Precision marketing method and system based on real-time data flow

Abstract

The invention relates to a precision marketing method and system based on real-time data flow. The method comprises the following steps: collecting customer behavior data in real time; preprocessing and dynamically marking the customer behavior data to generate preprocessed customer behavior data; performing real-time analysis on the preprocessed customer behavior data by using a deep learning algorithm, and generating a customer portrait in combination with historical behaviors, preference changes and situational features of the customer; according to the customer portrait and the customer real-time behavior, a marketing strategy is generated in real time through an intelligent decision routing technology, and recommendation content in the marketing strategy is dynamically adjusted based on the customer interaction behavior in the customer real-time behavior; according to the marketing strategy, the service recommendation corresponding to the marketing strategy is sent to the customer contact point, and through a real-time feedback mechanism, the marketing strategy is dynamically optimized according to the real-time response of the customer. The method has the effect of improving the marketing efficiency.

Description

Technical Field

[0001] The present application relates to the technical field of information and communication technology, and in particular to a precision marketing method and system based on real-time data stream. Background Art

[0002] At present, with the rapid development of information technology, the amount of data faced by enterprises in marketing activities is growing exponentially, and traditional marketing methods are difficult to meet the needs of precision and personalization. In order to improve customer conversion rate and satisfaction, enterprises are increasingly relying on real-time data flow and big data analysis technology, especially in the business field, through real-time collection and analysis of customer behavior data, quickly identify changes in customer needs and preferences, and thus formulate personalized marketing strategies.

[0003] Most existing precision marketing technologies rely on static data and preset rules, often lacking dynamic responses to real-time customer behavior, and making it difficult to achieve automatic adjustment and real-time optimization of marketing strategies. Existing systems often have problems such as poor timeliness and lack of flexibility in data processing, analysis, and strategy generation, making it difficult to cope with the challenges of rapidly changing customer behavior.

[0004] The above-mentioned existing technical solutions have the following defects: customer portrait updates and marketing strategy adjustments are not flexible enough to cope with complex customer behavior patterns and rapid changes in preferences, so there is room for improvement. Summary of the invention

[0005] In order to improve the efficiency of marketing, the present application provides a precision marketing method and system based on real-time data stream.

[0006] The above-mentioned invention objective of the present application is achieved through the following technical solutions: A precision marketing method based on real-time data stream, the precision marketing method based on real-time data stream comprising: Collect customer behavior data in real time; Preprocessing and dynamically marking the customer behavior data to generate preprocessed customer behavior data; Use a deep learning algorithm to perform real-time analysis on the pre-processed customer behavior data, and generate a customer profile by combining the customer's historical behavior, preference changes, and situational characteristics; Generate a marketing strategy in real time through intelligent decision routing technology according to the customer portrait and the customer's real-time behavior, and dynamically adjust the recommended content in the marketing strategy based on the customer's interactive behavior in the customer's real-time behavior; According to the marketing strategy, a service recommendation corresponding to the marketing strategy is sent to the customer contact point, and through a real-time feedback mechanism, the marketing strategy is dynamically optimized according to the real-time response of the customer.

[0007] By adopting the above technical solutions, by acquiring customer behavior data from different channels in real time, it is possible to fully capture customers' dynamic behaviors and demand changes, thereby achieving instant perception and response to customer activities, ensuring the timeliness of data and improving the pertinence and accuracy of marketing strategies; by pre-processing customer behavior data, inaccurate and irrelevant information in the data is eliminated, thereby improving data quality; and dynamically marking customer behavior data helps identify customers' interests, demand changes and potential behaviors, providing more accurate input for subsequent analysis; by performing real-time analysis of customer behavior data through deep learning algorithms, it is possible to deeply explore customers' potential needs, preferences and behavior patterns, and then generate accurate customer portraits, combining historical behaviors, preference changes and situational characteristics to ensure the comprehensiveness and real-time nature of the portraits, which helps predict customers' future behaviors and needs; through intelligent decision routing technology, personalized marketing strategies are generated based on customer portraits and real-time behaviors, which can respond to changes in customer needs in real time; at the same time, by dynamically adjusting recommended content, it is possible to ensure that marketing strategies can be quickly optimized based on customer interactive behaviors, maximizing customer experience and conversion rates; by sending service recommendations corresponding to marketing strategies to customer contact points, it is possible to directly reach customers and dynamically optimize based on their feedback, thereby improving marketing effectiveness and response speed.

[0008] In one example, the present application may be further configured as follows: the real-time collection of customer behavior data includes: Use a multi-source data acquisition module to obtain customer action data from different channels in real time, and integrate the action data into the customer behavior data. The action data includes website click streams, APP usage records, browsing history interaction data, and transaction data.

[0009] By adopting the above technical solution, through the multi-source data collection module, it is possible to obtain real-time customer behavior data from multiple channels at the same time, so that the system can comprehensively capture customer activities on different platforms, and provide richer and more comprehensive basic data for subsequent analysis; by integrating customer behavior data from different channels, the system can summarize and uniformly analyze customer activities at various contact points, which helps to eliminate data silos and enable customer behavior to be comprehensively and uniformly displayed in a cross-channel and cross-platform context; by collecting different types of behavior data such as website click streams, APP usage records, browsing history interaction data and transaction data, the system can obtain detailed customer behavior trajectories in multiple dimensions, which helps to comprehensively analyze customer needs and formulate accurate marketing strategies.

[0010] In one example, the present application may be further configured as follows: the preprocessing and dynamic marking of the customer behavior data includes: Cleaning the customer behavior data to remove noise and outliers; Performing time series analysis on the customer behavior data to identify customer behavior patterns and trends; Normalizing the customer behavior data; Marking the customer behavior data to identify the customer's interests and potential needs; The customer behavior data is classified to identify different types of customers and assign corresponding type tags to different types of customers.

