Online laundry potential customer stratified marketing lead system and method

Abstract

The present application discloses an online laundry potential customer hierarchical marketing lead system and method, relates to the technical field of marketing management, and the present application collects the basic information, order data and behavior data of all users in the network, processes the collected data through a data processing method, trains the processed data through an AI large model after processing is completed, generates a user portrait according to the training results and user basic information, determines potential users based on the generated user portrait, further performs deep mining on the potential users through a data analysis method, constructs a potential user hierarchical system, sets a marketing lead strategy based on the constructed potential user hierarchical system, collects marketing lead strategy feedback data in real time, finally constructs a marketing lead model based on the real-time collected marketing lead strategy feedback data, and performs real-time optimization adjustment and execution through a model evaluation method, thereby improving the accuracy and real-time performance of potential customer hierarchical marketing lead.

Description

Technical Field

[0001] This invention relates to the field of marketing management technology, specifically to a tiered marketing and lead generation system and method for online laundry potential customers. Background Technology

[0002] Traditional online laundry marketing methods employ a broad-based approach: ① Recommended solutions often fail to meet customer needs, significantly reducing marketing success rates and resulting in a severe waste of touchpoint resources. ② They rely solely on promotional methods to attract customers with specific needs, leading to widespread ineffective advertising and substantial wasted promotional costs. Summary of the Invention

[0003] The purpose of this invention is to provide a tiered marketing and lead generation system and method for online laundry potential customers, which solves the problems existing in the background art.

[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a method for tiered marketing and lead generation of potential online laundry customers, specifically including the following steps: S1. Collect basic information, historical order data, and historical behavior data of all users in the network, and process the collected historical order data and historical behavior data through data processing methods to obtain processed historical order data and historical behavior data; S2. Train the processed historical order data and historical behavior data using a large AI model, and generate user profiles based on the training results and basic user information; S3. Based on the generated user profile, potential users are initially identified; S4. Conduct in-depth analysis of the initially identified potential users through data analysis and construct a potential user segmentation system; S5. Set up marketing and lead generation strategies based on the constructed potential user segmentation system, and collect marketing and lead generation strategy feedback data in real time; S6. Build a marketing lead generation model based on real-time collected marketing lead generation strategy feedback data, and optimize and execute the built marketing lead generation model in real time through model evaluation.

[0005] Preferably, the process of collecting basic information, historical order data, and historical behavior data of all users within the network, and then processing the collected historical order data and historical behavior data to obtain the processed historical order data and historical behavior data includes the following steps: S11. Set the collection period and collect basic information of users on the website based on the set collection period; Set the basic information as the user's unique identification information; The basic information includes: age information, device number information, and gender information; S12. Collect historical behavior data and historical order data of the corresponding user based on the user's basic information; The historical order data includes: laundry order time, laundry order quantity, and laundry order amount; The historical behavioral data includes: search behavior data, collection behavior data, sharing behavior data, and payment behavior data; The collected historical behavior data is aggregated and a dataset is constructed based on the historical behavior data type. ; in, The table contains a collection of historical behavioral data. This represents payment behavior data from historical behavioral data. This represents search behavior data from historical behavioral data. This represents the collection behavior data in historical behavior data. This represents sharing behavior data within historical behavioral data; S13. The collected historical behavior data and historical order data are processed through data processing methods to obtain processed historical behavior data and historical order data.

[0006] Preferably, the process of processing the collected historical behavior data and historical order data to obtain processed historical behavior data and historical order data includes the following steps: Set the weight of behavioral data to search < favorite < share < pay, and set the score value of the four behaviors to [1,2,3,4] based on the weight of behavioral data. When a user performs any of the above four behaviors on a laundry product, the corresponding points will be added to that laundry product. Behavioral factors are set based on the points system for laundry products to prevent users from accumulating too many points due to multiple clicks on laundry products. The user's historical behavior data is organized by setting the time interval threshold T and the effective range of various behavioral data within the time interval threshold; Within a set time interval threshold T, when the various types of behavioral data in the statistical user historical behavior data are within the valid range, the behavior factor is set to 1; When the various types of behavioral data in the statistical user historical behavior data exceed the valid range, the behavior factor is set to 0.5; The collected historical behavior data is summarized and organized by multiplying by the corresponding behavior factor to obtain the processed historical behavior data. The collected historical order data is traversed to locate any missing historical order data. Missing historical order data is deleted and filtered, and then the processed historical order data is obtained by summarizing.

