A method for serialized delivery of e-commerce advertisements based on user journey mapping

Abstract

The present invention provides an e-commerce advertising serialization delivery method based on user journey mapping. By integrating Cookies, pixel tracking and SDK technologies to collect user behavior data, the hidden Markov model (HMM) and conditional random field (CRF) are used to analyze the user purchase stage and design personalized advertising sequences. A reinforcement learning model is used to determine the optimal delivery time, and real-time bidding technology is used to optimize advertising display. At the same time, data analysis tools are used to evaluate the delivery effect, and a recurrent neural network (RNN) model is used to update the user journey model in real time to adapt to market changes. This method realizes personalized and intelligent advertising delivery, improves conversion rate, reduces cost, and provides an efficient advertising delivery solution for e-commerce platforms.

Description

Technical Field

[0001] The present invention relates to the technical field of e-commerce advertising delivery, and more particularly to a method for serialized delivery of e-commerce advertisements based on user journey mapping. Background Art

[0002] With the popularization of the Internet and the rapid rise of e-commerce platforms, advertising technology has also undergone a transformation from traditional media to digital media. The advent of the big data era has provided rich user data resources for advertising. These data are analyzed through advanced data analysis technologies such as user behavior analysis and preference mining, enabling advertisements to reach target consumers more accurately.

[0003] The popularity of mobile devices has changed consumers' shopping habits, and advertising technology has also shifted towards mobile devices, developing advertising formats and delivery strategies that are suitable for mobile devices. In addition, the vigorous development of social media has provided new delivery channels for e-commerce advertising, and advertising technology has begun to focus on social network effects to achieve better communication effects.

[0004] The timing of advertising delivery with existing technologies is often based on preset rules or simple statistical analysis, which may not necessarily match the user's actual purchasing intention. In addition, the user journey model may be static or have a long update cycle, and cannot reflect changes in user behavior in a timely manner. Summary of the Invention

[0005] In order to overcome the problems of existing technology that advertising delivery is based on preset rules or simple statistical analysis and is consistent with the user's actual purchasing intention, and the user journey model takes a long time to update, the present invention designs an e-commerce advertising serialization delivery method based on user journey mapping, which can effectively overcome the problems.

[0006] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0007] A method for serialized delivery of e-commerce advertisements based on user journey mapping includes the following steps:

[0008] Collect user behavior data on e-commerce platforms;

[0009] Analyze the collected user behavior data, identify the user's purchase stage and build a user journey model;

[0010] Design different advertising sequences based on the different purchasing stages of users;

[0011] Choose the right time to place ads based on the user's purchase stage and user behavior data;

[0012] Deliver the designed ad sequence at the chosen time;

[0013] Collect advertising data and evaluate advertising effectiveness;

[0014] Update the user journey model based on the collected advertising data and user behavior data.

[0015] Preferably, the collecting of user behavior data on the e-commerce platform further includes:

[0016] Use Cookies, pixel tracking and SDK integration technologies to collect user behavior data.

[0017] Preferably, analyzing the collected user behavior data, identifying the user's purchase stage, and constructing a user journey model further includes:

[0018] Encode user behavior data and convert it into a format suitable for HMM model processing;

[0019] Determine the five basic elements of the HMM model: state set, observation set, initial state probability distribution, state transition probability distribution, and observation probability distribution;

[0020] A state set that defines the user's purchase stage;

[0021] Observation set, the user's specific behavior;

[0022] Initial state probability distribution: the probability of a user being in each purchase stage when they first visit;

[0023] State transition probability distribution, the probability that a user moves from one purchase stage to another;

[0024] Observation probability distribution: the probability that a user will exhibit a specific behavior under a specific state;

[0025] Use the Baum-Welch algorithm to train the HMM model and iteratively update the HMM model parameters until the HMM model converges;

[0026] For new user behavior data, the Viterbi algorithm is used to calculate the hidden state sequence, that is, the user's purchase stage.