[0011] By adopting the above technical solutions, by cleaning the customer behavior data, the noise and outliers in the data can be removed, thereby improving the quality and accuracy of the data; through time series analysis, the patterns and trends of customer behavior changes over time can be identified, thereby providing a basis for accurately predicting future customer behavior and formulating dynamic marketing strategies; by normalizing the customer behavior data, data of different dimensions or scales are converted into a unified standard range, ensuring that each feature participates in the analysis with equal weight, which helps to avoid certain features from having unfair effects on the analysis results due to differences in data magnitude, thereby improving the accuracy of the analysis results; by marking the customer behavior data, the customer's points of interest and potential needs can be accurately identified, so as to help the system better understand the customer's preferences and needs, and provide an accurate basis for the formulation of personalized marketing strategies; by classifying the customer behavior data, different types of customers can be identified, which helps to carry out refined management and personalized services for customers, thereby realizing more targeted marketing and service strategies.

[0012] In one example, the present application may be further configured as follows: the use of a deep learning algorithm to perform real-time analysis on the pre-processed customer behavior data, and combining the customer's historical behavior, preference changes, and situational characteristics to generate a customer profile includes: Using a deep neural network to extract features from the pre-processed customer behavior data to identify key feature data; Based on the historical behaviors, recurrent neural networks, and long short-term memory networks, a behavior prediction model is constructed; The key feature data is transmitted to the behavior prediction model, and the time series information in the pre-processed customer behavior data is modeled in combination with the preference changes and the situational features, the evolution process of customer preferences is analyzed, and the customer portrait is generated.

[0013] By adopting the above technical solution, the customer behavior data is extracted through deep neural networks, which can automatically identify the most representative and influential key features from a large amount of data, and can efficiently screen out data features that are crucial to the construction of customer portraits, reduce redundant information, and improve the calculation efficiency and accuracy of the model; by combining historical behavior data, recurrent neural networks and long short-term memory networks, a behavior prediction model that can process time series data can be constructed. The model can capture the temporal dependencies of customer behavior, identify long-term and short-term behavior patterns, and then accurately predict the customer's future behavior; by combining the extracted key feature data with the behavior prediction model, the model can not only analyze the temporal nature of the customer's historical behavior, but also consider the changes in customer preferences and situational characteristics. The modeling process can reveal the evolution of customer preferences, thereby generating a dynamic and accurate customer portrait that reflects the customer's latest needs and interest changes, and provides a basis for personalized marketing.

[0014] In one example, the present application may be further configured as follows: the real-time generation of marketing strategies through intelligent decision routing technology according to the customer portrait and the customer's real-time behavior includes: Based on the customer portrait, a decision tree algorithm is used to dynamically evaluate the customer's current and potential needs and generate a marketing decision path; Determine a marketing action based on the customer's real-time behavior and the marketing decision path; The marketing strategy is generated by combining the marketing decision path, the marketing action and the preset marketing rules.

[0015] By adopting the above technical solution, by using the decision tree algorithm, the current and potential needs of customers can be dynamically evaluated according to the information in the customer portrait, which can clarify the direction of customer needs, help identify the marketing method that best suits the customer, and improve marketing effectiveness; by obtaining customer behavior data in real time and combining it with the aforementioned marketing decision path, the system can flexibly determine the marketing actions suitable for the current situation, ensuring the immediate responsiveness of the marketing strategy, and can make dynamic adjustments based on the real-time behavior changes of customers, thereby improving the pertinence and timeliness of marketing; by combining the marketing decision path, determined marketing actions and preset marketing rules, the system can generate the final marketing strategy, ensuring that the marketing strategy can be personalized based on customer needs, while also complying with preset rules and business goals, thereby improving the efficiency and effectiveness of marketing activities.

[0016] In one example, the present application may be further configured as follows: dynamically adjusting the recommended content in the marketing strategy based on the customer interaction behavior in the customer's real-time behavior includes: Analyze the customer's interactive behaviors to obtain the customer's immediate needs and interest changes, and the customer's interactive behaviors include clicking, browsing, collecting, sharing and purchasing; According to the user's feedback on the recommended content, evaluating the effectiveness of the recommended content and generating a feedback result; Based on the feedback result, an adaptive learning algorithm is used to adjust the recommendation scheme, generate a recommendation adjustment strategy, and adjust the recommended content according to the recommendation adjustment strategy.

[0017] By adopting the above technical solutions, by analyzing the customer's interactive behavior, identifying the changes in customer needs and interest trends, and by accurately tracking the changes in customer behavior, it is possible to capture the customer's immediate needs at different time points, and help the marketing system quickly adjust strategies to better meet customer needs; by collecting and analyzing user feedback on recommended content, the system can evaluate the effectiveness of recommended content and generate specific feedback results, ensuring that the recommendation system can judge whether the recommendation meets its needs based on the user's response, thereby optimizing subsequent recommended content and improving the accuracy of recommendations and user satisfaction; by using adaptive learning algorithms to automatically adjust recommendation plans based on feedback results, the system can dynamically optimize recommended content to ensure continuous optimization of marketing strategies and accurate matching of customers' personalized needs.

[0018] In one example, the present application may be further configured as follows: the precision marketing method based on real-time data stream further includes: Acquire customer behavior data at each customer contact point in real time, monitor customer immediate response to the recommended content, and analyze changes in customer preferences to generate demand change data; According to the demand change data and real-time customer feedback, the display order, promotion channels and marketing methods of the recommended content are adjusted.

[0019] By adopting the above technical solutions, by acquiring the customer's behavioral data at each contact point in real time and monitoring their immediate response to the recommended content, the system can dynamically track changes in customer preferences, promptly identify fluctuations in customer demand, and help the system quickly capture changes in customer demand and generate corresponding demand change data; by analyzing demand change data and real-time feedback from customers, the system can accurately adjust the display order, promotion channels and marketing methods of recommended content, ensuring that marketing activities can respond in a timely manner according to changes in customer demand, and push content through the optimal channels and methods at the right time, thereby improving the recommendation effect and customer interactive experience.