[0007] Preferably, the step of training the processed historical order data and historical behavior data using an AI large model, and generating a user profile based on the training results and basic user information, includes the following steps: The processed historical behavior data and historical order data are aggregated and combined with basic user information to initially construct a set of static user profiles; ; in, Represents a collection of static user profiles. This represents the basic information of the k-th user. This represents the historical behavior data of the k-th user after processing. This represents the historical order data processed by the k-th user.

[0008] Preferably, the preliminary identification of potential users based on the generated user profile includes the following steps: S31. Construct a decision tree based on the generated user profile; S311. Summarize the generated user profiles to construct a sample set, select a sample from the sample set Q, and extract the user's basic information, historical behavior data, and historical order data contained in the sample. S312. Set that the sample contains user basic information, historical behavior data and historical order data, which have a total of R categories. Each time, select R category samples with replacement, use the selected R category samples as the training set to train a decision tree, and use the selected R category samples as the samples at the root node of the decision tree. S313. Compare the selected R class samples and, based on the comparison results of the selected R class samples, determine whether they all belong to the same class. When all R selected classifiers belong to the same class, the decision tree is labeled with that class. When the selected R classification samples do not belong to the same category, the root node of the decision tree is divided based on the CART algorithm to generate child nodes; S314. Then recursively call steps S312 and S313 on the partitioned child nodes until all nodes are partitioned, no new nodes are generated, the recursion ends, and a decision tree is generated. Selecting the root node of a decision tree based on the CART algorithm involves the following steps: Calculate the Gini coefficients for each category of data at the current root node of the decision tree, and set a Gini coefficient threshold. When the calculated Gini coefficient is less than the threshold, return to the root node of the decision tree and stop the recursion. When the calculated Gini coefficient is greater than the threshold, the category with the smallest calculated Gini coefficient and the corresponding data value are selected as the split point to split each category sample in the root node of the decision tree; When the calculated Gini coefficient is 0, it means that there is only 1 sample in the current node, and the current node is set as a leaf node; The CART algorithm is recursively called to perform the segmentation until the segmentation result no longer changes, at which point the recursion stops. S32. Traverse each sample in the sample set Q, construct a decision tree for each sample in the sample set Q, and after construction, summarize and obtain a random forest; S33. Extract features from the generated user profile using the obtained random forest to obtain user profile features.

[0009] Preferably, the step of extracting features from the generated user profile using the obtained random forest to obtain user profile features includes the following steps: The generated user profile is input into the obtained random forest. The random forest is used to classify and extract features from the generated user profile to obtain user profile features. The obtained user profile features are reconstructed into a learnable key-value dictionary structure; ; in, This represents the attribute dimension of the first user, obtained from user profile features. This represents the quantitative characteristics of the first user in this dimension, obtained from the generated user profile.

[0010] Preferably, the step of conducting in-depth data analysis on initially identified potential users and constructing a potential user stratification system includes the following steps: Summarize the key-value dictionary structure obtained in step S3; Set a scoring threshold to filter the initially identified potential users and obtain the filtered potential users; The filtering rules are as follows: ; in, Indicates that the sample is retained. This indicates the filtered potential users. Indicates the scoring threshold; The filtered potential users are input into an SVM, which performs feature mapping and outputs decision scores. The SVM output decision function score is used as a new feature and concatenated with the key-value dictionary structure to construct an enhanced feature vector. Input the constructed enhanced feature vector Regression output reveals potential users after in-depth analysis; Set a segmentation threshold and divide the potential users after in-depth mining, and summarize the resulting potential user stratification system.

