intention status of the first object. Based on the analysis of the first object's real-time car purchase intention status and the predicted car purchase intention status, a target marketing strategy is determined and delivered to the first object.

8. A potential customer discovery system, characterized in that, The system includes: The data acquisition module is used to acquire platform behavior data of the first object; the platform behavior data is obtained by unifying the identity of the original behavior data associated with the first object obtained from several different platforms. The feature extraction module is used to extract features based on the platform behavior data to obtain the car purchase behavior features and social behavior features of the first object; the car purchase behavior features are related to individual vehicle purchase behavior; the social behavior features are related to group vehicle purchase behavior. The analysis module is used to determine the individual purchase intention of the first object based on the car purchase behavior characteristics; determine the group purchase influence of the first object based on the social behavior characteristics; and analyze the individual purchase intention and the group purchase influence to determine the customer status of the object; the customer status indicates whether the first object is a potential customer who meets the requirements.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images