[0020] The second object of the invention is achieved by the following technical solutions: A precision marketing system based on real-time data stream, the precision marketing system based on real-time data stream comprising: Data collection module, used to collect customer behavior data in real time; A preprocessing module, used to preprocess and dynamically mark the customer behavior data to generate preprocessed customer behavior data; An analysis module is used to use a deep learning algorithm to perform real-time analysis on the pre-processed customer behavior data, and generate a customer profile by combining the customer's historical behavior, preference changes, and situational characteristics; A strategy generation module, for generating a marketing strategy in real time based on the customer profile and the customer's real-time behavior through intelligent decision routing technology, and dynamically adjusting the recommended content in the marketing strategy based on the customer's interactive behavior in the customer's real-time behavior; The optimization module is used to send service recommendations corresponding to the marketing strategy to customer contact points according to the marketing strategy, and dynamically optimize the marketing strategy according to the real-time response of the customer through a real-time feedback mechanism.

[0021] By adopting the above technical solutions, by acquiring customer behavior data from different channels in real time, it is possible to fully capture customers' dynamic behaviors and demand changes, thereby achieving instant perception and response to customer activities, ensuring the timeliness of data and improving the pertinence and accuracy of marketing strategies; by pre-processing customer behavior data, inaccurate and irrelevant information in the data is eliminated, thereby improving data quality; and dynamically marking customer behavior data helps identify customers' interests, demand changes and potential behaviors, providing more accurate input for subsequent analysis; by performing real-time analysis of customer behavior data through deep learning algorithms, it is possible to deeply explore customers' potential needs, preferences and behavior patterns, and then generate accurate customer portraits, combining historical behaviors, preference changes and situational characteristics to ensure the comprehensiveness and real-time nature of the portraits, which helps predict customers' future behaviors and needs; through intelligent decision routing technology, personalized marketing strategies are generated based on customer portraits and real-time behaviors, which can respond to changes in customer needs in real time; at the same time, by dynamically adjusting recommended content, it is possible to ensure that marketing strategies can be quickly optimized based on customer interactive behaviors, maximizing customer experience and conversion rates; by sending service recommendations corresponding to marketing strategies to customer contact points, it is possible to directly reach customers and dynamically optimize based on their feedback, thereby improving marketing effectiveness and response speed.

[0022] In summary, this application includes the following beneficial technical effects: 1. By acquiring customer behavior data from different channels in real time, it is possible to fully capture customers' dynamic behaviors and demand changes, thereby achieving instant perception and response to customer activities, ensuring the timeliness of data and improving the pertinence and accuracy of marketing strategies; by pre-processing customer behavior data, inaccurate and irrelevant information in the data is eliminated, thereby improving data quality; and dynamically marking customer behavior data helps identify customers' interests, demand changes and potential behaviors, providing more accurate input for subsequent analysis; 2. Real-time analysis of customer behavior data through deep learning algorithms can deeply explore customers' potential needs, preferences and behavioral patterns, and then generate accurate customer portraits. Combining historical behaviors, preference changes and situational characteristics, it ensures the comprehensiveness and real-time nature of the portraits, which helps predict customers' future behaviors and needs. Through intelligent decision-making routing technology, personalized marketing strategies are generated according to customer portraits and real-time behaviors, which can respond to changes in customer needs in real time. At the same time, by dynamically adjusting recommended content, it ensures that marketing strategies can be quickly optimized according to customer interactive behaviors, maximizing customer experience and conversion rates. By sending service recommendations corresponding to marketing strategies to customer touch points, it can directly reach customers and dynamically optimize based on their feedback, thereby improving marketing effectiveness and response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of a precision marketing method based on real-time data stream in one embodiment of the present application; Figure 2 This is a flowchart for implementing step S10 in a precision marketing method based on real-time data stream in one embodiment of the present application; Figure 3 This is a flowchart for implementing step S20 in a precision marketing method based on real-time data stream in one embodiment of the present application; Figure 4 This is a flowchart for implementing step S30 in a precision marketing method based on real-time data stream in one embodiment of the present application; Figure 5 This is a flowchart for implementing step S40 in a precision marketing method based on real-time data stream in one embodiment of the present application; Figure 6 This is a flowchart for implementing step S40 in a precision marketing method based on real-time data stream in one embodiment of the present application; Figure 7 This is a flowchart of an implementation of a precision marketing method based on real-time data stream in an embodiment of the present application; Figure 8 It is a principle block diagram of a precision marketing system based on real-time data stream in one embodiment of the present application. DETAILED DESCRIPTION

[0024] The present application is further described in detail below in conjunction with the accompanying drawings.

[0025] In one embodiment, if Figure 1 As shown, the present application discloses a precision marketing method based on real-time data stream, which specifically includes the following steps: S10: Collect customer behavior data in real time.

[0026] Specifically, by deploying multiple data collection modules at different customer contact points (such as e-commerce platforms, apps, social media, etc.), customer behavior data can be obtained in real time. These data include but are not limited to website clickstream data, page browsing records in apps, interactive behaviors on social platforms (such as likes, comments, shares), transaction records (such as purchases, payments), and other interactive behaviors (such as searches, followings, messages, etc.). These data can be collected in real time through API interfaces, SDKs, or browser plug-ins, and transmitted to the data storage module in the form of data streams to ensure the timeliness and comprehensiveness of customer behavior data.

[0027] S20: Preprocess and dynamically mark the customer behavior data to generate preprocessed customer behavior data.

[0028] Specifically, customer behavior data will first be preprocessed after collection, a process that includes data cleaning, denoising, and outlier removal. For example, data with missing values ​​or duplicate records will be filled or deleted, and unreasonable behavior data that exceeds the normal range will be filtered to ensure data quality. Next, by dynamically marking the customer's behavior data, the tags can include key indicators such as the customer's interests, activity frequency, and purchase intentions. For example, the time and frequency of each time a customer browses a certain product can be marked as "high interest", and the purchase of a certain product can be marked as "conversion behavior." These tags help with subsequent data analysis and the accurate generation of customer portraits.