[0011] Preferably, the step of setting marketing and lead generation strategies based on the constructed potential user segmentation system and collecting marketing and lead generation strategy feedback data in real time includes the following steps: Marketing segmentation and lead generation decisions; The marketing tiered lead generation decision-making model is shown below: ; in, This indicates a marketing tiered lead generation decision-making model. Indicates the number of user tiers. Indicates the first Marketing and lead generation time for tiered users Indicates the first Marketing and customer acquisition costs for tiered user segments; Real-time collection of marketing and lead generation strategy feedback data to calculate the conversion rate of marketing tiered lead generation decisions; Record changes in user behavior and order data after the implementation of marketing and traffic acquisition strategies, and calculate the conversion rate of marketing tiered traffic acquisition decisions. The formula for calculating the conversion rate of marketing tiered lead generation decisions is as follows: ; in, Indicates the first Decision conversion rate after implementing a segmented marketing and lead generation strategy for tiered users. Indicates the first Changes in user behavior and order data after the implementation of a segmented marketing and customer acquisition strategy for different user segments. This represents the average change in user behavior data and order data.

[0012] Preferably, the process of constructing a marketing lead generation model based on real-time collected marketing lead generation strategy feedback data, and optimizing and executing the constructed marketing lead generation model in real time through model evaluation, includes the following steps: S61. Construct a marketing lead generation model based on real-time collected marketing lead generation strategy feedback data; The marketing lead generation model is constructed by combining the marketing tiered lead generation decision conversion rate and the marketing tiered lead generation decision model. S62. Optimize and implement the marketing lead generation model in real time through model evaluation; The accuracy of the marketing lead generation model prediction under the current marketing lead generation strategy is calculated based on real-time collected behavioral data and order data. ; ; Wherein, TP represents the number of identifications that match the actual results; FP represents the number of identifications that differ from the actual results; and FN represents the number of missed detections. For accuracy, Recall rate; The relationship curve between precision and recall is set as follows: - The curve; the area enclosed by the curve and the horizontal axis represents the average accuracy. The calculation formula is: ; in, The average accuracy of water resource image data; Set a threshold for the average prediction accuracy. When the calculated average prediction accuracy is lower than the set threshold, adjust the marketing and lead generation strategy.

[0013] The present invention also provides an online laundry potential customer segmentation marketing and lead generation system, which is used to implement an online laundry potential customer segmentation marketing and lead generation method. The system includes: a data acquisition module, a data processing module, a user profile generation module, a potential user identification module, a user segmentation system construction module, a marketing and lead generation module, and a marketing and lead generation adjustment module. The data acquisition module is used to collect basic information, order data, and behavioral data of all users within the network. The data processing module is used to process the collected order data and behavior data to obtain processed order data and behavior data. The user profile generation module is used to generate user profiles based on the user's basic information, processed order data, and behavioral data. The potential user identification module is used to initially identify potential users based on the generated user profile. The user segmentation system construction module is used to conduct in-depth analysis of initially identified potential users through data analysis and to construct a potential user segmentation system. The marketing lead generation module is used to set up marketing lead generation strategies and collect feedback data on marketing lead generation strategies in real time to build a marketing lead generation model; The marketing lead generation adjustment module is used to optimize and execute the marketing lead generation model in real time through model evaluation.

