Internet-based information intelligent directional accurate delivery system and pushing method thereof

Abstract

The invention provides an internet-based information intelligent directional accurate delivery system, which is characterized by comprising an analysis platform, the analysis platform comprises an intelligent matching module and an analysis engine, the intelligent matching module captures user key data and makes a mixed feature set, and the analysis engine is used for analyzing the mixed feature set; the analysis engine reads data of the mixed feature set and constructs a superficial portrait of the user; the fusion platform generates a user deep portrait according to an algorithm of the fusion platform and in combination with the user interest vector; the method relates to the technical field of Internet information accurate release. Through cooperative work of the analysis platform, the fusion platform and the delivery platform, deep mining, portrait construction and accurate content delivery of user data are realized; and the analysis platform, the fusion platform and the delivery platform realize a dynamic optimized pushing strategy of data by adopting a distributed crawling technology, time sequence prediction, transfer learning, reinforcement learning and intelligent routing decision, so that the information delivery accuracy and the user experience are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of precise delivery of Internet information, and in particular to an Internet-based intelligent, targeted, precise delivery system for information and a method for delivering the information. Background Art

[0002] In today's internet information age, businesses and platforms need to accurately deliver personalized content (such as advertisements, news, and product recommendations) to users to increase user engagement and business conversion rates. However, traditional recommendation systems typically rely solely on historical user behavior data (such as click and purchase records), making it difficult to fully capture real-time changes in user interests and susceptible to data sparsity and cold start issues. Furthermore, existing delivery systems often fail to consider factors such as user device status, network environment, and cross-platform data integration, resulting in poor push results and wasted resources.

[0003] The existing technology for precise information delivery relies on the user-item interaction matrix, but it is difficult to handle sparse data and new user problems. Although some algorithms can learn high-order feature interactions, they have high computational costs and poor adaptability to real-time data changes. In addition, such algorithms have weak dynamic adjustment strategy capabilities and lack the ability to integrate multi-source heterogeneous data, so there is room for improvement. Summary of the Invention

[0004] The purpose of the present invention is to provide an Internet-based intelligent targeted information delivery system and a push method thereof, so as to solve the problems raised by the background technology in the above content and overcome the technical defects thereof.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: an Internet-based intelligent targeted and precise information delivery system, including an analysis platform, the analysis platform includes an intelligent matching module and an analysis engine, the intelligent matching module captures user key data and makes a hybrid feature set, the analysis engine reads the data of the hybrid feature set and constructs a user shallow portrait; a fusion platform, the fusion platform generates a user deep portrait based on its own algorithm and combined with the user interest vector; and a delivery platform, the delivery platform combines and transforms the user shallow portrait and the user deep portrait data and accurately connects and delivers them with the Internet.

[0006] An Internet-based intelligent targeted information delivery system and its delivery method, including

[0007] Step S1: Use distributed crawling technology to obtain user basic attribute data, historical interaction tracks, and third-party platform authorization data in real time to build a hybrid feature set;

[0008] Step S2: using a time series prediction algorithm to analyze the mixed feature set, generate a user interest map including the user's current operation behavior data stream, terminal device status, geographic location and network environment parameters, and construct a user shallow profile based on this;

[0009] Step S3: Use the transfer learning framework to map user features into a structured label system that includes consumption propensity index, content sensitivity level, and social influence weight. Generate a deep user profile with spatiotemporal correlation through a feature cross-network.