[0029] S30: Use deep learning algorithms to perform real-time analysis on pre-processed customer behavior data, and generate customer profiles based on the customer’s historical behavior, preference changes, and situational characteristics.

[0030] Specifically, by building deep neural networks (such as CNN, RNN, etc.), pre-processed customer behavior data can be analyzed in real time. In this process, historical behavior data will be used as input, combined with changes in customer preferences (for example, changes in customer interest in a certain type of goods or services over a period of time) and situational characteristics (such as behavior patterns during holidays and promotions) to train the model and automatically identify key factors that affect customer behavior. Through model training, dynamic and accurate customer portraits can be generated, including information such as customer consumption tendencies, preference categories, and purchase frequency. These portraits can be continuously updated and adjusted as customer behavior changes.

[0031] S40: Generate marketing strategies in real time based on customer portraits and real-time customer behaviors through intelligent decision routing technology, and dynamically adjust the recommended content in the marketing strategy based on customer interaction behaviors in real-time customer behaviors.

[0032] Specifically, by combining customer profile information and real-time behavior data (for example, customer clicks, browsing, purchases, etc. at a certain point in time), through intelligent decision-making technologies such as decision tree algorithms or reinforcement learning, customers' current and potential needs are dynamically evaluated to generate personalized marketing strategies. For example, if a customer has recently browsed a certain category of goods frequently, promotional information or new product recommendations related to the goods can be pushed. At the same time, based on the customer's interactive behavior (such as clicks, likes, comments, etc.), the recommended content can be adjusted in real time. If a customer shows a high interest in a certain recommended content (for example, a high click-through rate), the display frequency of the recommended content can be increased, otherwise the push frequency can be reduced.

[0033] S50: According to the marketing strategy, the service recommendation corresponding to the marketing strategy is sent to the customer contact point, and the marketing strategy is dynamically optimized according to the customer's real-time response through the real-time feedback mechanism.

[0034] Specifically, after the marketing strategy is generated, personalized recommendations are sent to the customer's touch points (such as mobile phones, computers, social platforms, etc.) through channels such as email, SMS push, and APP notifications. After the customer receives the recommendation content, based on their real-time response (such as clicks, purchases, shares, etc.), the system will record and analyze these feedbacks in real time to evaluate the effectiveness of the recommendation content. Through feedback data, the effect of the recommendation strategy can be calculated, such as click-through rate, conversion rate, etc. The recommendation strategy can be optimized through these data, and the recommendation content and push strategy can be dynamically adjusted. For example, if the click-through rate of a recommendation is low, the recommended product type can be adjusted or the marketing copy can be redesigned to increase user interest and responsiveness.

[0035] By adopting the above technical solutions, by acquiring customer behavior data from different channels in real time, it is possible to fully capture customers' dynamic behaviors and demand changes, thereby achieving instant perception and response to customer activities, ensuring the timeliness of data and improving the pertinence and accuracy of marketing strategies; by pre-processing customer behavior data, inaccurate and irrelevant information in the data is eliminated, thereby improving data quality; and dynamically marking customer behavior data helps identify customers' interests, demand changes and potential behaviors, providing more accurate input for subsequent analysis; by performing real-time analysis of customer behavior data through deep learning algorithms, it is possible to deeply explore customers' potential needs, preferences and behavior patterns, and then generate accurate customer portraits, combining historical behaviors, preference changes and situational characteristics to ensure the comprehensiveness and real-time nature of the portraits, which helps predict customers' future behaviors and needs; through intelligent decision routing technology, personalized marketing strategies are generated based on customer portraits and real-time behaviors, which can respond to changes in customer needs in real time; at the same time, by dynamically adjusting recommended content, it is possible to ensure that marketing strategies can be quickly optimized based on customer interactive behaviors, maximizing customer experience and conversion rates; by sending service recommendations corresponding to marketing strategies to customer contact points, it is possible to directly reach customers and dynamically optimize based on their feedback, thereby improving marketing effectiveness and response speed.

[0036] In one embodiment, if Figure 2 As shown, in step S10, customer behavior data is collected in real time, specifically including: S11: Use the multi-source data collection module to obtain customer action data from different channels in real time, and integrate the action data into customer behavior data. The action data includes website click streams, APP usage records, browsing history interaction data, and transaction data.

[0037] Specifically, using distributed data collection technology, it is first necessary to deploy a variety of data collection modules, through which customers' action data can be obtained in real time at different customer contact points. For example, website clickstream data can be tracked and recorded in real time through the JavaScript code of the web page or the browser plug-in, capturing the page elements clicked by each user, the dwell time, page jumps and other behaviors; APP usage records are collected through the SDK (software development kit) embedded in the mobile application, and these data include the time when the user opens the application, the pages visited, the functional modules browsed, the dwell time, etc.; browsing history interaction data can be obtained by analyzing the user's interactive behavior on different platforms, including the user's comments, likes, collections, sharing and other behaviors, all of which are uploaded to the data storage module in real time through the corresponding API interface; transaction data includes the user's order details, payment records, purchase frequency, etc., which can be obtained through the payment system of the e-commerce platform. These action data from different channels will be integrated in real time and converted into unified customer behavior data, and all data will be transmitted to a unified data warehouse or cloud storage through data streams to ensure the real-time and integrity of the data. In order to ensure that the integrated data can be managed and processed in a unified manner, all raw data will be standardized, including field unification and format conversion, for subsequent analysis and modeling.

[0038] In one embodiment, if Figure 3 As shown, in step S20, the customer behavior data is preprocessed and dynamically marked, specifically including: S21: Clean customer behavior data to remove noise and outliers.