[0014] The beneficial effects of this invention are as follows: This invention collects basic information, order data, and behavioral data of all users within the network, processes the collected data, and then trains the processed data using an AI model. Based on the training results and basic user information, user profiles are generated. Potential users are initially identified based on these user profiles. Further data analysis is used to deeply mine these potential users and construct a potential user segmentation system. Marketing and lead generation strategies are then set based on this system, and feedback data from these strategies is collected in real time. Finally, a marketing and lead generation model is built based on the real-time feedback data, and the model is evaluated and optimized in real time, improving the accuracy and real-time performance of segmented marketing and lead generation for potential customers. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the online laundry potential customer segmentation marketing and lead generation method of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] In a specific embodiment of the present invention, Reference Figure 1 As shown, this invention provides a tiered marketing and lead generation method for potential online laundry customers, comprising the following steps: S1. Collect basic information, historical order data, and historical behavior data of all users in the network, and process the collected historical order data and historical behavior data through data processing methods to obtain processed historical order data and historical behavior data; S2. Train the processed historical order data and historical behavior data using a large AI model, and generate user profiles based on the training results and basic user information; S3. Based on the generated user profile, potential users are initially identified; S4. Conduct in-depth analysis of the initially identified potential users through data analysis and construct a potential user segmentation system; S5. Set up marketing and lead generation strategies based on the constructed potential user segmentation system, and collect marketing and lead generation strategy feedback data in real time; S6. Build a marketing lead generation model based on real-time collected marketing lead generation strategy feedback data, and optimize and execute the built marketing lead generation model in real time through model evaluation. Furthermore, referring to Figure 1 As shown, the process involves collecting basic information, historical order data, and historical behavior data from all users within the network. The collected historical order data and historical behavior data are then processed using data processing methods to obtain the processed historical order data and historical behavior data. This process includes the following steps: S11. Set the collection period and collect basic information of users on the website based on the set collection period; Set the basic information as the user's unique identification information; The basic information includes: age information, device number information, and gender information; S12. Collect historical behavior data and historical order data of the corresponding user based on the user's basic information; The historical order data includes: laundry order time, laundry order quantity, and laundry order amount; The historical behavioral data includes: search behavior data, collection behavior data, sharing behavior data, and payment behavior data; The collected historical behavior data is aggregated and a dataset is constructed based on the historical behavior data type. ; in, The table contains a collection of historical behavioral data. This represents payment behavior data from historical behavioral data. This represents search behavior data from historical behavioral data. This represents the collection behavior data in historical behavior data. This represents sharing behavior data within historical behavioral data; S13. The collected historical behavior data and historical order data are processed through data processing methods to obtain processed historical behavior data and historical order data; Set the weight of behavioral data to search < favorite < share < pay, and set the score value of the four behaviors to [1,2,3,4] based on the weight of behavioral data. When a user performs any of the above four behaviors on a laundry product, the corresponding points will be added to that laundry product. Furthermore, behavioral factors are set based on the points system for laundry products to prevent laundry products from accumulating excessive points due to multiple clicks by users; The user's historical behavior data is organized by setting the time interval threshold T and the effective range of various behavioral data within the time interval threshold; Within a set time interval threshold T, when the various types of behavioral data in the statistical user historical behavior data are within the valid range, the behavior factor is set to 1; When the various types of behavioral data in the statistical user historical behavior data exceed the valid range, the behavior factor is set to 0.5; Furthermore, the collected historical behavior data is summarized and organized by multiplying the corresponding behavior factors to obtain the processed historical behavior data; Furthermore, the collected historical order data is traversed to locate any missing historical order data. Furthermore, the missing historical order data is deleted and filtered, and then the processed historical order data is obtained by summarizing. Furthermore, referring to Figure 1 The process involves training a large AI model on processed historical order and behavior data, and generating user profiles based on the training results and basic user information. This includes the following steps: The processed historical behavior data and historical order data are aggregated and combined with basic user information to initially construct a set of static user profiles; ; in, Represents a collection of static user profiles. This represents the basic information of the k-th user. This represents the historical behavior data of the k-th user after processing. This represents the historical order data processed by the k-th user; Furthermore, referring to Figure 1 The initial identification of potential users based on the generated user profiles includes the following steps: S31. Construct a decision tree based on the generated user profile; S311. Summarize the generated user profiles to construct a sample set, select a sample from the sample set Q, and extract the user's basic information, historical behavior data, and historical order data contained in the sample. S312. Set that the sample contains user basic information, historical behavior data and historical order data, which have a total of R categories. Each time, select R category samples with replacement, use the selected R category samples as the training set to train a decision tree, and use the selected R category samples as the samples at the root node of the decision tree. S313. Compare the selected R class samples and, based on the comparison results of the selected R class samples, determine whether they all belong to the same class. When all R selected classifiers belong to the same class, the decision tree is labeled with that class. When the selected R classification samples do not belong to the same category, the root node of the decision tree is divided based on the CART algorithm to generate child nodes; S314. Then recursively call steps S312 and S313 on the partitioned child nodes until all nodes are partitioned, no new nodes are generated, the recursion ends, and a decision tree is generated. Furthermore, the selection of the root node of the decision tree based on the CART algorithm includes the following steps: Calculate the Gini coefficients for each category of data at the current root