[0027] Preferably, the step of designing different advertisement sequences according to different purchasing stages of users further includes:

[0028] Extract features from user behavior data;

[0029] Convert the extracted features into the format of CRF model input;

[0030] Convert structured and unstructured data into numerical feature vectors;

[0031] Organize features into sequence format;

[0032] Build an input format that the CRF model understands;

[0033] Use the labeled user behavior data and corresponding advertising content as training data to train the CRF model.

[0034] Adjust the CRF model parameters and optimize the CRF model performance until the CRF model can accurately predict the advertising content corresponding to the user's purchase stage;

[0035] For each identified user purchase stage, use the trained CRF model for semantic matching to select the most appropriate advertising content;

[0036] Based on the prediction results of the CRF model, design advertising sequences for different purchase stages.

[0037] Preferably, selecting the timing for advertising delivery based on the user's purchase stage and user behavior data further includes:

[0038] Build a reinforcement learning model and define state-action pairs, where the state represents the user's current purchasing stage and behavior data, and the action represents the timing of ad delivery;

[0039] Design rewards, with positive responses from users such as clicks and purchases as positive rewards, and negative responses such as no responses as negative rewards;

[0040] By interacting with the environment, it learns the optimal advertising strategy to maximize the cumulative reward;

[0041] Select the best time to deliver ads based on the output of the reinforcement learning model;

[0042] Use ad servers and demand-side platforms to automatically deploy and schedule ads based on predetermined opportunities.

[0043] Preferably, delivering the designed advertisement sequence at a selected time further includes:

[0044] Real-time bidding technology is used to dynamically adjust ad display on the ad exchange platform to improve the efficiency and effectiveness of ad delivery.

[0045] Preferably, the collecting of advertising delivery data and evaluating the effectiveness of advertising delivery further includes:

[0046] Use data analysis tools to collect ad display, click and conversion data;

[0047] Use heat maps to conduct in-depth analysis of advertising data to evaluate ad impressions, click-through rates, and conversion rates;

[0048] Evaluate advertising effectiveness through multi-dimensional analysis and identify optimization points.

[0049] Preferably, updating the user journey model based on the collected advertising data and user behavior data further includes:

[0050] Use RNN models to process user behavior data and capture dynamic changes in user behavior;

[0051] Through online learning and continuous learning, the RNN model can be updated in real time to adapt to changes in user behavior;

[0052] Optimize the architecture and parameters of the RNN model to improve the model's predictive ability and stability.

[0053] A computer storage medium stores computer instructions, which, when called, are used to execute the above-mentioned method for serialized delivery of e-commerce advertisements based on user journey mapping.

[0054] Compared with existing technologies, the present invention has the following beneficial effects: By collecting user behavioral data on e-commerce platforms and utilizing technologies such as hidden Markov models (HMMs) and conditional random fields (CRFs), this method accurately identifies users' purchasing stages and designs personalized advertising sequences for users at different purchasing stages. This makes advertising more tailored to users' actual needs and improves advertising conversion rates. A reinforcement learning model is used to select the optimal timing for advertising based on user purchasing stages and behavioral data, thus achieving intelligent advertising delivery. Compared with traditional preset rules or simple statistical analysis, this method can more accurately grasp user needs and improve advertising delivery effectiveness. Through real-time bidding technology, advertising display can be dynamically adjusted on the advertising trading platform, improving the efficiency and effectiveness of advertising delivery. This enables a more reasonable allocation of advertising resources and reduces advertising costs. Recurrent neural networks (RNNs) are used to process user behavior data and capture dynamic changes in user behavior in real time. Through online learning and continuous learning, the RNN model can be quickly updated to adapt to changes in user behavior, shortening the update cycle of the user journey model. This method collects advertising data and conducts in-depth analysis through data analysis tools to evaluate advertising exposure, click-through rate, and conversion rate. Multi-dimensional analysis helps identify optimization points, thereby continuously optimizing advertising delivery strategies. Due to the high match between advertising content and user purchasing intentions, this method helps to improve user experience, reduce the harassment of invalid advertising, and improve user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the implementation methods or the description of the prior art. Obviously, the drawings described below are merely exemplary. For ordinary technicians in this field, other implementation drawings can be derived based on the provided drawings without any creative work.