[0039] Specifically, a distributed computing framework is used to clean customer behavior data to remove noise and outliers. During the data cleaning process, invalid or abnormal data is identified and removed by setting a reasonable threshold range. For example, for click stream data in customer behavior data, if a user's click behavior is abnormally frequent and exceeds the normal range, it may be an error operation or a malicious attack. Such behavior will be marked as abnormal and removed; for transaction data, data with too high or too low order amounts can be filtered by setting a reasonable transaction range to ensure the accuracy and representativeness of the data. In addition, it is necessary to identify and remove extreme values ​​by analyzing the data distribution. For example, the order time in some transaction records is 12 o'clock midnight. This type of data may not be representative due to data collection errors. These will also be deleted or corrected. The final output data set will only have normal data that truly reflects customer behavior.

[0040] S22: Perform time series analysis on customer behavior data to identify customer behavior patterns and trends.

[0041] Specifically, time series analysis is performed on customer behavior data to identify customer behavior patterns and trends. Customer behavior data will be sorted by timestamp to ensure that each record has an accurate time sequence, and each user's behavior data will be expanded along the timeline for continuous time series analysis. With the help of time series analysis methods, such as moving average or exponential smoothing, customer behavior fluctuations within a certain time range are analyzed to identify customers' periodic behavior patterns. For example, some customers shop at fixed times each week, and these behaviors can be labeled as "periodic purchases." In addition, time series analysis can also help identify long-term trends, such as the continued increase in purchase frequency of certain customers over a period of time, or the changing trend in interest in a certain type of product.

[0042] S23: Normalize customer behavior data.

[0043] Specifically, the purpose of normalization is to eliminate the influence of different dimensions and scales in the data, and to ensure that different types of customer behavior data can be compared and analyzed on the same scale. For example, different indicators such as purchase amount and browsing time can be processed through the minimum-maximum normalization method to convert each behavior data into an interval of 0 to 1. The formula is: (x-min) / (max-min), where x is the original data, min and max are the minimum and maximum values ​​of the data item, respectively. This can unify data originally in different dimensions to the same standard, avoiding the disproportionate impact of a certain type of data on the results due to excessive dimensional differences in subsequent model analysis.

[0044] S24: Mark customer behavior data to identify customers’ interests and potential needs.

[0045] Specifically, first, based on the customer's behavioral characteristics, such as browsing history, purchase history, etc., identify the products or services that the customer may be interested in. If a customer frequently browses a certain type of product but does not make a purchase, this type of product can be marked as the customer's potential demand. Conversely, if the customer has purchased a certain type of product, this type of product can be marked as his or her interest point. These marks can be achieved through rule judgment and algorithm recognition. For example, if a customer browses a certain product category more than 5 times but does not purchase it, the product category can be marked as a potential customer demand. In addition, through methods such as cluster analysis, each customer can also be tagged with multiple interest tags, such as "sports enthusiast", "high-end electronic product demander", etc., to help understand customer needs more accurately.

[0046] S25: Classify customer behavior data, identify different types of customers, and assign corresponding type tags to different types of customers.

[0047] Specifically, after being cleaned, normalized, labeled, and other processing, customer behavior data will be classified according to indicators such as customer behavior patterns, purchase frequency and amount. For example, the K-means clustering algorithm can be used to divide customers into "high-value customers", "medium-value customers" and "low-value customers". These classification labels are based on customer purchase frequency, historical transaction amount, activity and other data. In addition, customers can be further subdivided into types such as "potential customers", "loyal customers", and "returning customers" based on their behavior trajectories, such as the number of views on a certain category of goods, the number of items added to the shopping cart, etc., and each type of customer is assigned corresponding labels. These labels help to formulate different marketing strategies to more accurately meet the needs of various types of customers.

[0048] In one embodiment, if Figure 4 As shown, in step S30, a deep learning algorithm is used to perform real-time analysis on the pre-processed customer behavior data, and a customer profile is generated by combining the customer's historical behavior, preference changes and situational characteristics, specifically including: S31: Use deep neural network to extract features from pre-processed customer behavior data and identify key feature data.

[0049] Specifically, the preprocessed customer behavior data is first input into the deep neural network, and the important features in the data are gradually extracted through the multi-layer network structure. Each layer of the deep neural network will perform nonlinear transformation on the data to identify hidden patterns and relationships in the original data. For example, by inputting the customer's click behavior, browsing history and transaction records, the neural network can automatically learn the customer's interests and purchasing tendencies. For example, if a user frequently browses a certain brand of electronic products in the past week, this behavior will be extracted by the deep neural network as the user's purchase interest feature. In addition, the network weights are continuously adjusted through the back-propagation algorithm, so that each layer of the network can maximize the extraction of potential rules in the customer behavior data, thereby improving the accuracy and effectiveness of subsequent analysis.

[0050] S32: Construct a behavior prediction model based on historical behavior, recurrent neural networks, and long short-term memory networks.

[0051] Specifically, when building a behavior prediction model, the customer's historical behavior data, such as past purchase records, browsing history, etc., is first modeled through a recurrent neural network (RNN) for time series data. RNN can handle the time dependency in the data and help capture the temporal relationship of customer behavior. Then, the long short-term memory network (LSTM) is further used to strengthen the memory and analysis of long-term dependent information. Through its gating mechanism, LSTM can better handle long-term trends in customer behavior and changes in potential demand. For example, if a customer has frequently purchased sports shoes in the past two months, LSTM can capture this preference and predict that the customer may continue to purchase similar products in the future. Based on this information, the behavior prediction model can accurately predict the customer's future purchasing tendencies and behavior changes, thereby providing support for marketing decisions.

[0052] S33: Transfer key feature data to the behavior prediction model, combine preference changes and situational characteristics, model the time series information in the pre-processed customer behavior data, analyze the evolution of customer preferences, and generate customer portraits.