node of the decision tree, and set a Gini coefficient threshold. When the calculated Gini coefficient is less than the threshold, return to the root node of the decision tree and stop the recursion. When the calculated Gini coefficient is greater than the threshold, the category with the smallest calculated Gini coefficient and the corresponding data value are selected as the split point to split each category sample in the root node of the decision tree; When the calculated Gini coefficient is 0, it means that there is only 1 sample in the current node, and the current node is set as a leaf node; Furthermore, the CART algorithm is recursively called for segmentation until the segmentation result no longer changes, at which point the recursion stops. S32. Traverse each sample in the sample set Q, construct a decision tree for each sample in the sample set Q, and after construction, summarize and obtain a random forest; S33. Extract features from the generated user profile using the obtained random forest to obtain user profile features; The generated user profile is input into the obtained random forest. The random forest is used to classify and extract features from the generated user profile to obtain user profile features. Furthermore, the obtained user profile features are reconstructed into a learnable key-value dictionary structure; ; in, This represents the attribute dimension of the first user, obtained from user profile features. This represents the quantitative characteristics of the first user in this dimension, obtained from the generated user profile. Furthermore, referring to Figure 1 The process of conducting in-depth analysis of initially identified potential users through data analysis and constructing a potential user segmentation system includes the following steps: Summarize the key-value dictionary structure obtained in step S3; Furthermore, a scoring threshold is set to filter the initially identified potential users, resulting in a filtered pool of potential users. The filtering rules are as follows: ; in, Indicates that the sample is retained. This indicates the filtered potential users. Indicates the scoring threshold; Furthermore, the filtered potential users are input into an SVM, which performs feature mapping and outputs decision scores. ; in, For Gaussian kernel function, Represents the Lagrange multiplier. Indicates the bias term. Let the true label corresponding to the i-th support vector be defined. For potential users, For non-potential users, Represents the SVM decision function. This represents the score of the SVM output decision function; Furthermore, the SVM output decision function score is used as a new feature and concatenated with the key-value dictionary structure to construct an enhanced feature vector; Furthermore, the constructed enhanced feature vector is input Regression output reveals potential users after in-depth analysis; ; in, This represents the sigmoid function. express Regression weights express The bias of regression, express Regression decision function, Represents the enhanced feature vector. This indicates potential users after in-depth analysis; Furthermore, a segmentation threshold is set and potential users after in-depth mining are divided, and the resulting potential user stratification system is summarized. Furthermore, referring to Figure 1 The steps involved in setting up marketing and lead generation strategies based on the constructed potential user segmentation system and collecting real-time feedback data on these strategies are as follows: Marketing segmentation and lead generation decisions; The marketing tiered lead generation decision-making model is shown below: ; in, This indicates a marketing tiered lead generation decision-making model. Indicates the number of user tiers. Indicates the first Marketing and lead generation time for tiered users Indicates the first Marketing and customer acquisition costs for tiered user segments; Furthermore, real-time data collection of marketing and lead generation strategy feedback is used to calculate the conversion rate of marketing tiered lead generation decisions; Record changes in user behavior and order data after the implementation of marketing and traffic acquisition strategies, and calculate the conversion rate of marketing tiered traffic acquisition decisions. The formula for calculating the conversion rate of marketing tiered lead generation decisions is as follows: ; in, Indicates the first Decision conversion rate after implementing a segmented marketing and lead generation strategy for tiered users. Indicates the first Changes in user behavior and order data after the implementation of a segmented marketing and customer acquisition strategy for different user segments. This represents the average change in customer behavior data and order data. Furthermore, referring to Figure 1 The marketing lead generation model is built based on real-time collected feedback data of marketing lead generation strategies. The model is then evaluated and optimized in real-time. The process includes the following steps: S61. Construct a marketing lead generation model based on real-time collected marketing lead generation strategy feedback data; The marketing lead generation model is constructed by combining the marketing tiered lead generation decision conversion rate and the marketing tiered lead generation decision model. S62. Optimize and implement the marketing lead generation model in real time through model evaluation; The accuracy of the marketing lead generation model prediction under the current marketing lead generation strategy is calculated based on real-time collected behavioral data and order data. ; ; Wherein, TP represents the number of identifications that match the actual results; FP represents the number of identifications that differ from the actual results; and FN represents the number of missed detections. For accuracy, Recall rate; The relationship curve between precision and recall is set as follows: - The curve; the area enclosed by the curve and the horizontal axis represents the average accuracy. The calculation formula is: ; in, The average accuracy of water resource image data; Set a threshold for the average prediction accuracy. When the calculated average prediction accuracy is lower than the set threshold, adjust the marketing and lead generation strategy. In one specific embodiment, the online laundry potential customer segmented marketing and lead generation system is used to implement an online laundry potential customer segmented marketing and lead generation method. The system includes: a data acquisition module, a data processing module, a user profile generation module, a potential user identification module, a user segmentation system construction module, a marketing and lead generation module, and a marketing and lead generation adjustment module. The data acquisition module is used to collect basic information, order data, and behavioral data of all users within the network. The data processing module is used to process the collected order data and behavior data to obtain processed order data and behavior data. The user profile generation module is used to generate user profiles based on the user's basic information, processed order data, and behavioral data. The potential user identification module is used to initially identify potential users based on the generated user profile. The user segmentation system construction module is used to conduct in-depth analysis of initially identified potential users through data analysis and to construct a potential user segmentation system. The marketing lead generation module is used to set up marketing lead generation strategies and collect feedback data on marketing lead generation strategies in real time to build a marketing lead generation model; The marketing lead generation adjustment module is used to optimize and execute the marketing lead generation model in real time through model evaluation.