[0056] Figure 1 A step-by-step diagram of a method for serializing e-commerce advertising based on user journey mapping. DETAILED DESCRIPTION

[0057] The accompanying drawings are for illustrative purposes only and are not to be construed as limiting this patent;

[0058] In order to better illustrate this embodiment, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product size;

[0059] It is understandable to those skilled in the art that some well-known structures and descriptions thereof may be omitted in the drawings.

[0060] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0061] Example

[0062] A method for serialized delivery of e-commerce advertisements based on user journey mapping, such as Figure 1 As shown, the following steps are included:

[0063] Collect user behavior data on e-commerce platforms, such as browsing history, search history, purchase history, interaction history, etc.

[0064] Analyze the collected user behavior data, identify the user's purchase stage and build a user journey model; for example, the browsing stage, comparison stage, purchase stage, after-sales stage, etc.

[0065] Design different advertising sequences based on the different purchasing stages of users;

[0066] Choose the right time to place ads based on the user's purchase stage and user behavior data;

[0067] Deliver the designed ad sequence at the chosen time;

[0068] Collect advertising data, such as click-through rate and conversion rate, to evaluate advertising effectiveness;

[0069] Update the user journey model based on collected advertising data and user behavior data. For example, adjust the user purchase stage segmentation and user profile.

[0070] The collection of user behavior data on the e-commerce platform further includes:

[0071] Use Cookies, pixel tracking and SDK integration technologies to collect user behavior data.

[0072] Analyzing the collected user behavior data, identifying the user's purchase stage, and constructing a user journey model further includes:

[0073] Encode user behavior data and convert it into a format suitable for HMM model processing; encoding user behavior data is to convert various user behaviors into a digital format that can be understood by the machine learning model.

[0074] A Hidden Markov Model (HMM) is a statistical model that describes a series of possible states a system can be in and the probabilities of transitions between these states. The state of the system cannot be observed directly but can be inferred from observed events. HMM is widely used in fields such as speech recognition, natural language processing, bioinformatics, and time series analysis.

[0075] Determine the five basic elements of the HMM model: state set, observation set, initial state probability distribution, state transition probability distribution, and observation probability distribution.

[0076] A set of states that define the user's purchasing stage; such as awareness stage, consideration stage, decision stage, purchase stage, etc.

[0077] In the awareness stage, users begin to realize their needs or become interested in a certain type of product.

[0078] In the consideration stage, users are considering different brands or product options.

[0079] The decision stage is when users compare and evaluate options before making a final purchase decision.

[0080] In the purchase stage, users actually make purchases.

[0081] Observation set, the user's specific behavior; such as browsing products, adding to shopping carts, searching for keywords, etc.

[0082] Browsing products, users view the product details page.

[0083] Add to cart: User adds the item to the shopping cart.

[0084] Search keywords, users search for specific products or services on the platform.

[0085] Watch Video: Users watch a product introduction video.

[0086] Read reviews, where users read reviews from other users.

[0087] Complete the purchase, the user completes the payment and purchases the product.

[0088] Initial state probability distribution: the probability of a user being in each purchase stage when they first visit;

[0089] Describes the probability that a user is in the awareness stage, consideration stage, decision stage, or purchase stage when visiting an e-commerce platform for the first time.

[0090] State transition probability distribution, the probability of a user moving from one stage to another; describes the probability of a user moving from one purchasing stage to another, such as the probability of moving from the awareness stage to the consideration stage.