[0053] Specifically, key feature data extracted from the deep neural network, such as customers' interest tags, purchasing preferences, etc., are transmitted to the constructed behavior prediction model. Combined with changes in customer preferences and situational characteristics, such as shopping behavior in different seasons or during promotional activities, the model will dynamically adjust the prediction output, perform time series modeling, and analyze the evolution of customer behavior over time. For example, a customer frequently bought winter coats in the past winter, and gradually turned to buying spring clothes in the spring. This change can be clearly identified through the combination of LSTM and historical behavior data. Finally, the behavior prediction model will generate a customer portrait and form an in-depth understanding of the customer's long-term behavior trends. The customer portrait can not only reflect current preferences, but also predict future behavior patterns, such as predicting the customer's potential demand for sports equipment in the future, thereby providing data support for precision marketing.

[0054] In one embodiment, if Figure 5 As shown, in step S40, a marketing strategy is generated in real time based on the customer profile and the customer's real-time behavior through intelligent decision routing technology, which specifically includes: S41: Based on customer portraits, use the decision tree algorithm to dynamically evaluate customers' current and potential needs and generate marketing decision paths.

[0055] Specifically, first, based on the features extracted from the customer portrait, such as the customer's purchase history, browsing history, interest preferences, etc., a decision tree algorithm is used to dynamically evaluate the customer's current and potential needs. The decision tree continuously evaluates and divides the customer's demand levels by splitting nodes according to the customer's different behaviors and characteristics. For example, if a customer has browsed running shoes many times in the past three months and has a purchase record, the decision tree may infer that the customer currently has a strong demand for running shoes. If the customer has no such purchase behavior in the past few months, it may be judged that the customer's demand is potential demand. Subsequently, the decision tree generates marketing decision paths based on these judgments. These paths guide how marketing activities should be adjusted according to changes in customer needs. For example, if the current evaluation is a strong demand, relevant products will be recommended and displayed first. If it is a potential demand, promotional reminders will be given through coupons or targeted advertising.

[0056] S42: Determine marketing actions based on real-time customer behavior and marketing decision paths.

[0057] Specifically, based on the customer's real-time behavior, such as clicking, browsing, and adding to the shopping cart, combined with the marketing decision path generated in the previous step, the marketing actions to be taken are decided in real time. For example, if a customer does not immediately purchase a specific brand of sports shoes after browsing, and the decision path determines that the customer currently has a high demand for this type of product, then the marketing action can be automatically triggered to stimulate the customer's purchasing behavior by popping up discount information or sending promotional emails. On the other hand, if the customer only stays on the product details page and does not add to the shopping cart, and the marketing path determines that the customer is highly price sensitive, then the marketing action of displaying limited-time discounts may be adopted to incentivize customers to place orders through limited-time offers, thereby improving marketing efficiency.

[0058] S43: Generate marketing strategies by combining marketing decision paths, marketing actions and preset marketing rules.

[0059] Specifically, the generation of marketing strategies combines the marketing decision path and marketing actions, and at the same time, adjusts and optimizes according to preset marketing rules (such as promotion intensity, event time, user preferences, etc.). For example, if a customer has purchased high-end brand sports shoes in the past month and is currently in a network-wide promotion period, the strategy can be automatically adjusted to provide discount coupons or gifts to the customer according to the marketing rules. If the customer's purchase frequency is low, low-price promotions may be pushed first to increase their purchasing motivation. The marketing strategy will also be dynamically adjusted according to the customer's response to different types of discounts. For example, if the customer responds more to discount coupons but not to gifts, more discount coupons will be used in future marketing activities to improve the accuracy and effectiveness of the strategy, ultimately forming a set of personalized marketing strategies that adapt to changes in customer needs.

[0060] In one embodiment, if Figure 6 As shown, in step S40, dynamically adjusting the recommended content in the marketing strategy based on the customer interaction behavior in the real-time behavior of the customer specifically includes: S44: Analyze customer interaction behaviors to obtain customers’ immediate needs and interest changes. Customer interaction behaviors include clicking, browsing, collecting, sharing and purchasing.

[0061] Specifically, first, by collecting customer interaction data on different platforms, including clicking on a product link, browsing product pages, collecting products, sharing product information, and completing purchases, each behavior is counted and analyzed. Click data can indicate a customer's initial interest in a product, browsing behavior can provide a customer's attention to product details, collecting and sharing behaviors can reflect a customer's higher interest in a product and their potential purchase intentions, and purchasing behavior indicates that the customer has completed a transaction. Combined with the frequency and duration of these behaviors, the time window is used to analyze changes in customer demand in different time periods. For example, if a customer frequently clicks on a brand of sports shoes and repeatedly views product details but does not complete the purchase, and at the same time shares the product with others, it can be inferred that the customer has a strong interest in the brand. By analyzing the dynamic changes of these behaviors, we can further infer the customer's current demand changes. For example, the customer may be waiting for a discount or have other purchase plans. Such analysis helps to capture fluctuations in customer demand in a timely manner.

[0062] S45: Based on the user's feedback on the recommended content, the effectiveness of the recommended content is evaluated and a feedback result is generated.

[0063] Specifically, analysis is conducted based on customers' real-time feedback on recommended content. These feedbacks can be customers' click-through rate, dwell time, whether they add products to the shopping cart, whether they share products, and other behaviors. These customer reactions directly reflect the effectiveness of the recommended content. For example, if a recommended shoe model is clicked by most users and has a high purchase conversion rate, then the effectiveness of this recommended content is strong. On the contrary, if the recommended product is ignored or only clicked but no subsequent actions are taken, such as adding to the shopping cart or purchasing, then it means that the recommended content is less attractive. This feedback generates feedback results by comparing historical data with current data to evaluate the effectiveness of recommendations. These feedback results include which product recommendations are favored by customers and which fail to arouse customers' interest, thereby providing a basis for optimizing the next recommendation strategy.