[0019] It should be noted that the above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A segmented marketing and lead generation method for potential online laundry customers, characterized by: Includes the following steps: S1. Collect basic information, historical order data, and historical behavior data of all users in the network, and process the collected historical order data and historical behavior data through data processing methods to obtain processed historical order data and historical behavior data; S2. Train the processed historical order data and historical behavior data using a large AI model, and generate user profiles based on the training results and basic user information; S3. Based on the generated user profile, potential users are initially identified; S4. Conduct in-depth analysis of the initially identified potential users through data analysis and construct a potential user segmentation system; S5. Set up marketing and lead generation strategies based on the constructed potential user segmentation system, and collect marketing and lead generation strategy feedback data in real time; S6. Build a marketing lead generation model based on real-time collected marketing lead generation strategy feedback data, and optimize and execute the built marketing lead generation model in real time through model evaluation.

2. The online laundry potential customer segmentation marketing and lead generation method according to claim 1, characterized in that, The process of collecting basic information, historical order data, and historical behavior data of all users within the network, and then processing the collected historical order data and historical behavior data to obtain the processed historical order data and historical behavior data includes the following steps: S11. Set the collection period and collect basic information of users on the website based on the set collection period; Set the basic information as the user's unique identification information; The basic information includes: age information, device number information, and gender information; S12. Collect historical behavior data and historical order data of the corresponding user based on the user's basic information; The historical order data includes: laundry order time, laundry order quantity, and laundry order amount; The historical behavioral data includes: search behavior data, collection behavior data, sharing behavior data, and payment behavior data; The collected historical behavior data is aggregated and a dataset is constructed based on the historical behavior data type. ; in, The table contains a collection of historical behavioral data. This represents payment behavior data from historical behavioral data. This represents search behavior data from historical behavioral data. This represents the collection behavior data in historical behavior data. This represents sharing behavior data within historical behavioral data; S13. The collected historical behavior data and historical order data are processed through data processing methods to obtain processed historical behavior data and historical order data.