[0091] Observation probability distribution: the probability that a user will exhibit a specific behavior under a specific state;

[0092] During the awareness stage, users may be more likely to search for keywords or browse products.

[0093] During the consideration stage, users may be more likely to compare different products and read reviews.

[0094] During the decision stage, users may be more likely to add items to their cart or revisit items that have already been added to their cart.

[0095] During the purchase stage, the probability of users completing a purchase is highest.

[0096] Use the Baum-Welch algorithm to train the HMM model and iteratively update the HMM model parameters until the HMM model converges;

[0097] The Baum-Welch algorithm is an algorithm used to estimate the parameters of a Hidden Markov Model (HMM). In particular, during the training phase of the model, this algorithm is a special expectation-maximization (EM) algorithm that is used to iteratively adjust the parameters of the HMM until they converge to the optimal solution.

[0098] For new user behavior data, the Viterbi algorithm is used to calculate the hidden state sequence, that is, the user's purchase stage.

[0099] The Viterbi algorithm is a dynamic programming algorithm used to solve the decoding problem in Hidden Markov Models (HMMs), that is, given a series of observations and model parameters, find the state sequence that is most likely to produce these observations.

[0100] The designing of different advertising sequences according to different purchasing stages of users further includes:

[0101] Extract features from user behavior data;

[0102] Including user behavior type, timestamp, product category, user interaction level, etc.

[0103] Convert the extracted features into the format of CRF model input;

[0104] Conditional Random Field (CRF) is a statistical modeling method that is often used in structure prediction problems, especially in natural language processing (NLP) and computer vision.

[0105] Numerical normalization normalizes or standardizes continuous numerical features to make them on the same scale.

[0106] Category encoding, one-hot encoding (One-Hot Encoding) or label encoding (Label Encoding) of categorical features.

[0107] Missing value processing: fill in or delete missing feature values.

[0108] Convert structured and unstructured data into numerical feature vectors;

[0109] For structured data, such as user behavior statistics, it is directly converted into a numerical feature vector.

[0110] For unstructured data, such as text reviews, use NLP techniques to extract features, such as Bag of Words, TF-IDF, or Word Embedding.

[0111] Organize features into sequence format; since the CRF model processes sequence data, each user's behavior trajectory is regarded as a sequence, and each behavior is regarded as a sequence element.

[0112] Create a feature vector for each sequence element, containing all the features at that moment.

[0113] Build an input format that the CRF model understands;

[0114] The feature sequence is converted into the input format of the CRF model, which is a two-dimensional array, where each row represents a sequence element and each column represents a feature.

[0115] Label each sequence element with the correct label, i.e., the user’s purchase stage, for model learning.

[0116] Use the labeled user behavior data and corresponding advertising content as training data to train the CRF model.

[0117] Adjust model parameters and optimize model performance until the model can accurately predict the advertising content corresponding to the user's purchase stage.

[0118] For each identified user purchase stage, the trained CRF model is used for semantic matching to select the most appropriate advertising content.

[0119] Based on the prediction results of the CRF model, design advertising sequences for different purchase stages.

[0120] The selecting of the timing for advertising delivery based on the user's purchase stage and user behavior data further includes:

[0121] Build a reinforcement learning model and define state-action pairs, where the state represents the user's current purchasing stage and behavior data, and the action represents the timing of ad delivery;

[0122] Choose an appropriate reinforcement learning algorithm, such as Q-learning, SARSA, or Deep Q Network (DQN).

[0123] Design rewards, with positive responses from users such as clicks and purchases as positive rewards, and negative responses such as no responses as negative rewards;

[0124] By interacting with the environment, it learns the optimal advertising strategy to maximize the cumulative reward function;

[0125] Select the best time to deliver ads based on the output of the reinforcement learning model;

[0126] Use ad servers and demand-side platforms (DSPs) to automatically deploy and schedule ads based on pre-defined opportunities.