[0064] S46: Based on the feedback results, the recommendation scheme is adjusted using an adaptive learning algorithm to generate a recommendation adjustment strategy, and the recommendation content is adjusted according to the recommendation adjustment strategy.

[0065] Specifically, by inputting the feedback results into the adaptive learning algorithm, the algorithm will automatically adjust the recommendation plan based on the customer's interaction data and feedback results to improve the recommendation effect. If the customer responds well to a specific style of shoes, the adaptive learning algorithm will increase the weight of this shoe in future recommendations, while reducing the recommendation frequency of other styles. Through the iteration of the algorithm, the display order of the recommended content, the recommended product types, and the recommended display methods can be dynamically adjusted. If the recommendation frequency of certain products is too high but cannot generate enough user conversions, the algorithm will reduce the display frequency of the product in future recommendations. Conversely, if the feedback on certain products is very positive, the recommendation algorithm will increase the frequency of its appearance in the user's recommendation list. Ultimately, through continuous feedback learning, a recommendation adjustment strategy is generated to optimize customer experience and improve marketing effectiveness, and the recommended content is adjusted based on the recommendation adjustment strategy.

[0066] In one embodiment, if Figure 7 As shown, the precision marketing method based on real-time data stream also includes: S60: Obtain customer behavior data at each customer contact point in real time, monitor customers’ immediate reactions to recommended content, analyze changes in customer preferences, and generate demand change data.

[0067] Specifically, customer contact points include channels such as websites, apps, social media, and advertising platforms. Every interactive behavior of the customer is recorded, for example, browsing a product page on a website, clicking a buy button, liking or sharing product information on a social media platform, or adding products to a shopping cart in an app. These behaviors represent the customer's immediate response to the recommended content. If a customer stays at a certain contact point for a long time and clicks frequently, it indicates that the customer has a strong interest in the recommended content. By calculating the customer's behavior trajectory and stay time, we can further analyze their reactions at different contact points. If a customer shows frequent interaction with a certain type of product at multiple contact points, it can be inferred that they have a high demand for this type of product. Customer preferences gradually change through these interactive behaviors. If a certain product or brand frequently appears in a customer's browsing history or interaction record over a period of time, the customer's preference may have changed. Finally, by aggregating these instant feedback and behavioral data, we generate demand change data.

[0068] S70: Adjust the display order, promotion channels and marketing methods of recommended content based on demand change data and real-time customer feedback.

[0069] Specifically, demand change data and customer feedback may include customers' click behavior, purchase behavior, sharing behavior, and browsing time on recommended content. By analyzing these data, we can discover changing trends in customer demand. For example, when the number of views of a specific product increases sharply, or customers frequently view a certain type of product in a specific time period, it can be inferred that customer demand has changed. Based on this information, the display order of recommended content can be adjusted, such as giving priority to displaying such products to customers. If customer demand tends to favor a certain category of goods, the frequency of related products can be increased in subsequent recommendations. Promotion channels and marketing methods also need to be adjusted according to changes in customer preferences. For example, if a customer frequently shares information about a certain product on a social media platform, the promotion of the related product can be increased on that platform. If the customer feedbacks a higher purchase intention through the APP, more accurate coupons or limited-time discount information can be pushed through the APP. The display order and marketing methods of recommended content can be dynamically adjusted based on changes in customer demand.

[0070] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0071] In one embodiment, a precision marketing system based on real-time data stream is provided, and the precision marketing system based on real-time data stream corresponds one-to-one with the precision marketing method based on real-time data stream in the above embodiment. Figure 8 As shown, the precision marketing system based on real-time data stream includes a data acquisition module, a preprocessing module, an analysis module, a strategy generation module and an optimization module. The detailed description of each functional module is as follows: Data collection module, used to collect customer behavior data in real time; A preprocessing module is used to preprocess and dynamically mark customer behavior data to generate preprocessed customer behavior data; The analysis module is used to use deep learning algorithms to perform real-time analysis on pre-processed customer behavior data, and generate customer profiles based on the customer's historical behavior, preference changes, and situational characteristics; The strategy generation module is used to generate marketing strategies in real time based on customer profiles and real-time customer behaviors through intelligent decision routing technology, and dynamically adjust the recommended content in the marketing strategy based on customer interaction behaviors in real-time customer behaviors; The optimization module is used to send service recommendations corresponding to the marketing strategy to the customer contact points according to the marketing strategy, and dynamically optimize the marketing strategy according to the customer's real-time response through the real-time feedback mechanism.

[0072] Optionally, the data acquisition module includes: The action acquisition submodule is used to use the multi-source data acquisition module to obtain the customer's action data in different channels in real time, and integrate the action data into customer behavior data. The action data includes website click streams, APP usage records, browsing history interaction data, and transaction data.

[0073] Optionally, the preprocessing module includes: The cleaning submodule is used to clean customer behavior data and remove noise and outliers; The time series analysis submodule is used to perform time series analysis on customer behavior data and identify customer behavior patterns and trends; Normalization submodule, used to normalize customer behavior data; The tagging submodule is used to tag customer behavior data and mark the customer's interests and potential needs; The classification submodule is used to classify customer behavior data, identify different types of customers, and assign corresponding type tags to different types of customers.

[0074] Optional analysis modules include: The feature extraction submodule is used to extract features from the pre-processed customer behavior data using a deep neural network to identify key feature data; The model building submodule is used to build a behavior prediction model based on historical behavior, recurrent neural network, and long short-term memory network; The modeling submodule is used to transfer key feature data to the behavior prediction model, combine preference changes and situational characteristics, model the time series information in the pre-processed customer behavior data, analyze the evolution of customer preferences, and generate customer portraits.

[0075] Optionally, the strategy generation module includes: The path determination submodule is used to dynamically evaluate the current and potential needs of customers based on customer portraits and use decision tree algorithms to generate marketing decision paths; The marketing action determination submodule is used to determine marketing actions based on real-time customer behavior and marketing decision paths; The strategy generation submodule is used to generate marketing strategies by combining marketing decision paths, marketing actions and preset marketing rules.