[0010] Step S31: using named entity recognition technology to extract explicit attribute data of user age, occupation, and education level;

[0011] Step S32: Analyze the correlation between product categories in the user's historical orders through a convolutional neural network, and construct a product feature map based on the attention mechanism;

[0012] Step S33: Mapping multi-source heterogeneous data to the interest prediction model using a transfer learning adapter to generate a deep user profile with cross-domain feature fusion;

[0013] Step S4: Dynamically reconstruct the content recommendation priority queue based on the reinforcement learning decision network, and establish a closed-loop optimization mechanism for push strategy and user feedback signals;

[0014] Step S41: Generate an exponential decay function based on the user data timestamp and dynamically adjust the confidence weight of the historical behavior label;

[0015] Step S42: resolving conflicts among conflicting data labels and generating an improved label correction solution;

[0016] Step S43: combining the real-time feedback signal with the output of the strategy calculation sub-model to generate an optimal strategy combination that minimizes the user churn rate;

[0017] Step S5: Based on the remaining power of the user device, the real-time status of the network bandwidth and the platform traffic cost coefficient, the optimal delivery channel combination is calculated through the intelligent routing decision maker to achieve precise delivery of content adapted to cross-platform protocols.

[0018] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the hybrid feature set is constructed by obtaining user basic attribute data, historical interaction tracks and authorization data of third-party platforms through the distributed crawling technology of the intelligent matching module, and the analysis engine has a built-in time series prediction algorithm that can generate a user interest map, and the user interest map includes the user's current operation behavior data stream, terminal device status, geographic location and network environment parameters.

[0019] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the fusion platform includes a fusion model and a strategy unit, the fusion model uses a transfer learning framework to extract user features and map them into a structured label system including a consumption propensity index, content sensitivity level, and social influence weight, the fusion model generates a user deep portrait with spatiotemporal correlation through a feature cross-network, the strategy unit includes a decision network based on reinforcement learning, the strategy unit dynamically reconstructs the content recommendation priority queue according to the user deep portrait and establishes a closed-loop optimization mechanism of the push strategy and user feedback signal.

[0020] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the fusion model also includes an attribute mining sub-module and a transfer learning adapter. The attribute mining sub-module uses named entity recognition technology to extract data on user age, occupation, and education level, and combines convolutional neural networks to analyze the correlation between product categories in users' historical orders to construct a product feature map based on the attention mechanism. The transfer learning adapter maps the multi-source heterogeneous data associated with the user to the interest prediction model for use.

[0021] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the strategy unit includes a strategy calculation sub-model and a strategy generation sub-model, the strategy calculation sub-model generates a decay function according to the timestamp of the customer data and dynamically adjusts the confidence weight of the historical label, the decay function corrects the generated contradictory data and generates an optimal label correction solution, the strategy generation sub-model combines with the strategy calculation sub-model to generate the optimal strategy combination.

[0022] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the delivery platform includes a protocol adaptation layer and an intelligent routing decision maker, the protocol adaptation layer stores the protocols and rules required to be adapted by the Internet platform, and the intelligent routing decision maker calculates the optimal delivery channel combination based on the remaining power of the user device, the real-time status of the network bandwidth, and the platform traffic cost coefficient.

[0023] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the delivery platform also includes an intelligent routing selection algorithm, and the intelligent routing selection algorithm collaboratively pushes data content according to the interface protocol characteristics and rules of the target platform.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] This application achieves in-depth mining of user data, portrait construction and precise content delivery through the collaborative work of the analysis platform, fusion platform and delivery platform. In addition, the analysis platform, fusion platform and delivery platform implement a dynamic optimization and push strategy for data by adopting distributed crawling technology, time series prediction, transfer learning, reinforcement learning and intelligent routing decision-making, effectively improving the accuracy of information delivery and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The disclosure of the present invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the accompanying drawings, the same reference numerals are used to refer to the same components. Among them:

[0027] Figure 1 Schematically shows a general schematic diagram of the process steps proposed according to one embodiment of the present invention;

[0028] Figure 2 Schematically shows one of the process step sub-diagrams proposed according to one embodiment of the present invention;

[0029] Figure 3 The second schematic diagram of the process steps proposed in accordance with one embodiment of the present invention is schematically shown. DETAILED DESCRIPTION

[0030] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can propose a variety of interchangeable structural modes and implementation modes. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present invention and should not be regarded as the entire invention or as a limitation or restriction of the technical solution of the present invention.