[0127] The delivering of the designed advertisement sequence at the selected timing further includes:

[0128] Real-time bidding (RTB) technology is used to dynamically adjust ad display on ad trading platforms to improve the efficiency and effectiveness of ad delivery.

[0129] The collecting of advertising data and evaluating the effectiveness of advertising further includes:

[0130] Use data analysis tools, such as Google Analytics, to collect ad impression, click, and conversion data;

[0131] Use heat maps to conduct in-depth analysis of advertising data to evaluate ad impressions, click-through rates, and conversion rates;

[0132] Evaluate advertising effectiveness and identify optimization points through multi-dimensional analysis, such as ROI, CPC, and CPL.

[0133] Updating the user journey model based on the collected advertising data and user behavior data further includes:

[0134] Use RNN models to process user behavior data and capture dynamic changes in user behavior;

[0135] Through online learning and continuous learning, the RNN model can be updated in real time to adapt to changes in user behavior;

[0136] Optimize the architecture and parameters of the RNN model to improve the model's predictive ability and stability.

[0137] A computer storage medium stores computer instructions, which, when called, are used to execute the above-mentioned method for serialized delivery of e-commerce advertisements based on user journey mapping.

[0138] In specific implementation, on a large e-commerce platform, by deploying Cookies, pixel tracking and SDK integration technology, user behavior data is collected, including browsing history, search history, click behavior, shopping cart additions and order history.

[0139] The collected user behavior data is encoded into a numerical format to facilitate HMM model processing.

[0140] Define the basic elements of the HMM model, including state sets such as "awareness stage", "consideration stage", and "decision stage", observation sets such as "browse products", "add to shopping cart", and "view reviews", initial state probability distribution, state transition probability distribution, and observation probability distribution.

[0141] The HMM model is trained using the Baum-Welch algorithm until the model converges.

[0142] For new user behavior data, the Viterbi algorithm is applied to calculate the most likely hidden state sequence, that is, the user's purchase stage.

[0143] Extract features from user behavior data, such as the product categories browsed by users and their dwell time.

[0144] Convert the features into the CRF model input format and construct the feature vector.

[0145] Use the labeled user behavior data and corresponding advertising content as training data to train the CRF model.

[0146] Adjust the CRF model parameters until the model can accurately predict the advertising content corresponding to the user's purchase stage.

[0147] For each identified purchase stage, the CRF model is used to select the most appropriate advertising content and design the advertising sequence.

[0148] Build a reinforcement learning model and define state-action pairs, where the state is the user's current purchase stage and behavior data, and the action is the timing of ad delivery.

[0149] Design a reward system where user clicks and purchases are positive rewards, and non-response is negative reward.

[0150] Learn the optimal delivery strategy by interacting with the environment.

[0151] Choose the best time to launch based on the model output.

[0152] Using real-time bidding technology, we dynamically adjust ad display on the ad exchange platform to improve delivery efficiency and effectiveness.

[0153] Use data analysis tools to collect ad impressions, clicks, and conversion data.

[0154] Use heat maps to drill down into your data and evaluate your advertising effectiveness.

[0155] Identify optimization points through multi-dimensional analysis.

[0156] Use the RNN model to process user behavior data and update the user journey model in real time to adapt to changes in user behavior.

[0157] Optimize the RNN model architecture and parameters to improve prediction ability and stability.