[0076] The demand analysis submodule is used to analyze customer interaction behaviors and obtain customers' immediate needs and interest changes. Customer interaction behaviors include clicks, browsing, favorites, sharing, and purchases. The feedback analysis submodule is used to evaluate the effectiveness of the recommended content and generate feedback results based on the user's feedback on the recommended content; The recommendation adjustment submodule is used to adjust the recommendation scheme based on the feedback results using an adaptive learning algorithm, generate a recommendation adjustment strategy, and adjust the recommendation content according to the recommendation adjustment strategy.

[0077] Optionally, the precision marketing system based on real-time data stream also includes: The immediate response monitoring module is used to obtain customer behavior data at each customer contact point in real time, monitor customers' immediate responses to recommended content, and analyze changes in customer preferences to generate demand change data; The sequence adjustment module is used to adjust the display order, promotion channels and marketing methods of recommended content according to demand change data and real-time customer feedback.

[0078] For the specific definition of a precision marketing system based on real-time data stream, please refer to the definition of a precision marketing method based on real-time data stream in the above text, which will not be repeated here. Each module in the above-mentioned precision marketing system based on real-time data stream can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0079] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0080] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A precision marketing method based on real-time data stream, characterized in that: The precision marketing method based on real-time data stream includes: Collect customer behavior data in real time; Preprocessing and dynamically marking the customer behavior data to generate preprocessed customer behavior data; Use a deep learning algorithm to perform real-time analysis on the pre-processed customer behavior data, and generate a customer profile by combining the customer's historical behavior, preference changes, and situational characteristics; Generate a marketing strategy in real time through intelligent decision routing technology according to the customer portrait and the customer's real-time behavior, and dynamically adjust the recommended content in the marketing strategy based on the customer's interactive behavior in the customer's real-time behavior; According to the marketing strategy, a service recommendation corresponding to the marketing strategy is sent to the customer contact point, and through a real-time feedback mechanism, the marketing strategy is dynamically optimized according to the real-time response of the customer.

2. The precision marketing method based on real-time data stream according to claim 1, characterized in that: The real-time collection of customer behavior data includes: Use a multi-source data acquisition module to obtain customer action data from different channels in real time, and integrate the action data into the customer behavior data. The action data includes website click streams, APP usage records, browsing history interaction data, and transaction data.

3. The precision marketing method based on real-time data stream according to claim 1, characterized in that: The preprocessing and dynamic marking of the customer behavior data includes: Cleaning the customer behavior data to remove noise and outliers; Performing time series analysis on the customer behavior data to identify customer behavior patterns and trends; Normalizing the customer behavior data; Marking the customer behavior data to identify the customer's interests and potential needs; The customer behavior data is classified to identify different types of customers and assign corresponding type tags to different types of customers.

4. The precision marketing method based on real-time data stream according to claim 1, characterized in that: The use of a deep learning algorithm to analyze the pre-processed customer behavior data in real time, combining the customer's historical behavior, preference changes and situational characteristics, to generate a customer profile includes: Using a deep neural network to extract features from the pre-processed customer behavior data to identify key feature data; Based on the historical behaviors, recurrent neural networks, and long short-term memory networks, a behavior prediction model is constructed; The key feature data is transmitted to the behavior prediction model, and the time series information in the pre-processed customer behavior data is modeled in combination with the preference changes and the situational features, the evolution process of customer preferences is analyzed, and the customer portrait is generated.

5. The precision marketing method based on real-time data stream according to claim 1, characterized in that: Generating a marketing strategy in real time by intelligent decision routing technology based on the customer portrait and the customer's real-time behavior includes: Based on the customer portrait, a decision tree algorithm is used to dynamically evaluate the customer's current and potential needs and generate a marketing decision path; Determine a marketing action based on the customer's real-time behavior and the marketing decision path; The marketing strategy is generated by combining the marketing decision path, the marketing action and the preset marketing rules.

6. The precision marketing method based on real-time data stream according to claim 1, characterized in that: The dynamically adjusting the recommended content in the marketing strategy based on the customer interaction behavior in the customer's real-time behavior includes: Analyze the customer interaction behaviors to obtain the customer's immediate needs and interest changes, and the customer interaction behaviors include clicking, browsing, collecting, sharing and purchasing; According to the user's feedback on the recommended content, evaluating the effectiveness of the recommended content and generating a feedback result; Based on the feedback result, an adaptive learning algorithm is used to adjust the recommendation scheme, generate a recommendation adjustment strategy, and adjust the recommended content according to the recommendation adjustment strategy.

7. The precision marketing method based on real-time data stream according to claim 1, characterized in that: The precision marketing method based on real-time data stream also includes: Acquire customer behavior data at each customer contact point in real time, monitor customer immediate response to the recommended content, analyze changes in customer preferences, and generate demand change data; According to the demand change data and real-time customer feedback, the display order, promotion channels and marketing methods of the recommended content are adjusted.

8. A precision marketing system based on real-time data stream, characterized in that: The precision marketing system based on real-time data stream includes: Data collection module, used to collect customer behavior data in real time; A preprocessing module, used to preprocess and dynamically mark the customer behavior data to generate preprocessed customer behavior data; An analysis module is used to use a deep learning algorithm to perform real-time analysis on the pre-processed customer behavior data, and generate a customer profile by combining the customer's historical behavior, preference changes, and situational characteristics; A strategy generation module, for generating a marketing strategy in real time based on the customer profile and the customer's real-time behavior through intelligent decision routing technology, and dynamically adjusting the recommended content in the marketing strategy based on the customer's interactive behavior in the customer's real-time behavior; The optimization module is used to send service recommendations corresponding to the marketing strategy to customer contact points according to the marketing strategy, and dynamically optimize the marketing strategy according to the real-time response of the customer through a real-time feedback mechanism.

Images