[0031] According to one embodiment of the present invention, in combination with the accompanying drawings, an Internet-based intelligent, targeted, and precise information delivery system is shown, including an analysis platform, wherein the analysis platform includes an intelligent matching module and an analysis engine, wherein the intelligent matching module captures key user data and generates a hybrid feature set, wherein the capture adopts a distributed computing architecture (such as Hadoop, Spark, etc.) to efficiently capture user data, including basic user attribute data (such as gender, age, registration information, etc.); historical interaction trajectories (such as browsing records, click behaviors, length of stay, etc.) and then third-party platform authorization data (such as social media behaviors, e-commerce purchase records, etc.), and the hybrid feature set is a structured feature set integrated with the above data, and the analysis engine reads the data of the hybrid feature set and constructs a user shallow profile; a fusion platform, wherein the fusion platform generates a user deep profile based on its own algorithm and in combination with the user interest vector; and a delivery platform, wherein the delivery platform combines and transforms the user shallow profile and the user deep profile data and precisely connects and delivers them with the Internet.

[0032] This application achieves in-depth mining of user data, portrait construction and precise content delivery through the collaborative work of the analysis platform, fusion platform and delivery platform. In addition, the analysis platform, fusion platform and delivery platform implement a dynamic optimization and push strategy for data by adopting distributed crawling technology, time series prediction, transfer learning, reinforcement learning and intelligent routing decision-making, effectively improving the accuracy of information delivery and user experience.

[0033] An Internet-based intelligent targeted information delivery system and its delivery method, including

[0034] Step S1: Use distributed crawling technology to obtain user basic attribute data, historical interaction tracks, and third-party platform authorization data in real time to build a hybrid feature set;

[0035] Step S2: Using a time series prediction algorithm (such as LSTM or Prophet), the time series prediction algorithm here can analyze the trend of user behavior changes over time through the user's current operation behavior data stream (such as the content being browsed, click frequency), terminal device status (such as device model, screen resolution, system version), geographic location (such as GPS positioning, IP address) and network environment parameters (such as Wi-Fi / 4G / 5G status, latency, bandwidth) to generate a user interest mind map, parse the mixed feature set, generate a user interest map containing the user's current operation behavior data stream, terminal device status, geographic location and network environment parameters, and construct a user shallow profile based on this. Therefore, the user shallow profile can construct the user's short-term interests and behavior patterns based on the above data;

[0036] Step S3: Use a transfer learning framework (such as BERT or ResNet) to map user features into a structured label system that includes a consumption propensity index (a quantitative assessment of user purchase intention), a content sensitivity level (the user's preference for different types of content), and a social influence weight (the user's ability to spread information in social networks). This is then mapped through a feature cross-network (such as DeepFM, DCN, etc.). The feature cross-network can enhance the spatiotemporal correlation of the profile by analyzing the interactive relationship between features, and the feature cross-network can generate a deep user profile with spatiotemporal correlation.

[0037] Step S31: using named entity recognition technology to extract explicit attribute data of user age, occupation, and education level;

[0038] Step S32: Analyze the correlation between product categories in the user's historical orders through a convolutional neural network, and construct a product feature map based on the attention mechanism;

[0039] Step S33: Mapping multi-source heterogeneous data to the interest prediction model using a transfer learning adapter to generate a deep user profile with cross-domain feature fusion;

[0040] Step S4: Dynamically reconstruct the content recommendation priority queue based on the reinforcement learning decision network, and establish a closed-loop optimization mechanism for push strategy and user feedback signals;

[0041] Step S41: Generate an exponential decay function based on the user data timestamp and dynamically adjust the confidence weight of the historical behavior label;

[0042] Step S42: resolving conflicts among conflicting data labels and generating an improved label correction solution;

[0043] Step S43: combining the real-time feedback signal with the output of the strategy calculation sub-model to generate an optimal strategy combination that minimizes the user churn rate;

[0044] Step S5: Based on the remaining power of the user device, the real-time status of the network bandwidth and the platform traffic cost coefficient, the optimal delivery channel combination is calculated through the intelligent routing decision maker to achieve precise delivery of content adapted to cross-platform protocols.