3. The online laundry potential customer segmentation marketing and lead generation method according to claim 2, characterized in that, The process of processing the collected historical behavior data and historical order data to obtain the processed historical behavior data and historical order data includes the following steps: Set the weight of behavioral data to search < favorite < share < pay, and set the score value of the four behaviors to [1,2,3,4] based on the weight of behavioral data. When a user performs any of the above four behaviors on a laundry product, the corresponding points will be added to that laundry product. Behavioral factors are set based on the points system for laundry products to prevent users from accumulating too many points due to multiple clicks on laundry products. The user's historical behavior data is organized by setting the time interval threshold T and the effective range of various behavioral data within the time interval threshold; Within a set time interval threshold T, when the various types of behavioral data in the statistical user historical behavior data are within the valid range, the behavior factor is set to 1; When the various types of behavioral data in the statistical user historical behavior data exceed the valid range, the behavior factor is set to 0.5; The collected historical behavior data is summarized and organized by multiplying by the corresponding behavior factor to obtain the processed historical behavior data. The collected historical order data is traversed to locate any missing historical order data. Missing historical order data is deleted and filtered, and then the processed historical order data is obtained by summarizing.

4. The online laundry potential customer segmentation marketing and lead generation method according to claim 1, characterized in that, The process of training a large AI model on processed historical order data and historical behavior data, and generating user profiles based on the training results and basic user information, includes the following steps: The processed historical behavior data and historical order data are aggregated and combined with basic user information to initially construct a set of static user profiles; ； in, Represents a collection of static user profiles. This represents the basic information of the k-th user. This represents the historical behavior data of the k-th user after processing. This represents the historical order data processed by the k-th user.

5. The online laundry potential customer segmentation marketing and lead generation method according to claim 1, characterized in that, The process of initially identifying potential users based on the generated user profile includes the following steps: S31. Construct a decision tree based on the generated user profile; S311. Summarize the generated user profiles to construct a sample set, select a sample from the sample set Q, and extract the user's basic information, historical behavior data, and historical order data contained in the sample. S312. Set that the sample contains user basic information, historical behavior data and historical order data, which have a total of R categories. Each time, select R category samples with replacement, use the selected R category samples as the training set to train a decision tree, and use the selected R category samples as the samples at the root node of the decision tree. S313. Compare the selected R class samples and, based on the comparison results of the selected R class samples, determine whether they all belong to the same class. When all R selected classifiers belong to the same class, the decision tree is labeled with that class. When the selected R classification samples do not belong to the same category, the root node of the decision tree is divided based on the CART algorithm to generate child nodes; S314. Then recursively call steps S312 and S313 on the partitioned child nodes until all nodes are partitioned, no new nodes are generated, the recursion ends, and a decision tree is generated. Selecting the root node of a decision tree based on the CART algorithm involves the following steps: Calculate the Gini coefficients for each category of data at the current root node of the decision tree, and set a Gini coefficient threshold. When the calculated Gini coefficient is less than the threshold, return to the root node of the decision tree and stop the recursion. When the calculated Gini coefficient is greater than the threshold, the category with the smallest calculated Gini coefficient and the corresponding data value are selected as the split point to split each category sample in the root node of the decision tree; When the calculated Gini coefficient is 0, it means that there is only 1 sample in the current node, and the current node is set as a leaf node; The CART algorithm is recursively called to perform the segmentation until the segmentation result no longer changes, at which point the recursion stops. S32. Traverse each sample in the sample set Q, construct a decision tree for each sample in the sample set Q, and after construction, summarize and obtain a random forest; S33. Extract features from the generated user profile using the obtained random forest to obtain user profile features.

6. The online laundry potential customer segmentation marketing and lead generation method according to claim 5, characterized in that, The step of extracting features from the generated user profile using the obtained random forest to obtain user profile features includes the following steps: The generated user profile is input into the obtained random forest. The random forest is used to classify and extract features from the generated user profile to obtain user profile features. The obtained user profile features are reconstructed into a learnable key-value dictionary structure; ； in, This represents the attribute dimension of the first user, obtained from user profile features. This represents the quantitative characteristics of the first user in this dimension, obtained from the generated user profile.