[0158] The same or similar reference numerals correspond to the same or similar components;

[0159] The terms used in the drawings to describe positional relationships are for illustrative purposes only and should not be construed as limiting this patent;

[0160] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not limitations on the implementation methods of the present invention. For ordinary technicians in the field, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all implementation methods here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A method for serialized delivery of e-commerce advertisements based on user journey mapping, characterized in that: The following steps are involved: Collect user behavior data on e-commerce platforms; Analyze the collected user behavior data, identify the user's purchase stage and build a user journey model; Analyzing the collected user behavior data, identifying the user's purchase stage, and constructing a user journey model further includes: Encode user behavior data and convert it into a format suitable for HMM model processing; Determine the five basic elements of the HMM model: state set, observation set, initial state probability distribution, state transition probability distribution, and observation probability distribution; A state set that defines the user's purchase stage; Observation set, the user's specific behavior; Initial state probability distribution: the probability of a user being in each purchase stage when they first visit; State transition probability distribution, the probability that a user moves from one purchase stage to another; Observation probability distribution: the probability that a user will exhibit a specific behavior under a specific state; Use the Baum-Welch algorithm to train the HMM model and iteratively update the HMM model parameters until the HMM model converges; For new user behavior data, the Viterbi algorithm is used to calculate the hidden state sequence, that is, the user's purchase stage; Design different advertising sequences based on the different purchasing stages of users; The designing of different advertising sequences according to different purchasing stages of users further includes: Extract features from user behavior data; Convert the extracted features into the format of CRF model input; Convert structured and unstructured data into numerical feature vectors; Organize features into sequence format; Build an input format that the CRF model understands; Use the labeled user behavior data and corresponding advertising content as training data to train the CRF model; Adjust the CRF model parameters and optimize the CRF model performance until the CRF model can accurately predict the advertising content corresponding to the user's purchase stage; For each identified user purchase stage, use the trained CRF model for semantic matching to select the most appropriate advertising content; Design advertising sequences targeting different purchase stages based on the prediction results of the CRF model; Choose the right time to place ads based on the user's purchase stage and user behavior data; The selecting of the timing for advertising delivery based on the user's purchase stage and user behavior data further includes: Build a reinforcement learning model and define state-action pairs, where the state represents the user's current purchasing stage and behavior data, and the action represents the timing of ad delivery; Design rewards, with positive responses from users such as clicks and purchases as positive rewards, and negative responses such as no responses as negative rewards; By interacting with the environment, it learns the optimal advertising strategy to maximize the cumulative reward; Select the best time to deliver ads based on the output of the reinforcement learning model; Use ad servers and demand-side platforms to automatically deploy and schedule ads based on pre-determined timings; Deliver the designed ad sequence at the chosen time; Collect advertising data and evaluate advertising effectiveness; Update the user journey model based on the collected advertising data and user behavior data.

2. The method for serialized delivery of e-commerce advertisements based on user journey mapping according to claim 1, characterized in that: The collection of user behavior data on the e-commerce platform further includes: Use Cookies, pixel tracking and SDK integration technologies to collect user behavior data.

3. The method for serialized delivery of e-commerce advertisements based on user journey mapping according to claim 1, characterized in that: The delivering of the designed advertisement sequence at the selected timing further includes: Real-time bidding technology is used to dynamically adjust ad display on the ad exchange platform to improve the efficiency and effectiveness of ad delivery.

4. The method for serialized delivery of e-commerce advertisements based on user journey mapping according to claim 1, characterized in that: The collecting of advertising data and evaluating the effectiveness of advertising further includes: Use data analysis tools to collect ad display, click and conversion data; Use heat maps to conduct in-depth analysis of advertising data to evaluate ad impressions, click-through rates, and conversion rates; Evaluate advertising effectiveness through multi-dimensional analysis and identify optimization points.

5. The method for serialized delivery of e-commerce advertisements based on user journey mapping according to claim 1, characterized in that: Updating the user journey model based on the collected advertising data and user behavior data further includes: Use RNN models to process user behavior data and capture dynamic changes in user behavior; Through online learning and continuous learning, the RNN model can be updated in real time to adapt to changes in user behavior; Optimize the architecture and parameters of the RNN model to improve the model's predictive ability and stability.

6. A computer storage medium, characterized in that The computer storage medium stores computer instructions, which, when called, are used to execute the method for serialized delivery of e-commerce advertisements based on user journey mapping as described in any one of claims 1 to 5.

Images