[0045] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the hybrid feature set is constructed by obtaining user basic attribute data, historical interaction tracks and authorization data of third-party platforms through the distributed crawling technology of the intelligent matching module, and the analysis engine has a built-in time series prediction algorithm that can generate a user interest map, and the user interest map includes the user's current operation behavior data stream, terminal device status, geographic location and network environment parameters.

[0046] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the fusion platform includes a fusion model and a strategy unit, the fusion model uses a transfer learning framework to extract user features and map them into a structured label system including a consumption propensity index, content sensitivity level, and social influence weight, the fusion model generates a user deep portrait with spatiotemporal correlation through a feature cross-network, the strategy unit includes a decision network based on reinforcement learning, the strategy unit dynamically reconstructs the content recommendation priority queue according to the user deep portrait and establishes a closed-loop optimization mechanism of the push strategy and user feedback signal.

[0047] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the fusion model also includes an attribute mining sub-module and a transfer learning adapter. The attribute mining sub-module uses named entity recognition technology to extract data on user age, occupation, and education level, and combines convolutional neural networks to analyze the correlation between product categories in users' historical orders to construct a product feature map based on the attention mechanism. The transfer learning adapter maps the multi-source heterogeneous data associated with the user to the interest prediction model for use.

[0048] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the strategy unit includes a strategy calculation sub-model and a strategy generation sub-model, the strategy calculation sub-model generates a decay function according to the timestamp of the customer data and dynamically adjusts the confidence weight of the historical label, the decay function corrects the generated contradictory data and generates an optimal label correction solution, the strategy generation sub-model combines with the strategy calculation sub-model to generate the optimal strategy combination.

[0049] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the delivery platform includes a protocol adaptation layer and an intelligent routing decision maker, the protocol adaptation layer stores the protocols and rules required to be adapted by the Internet platform, and the intelligent routing decision maker calculates the optimal delivery channel combination based on the remaining power of the user device, the real-time status of the network bandwidth, and the platform traffic cost coefficient.

[0050] As a further solution of the present invention: an Internet-based intelligent targeted and precise information delivery system, the delivery platform also includes an intelligent routing selection algorithm, and the intelligent routing selection algorithm collaboratively pushes data content according to the interface protocol characteristics and rules of the target platform.

[0051] Working principle: This application realizes in-depth mining of user data, portrait construction and precise content delivery through the collaborative work of analysis platform, fusion platform and delivery platform. In addition, the analysis platform, fusion platform and delivery platform realize dynamic optimization and push strategy of data by adopting distributed crawling technology, time series prediction, transfer learning, reinforcement learning and intelligent routing decision, which effectively improves the accuracy of information delivery and user experience.

[0052] The technical scope of the present invention is not limited to the contents of the above description. Those skilled in the art can make various deformations and modifications to the above embodiments without departing from the technical idea of ​​the present invention, and these deformations and modifications should all fall within the protection scope of the present invention.

Claims

1. An Internet-based intelligent, targeted, and precise information delivery system, characterized in that: The analysis platform includes an intelligent matching module and an analysis engine. The intelligent matching module captures key user data and generates a hybrid feature set. The analysis engine reads the data in the hybrid feature set and constructs a superficial user profile. A fusion platform that generates a deep user profile based on its own algorithm and combined with user interest vectors; as well as The delivery platform combines and transforms user superficial profiles and user deep profile data and accurately connects them with the Internet for delivery.

2. The Internet-based intelligent targeted information delivery system according to claim 1, characterized in that: The hybrid feature set is constructed by obtaining user basic attribute data, historical interaction trajectories and authorization data of third-party platforms through the distributed crawling technology of the intelligent matching module. The analysis engine has a built-in time series prediction algorithm that can generate a user interest map. The user interest map includes the user's current operation behavior data stream, terminal device status, geographic location and network environment parameters.