7. The online laundry potential customer segmentation marketing and lead generation method according to claim 1, characterized in that, The process of conducting in-depth analysis of initially identified potential users and constructing a potential user segmentation system through data analysis includes the following steps: Summarize the key-value dictionary structure obtained in step S3; Set a scoring threshold to filter the initially identified potential users and obtain the filtered potential users; The filtering rules are as follows: ； in, Indicates that the sample is retained. This indicates the filtered potential users. Indicates the scoring threshold; The filtered potential users are input into an SVM, which performs feature mapping and outputs decision scores. The SVM output decision function score is used as a new feature and concatenated with the key-value dictionary structure to construct an enhanced feature vector. Input the constructed enhanced feature vector Regression output reveals potential users after in-depth analysis; Set a segmentation threshold and divide the potential users after in-depth mining, and summarize the resulting potential user stratification system.

8. The online laundry potential customer segmentation marketing and lead generation method according to claim 1, characterized in that, The process of setting up marketing and lead generation strategies based on the constructed potential user segmentation system and collecting real-time feedback data on these strategies includes the following steps: Marketing segmentation and lead generation decisions; The marketing tiered lead generation decision-making model is shown below: ； in, This indicates a marketing tiered lead generation decision-making model. Indicates the number of user tiers. Indicates the first Marketing and lead generation time for tiered users Indicates the first Marketing and customer acquisition costs for tiered user segments; Real-time collection of marketing and lead generation strategy feedback data to calculate the conversion rate of marketing tiered lead generation decisions; Record changes in user behavior and order data after the implementation of marketing and traffic acquisition strategies, and calculate the conversion rate of marketing tiered traffic acquisition decisions. The formula for calculating the conversion rate of marketing tiered lead generation decisions is as follows: ； in, Indicates the first Decision conversion rate after implementing a segmented marketing and lead generation strategy for tiered users. Indicates the first Changes in user behavior and order data after the implementation of a segmented marketing and customer acquisition strategy for different user segments. This represents the average change in user behavior data and order data.

9. The online laundry potential customer segmentation marketing and lead generation method according to claim 1, characterized in that, The process of constructing a marketing lead generation model based on real-time collected feedback data from marketing lead generation strategies, and then optimizing and implementing the model in real-time through model evaluation, includes the following steps: S61. Construct a marketing lead generation model based on real-time collected marketing lead generation strategy feedback data; The marketing lead generation model is constructed by combining the marketing tiered lead generation decision conversion rate and the marketing tiered lead generation decision model. S62. Optimize and implement the marketing lead generation model in real time through model evaluation; The accuracy of the marketing lead generation model prediction under the current marketing lead generation strategy is calculated based on real-time collected behavioral data and order data. ； ； Wherein, TP represents the number of identifications that match the actual results; FP represents the number of identifications that differ from the actual results; and FN represents the number of missed detections. For accuracy, Recall rate; The relationship curve between precision and recall is set as follows: - The curve; the area enclosed by the curve and the horizontal axis represents the average accuracy. The calculation formula is: ； in, The average accuracy of water resource image data; Set a threshold for the average prediction accuracy. When the calculated average prediction accuracy is lower than the set threshold, adjust the marketing and lead generation strategy.

10. A system for implementing the online laundry potential customer segmentation marketing and lead generation method as described in claim 1, characterized in that, The system includes: a data acquisition module, a data processing module, a user profile generation module, a potential user identification module, a user segmentation system construction module, a marketing and traffic generation module, and a marketing and traffic generation adjustment module. The data acquisition module is used to collect basic information, order data, and behavioral data of all users within the network. The data processing module is used to process the collected order data and behavior data to obtain processed order data and behavior data. The user profile generation module is used to generate user profiles based on the user's basic information, processed order data, and behavioral data. The potential user identification module is used to initially identify potential users based on the generated user profile. The user segmentation system construction module is used to conduct in-depth analysis of initially identified potential users through data analysis and to construct a potential user segmentation system. The marketing lead generation module is used to set up marketing lead generation strategies and collect feedback data on marketing lead generation strategies in real time to build a marketing lead generation model; The marketing lead generation adjustment module is used to optimize and execute the marketing lead generation model in real time through model evaluation.

Images