3. The Internet-based intelligent targeted information delivery system according to claim 2, characterized in that: The fusion platform includes a fusion model and a strategy unit. The fusion model uses a transfer learning framework to extract user features and map them into a structured label system that includes a consumption propensity index, content sensitivity level, and social influence weight. The fusion model generates a deep user portrait with spatiotemporal correlation through a feature cross-network. The strategy unit includes a decision network based on reinforcement learning. The strategy unit dynamically reconstructs the content recommendation priority queue according to the deep user portrait and establishes a closed-loop optimization mechanism for push strategies and user feedback signals.

4. The Internet-based intelligent targeted information delivery system according to claim 3, characterized in that: The fusion model also includes an attribute mining submodule and a transfer learning adapter. The attribute mining submodule uses named entity recognition technology to extract data on user age, occupation, and education level, and combines it with a convolutional neural network to analyze the correlation between product categories in users' historical orders to construct a product feature map based on an attention mechanism. The transfer learning adapter maps the multi-source heterogeneous data associated with the user to the interest prediction model for use.

5. The Internet-based intelligent targeted information delivery system according to claim 4, characterized in that: The strategy unit includes a strategy calculation sub-model and a strategy generation sub-model. The strategy calculation sub-model generates a decay function based on the timestamp of the customer data and dynamically adjusts the confidence weight of the historical label. The decay function corrects the generated contradictory data and generates the optimal label correction solution. The strategy generation sub-model combines with the strategy calculation sub-model to generate the optimal strategy combination.

6. The Internet-based intelligent targeted information delivery system according to claim 5, characterized in that: The delivery platform includes a protocol adaptation layer and an intelligent routing decision maker. The protocol adaptation layer stores the protocols and rules that the Internet platform needs to adapt. The intelligent routing decision maker calculates the optimal delivery channel combination based on the remaining power of the user device, the real-time status of the network bandwidth and the platform traffic cost coefficient.

7. The Internet-based intelligent targeted information delivery system according to claim 6, characterized in that: The delivery platform also includes an intelligent routing selection algorithm, which coordinates the push of data content based on the interface protocol characteristics and rules of the target platform.

8. The Internet-based intelligent targeted information delivery system and its delivery method according to claim 7, characterized in that: include Step S1: Use distributed crawling technology to obtain user basic attribute data, historical interaction tracks, and third-party platform authorization data in real time to build a hybrid feature set; Step S2: using a time series prediction algorithm to analyze the mixed feature set, generate a user interest map including the user's current operation behavior data stream, terminal device status, geographic location and network environment parameters, and construct a user shallow profile based on this; Step S3: Use the transfer learning framework to map user features into a structured label system that includes consumption propensity index, content sensitivity level, and social influence weight. Generate a deep user profile with spatiotemporal correlation through a feature cross-network. Step S4: Dynamically reconstruct the content recommendation priority queue based on the reinforcement learning decision network, and establish a closed-loop optimization mechanism for push strategy and user feedback signals; Step S5: Based on the remaining power of the user device, the real-time status of the network bandwidth and the platform traffic cost coefficient, the optimal delivery channel combination is calculated through the intelligent routing decision maker to achieve precise delivery of content adapted to cross-platform protocols.

9. The Internet-based intelligent targeted information delivery system and its push method according to claim 8, characterized in that: The step S3 includes Step S31: using named entity recognition technology to extract explicit attribute data of user age, occupation, and education level; Step S32: Analyze the correlation between product categories in the user's historical orders through a convolutional neural network, and construct a product feature map based on the attention mechanism; Step S33: Use the transfer learning adapter to map multi-source heterogeneous data to the interest prediction model to generate a deep user profile with cross-domain feature fusion.

10. The Internet-based intelligent targeted information delivery system and its push method according to claim 9, characterized in that: The step S4 includes Step S41: Generate an exponential decay function based on the user data timestamp and dynamically adjust the confidence weight of the historical behavior label; Step S42: resolving conflicts among conflicting data labels and generating an improved label correction solution; Step S43: combining the real-time feedback signal with the output of the strategy calculation sub-model to generate an optimal strategy combination that minimizes the user churn rate.

Images