Tourist behavior multi-modal data fusion and precision marketing method based on AI and block chain

Abstract

The invention discloses a tourist behavior multi-modal data fusion and precision marketing method based on AI and a block chain, and the method comprises the following steps: firstly, fusing multi-source heterogeneous data through a federated learning technology, constructing a tourist multi-modal feature vector on the premise of protecting data privacy, and solving a cross-platform data island problem; secondly, designing a hierarchical authority control mechanism based on an alliance chain technology, and realizing safe sharing and compliance use of sensitive data by adopting homomorphic encryption and zero-knowledge proof; and finally, combining a reinforcement learning model and a multi-objective optimization algorithm to dynamically generate a personalized recommendation strategy, and realizing second-level response and strategy iteration through an online learning mechanism. The method provides safe, accurate and adaptive marketing decision support for the smart scenic spot.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence and data security technology, and specifically relates to a method for multimodal data fusion and precision marketing of tourist behavior based on AI and blockchain. It achieves data security collaboration through federated learning and blockchain technology, and improves marketing effectiveness by combining dynamic recommendation algorithms. Background Art

[0002] In traditional scenic spot marketing models, insights into visitor behavior often rely on manual experience, an approach that is not only inefficient but also lacks depth and precision. Scenic spot managers struggle to extract valuable visitor behavioral characteristics from massive amounts of data, significantly compromising the accuracy of recommendation strategies and failing to meet visitors' personalized needs, impacting both the visitor experience and the scenic spot's market competitiveness. Furthermore, cross-platform sharing of visitor privacy data presents significant security challenges. For example, sensitive data such as visitor location information and spending history can be easily leaked when transferred between platforms, posing potential privacy risks to visitors. However, the current lack of effective data usage traceability mechanisms makes it difficult to determine responsibility in the event of a data leak, and it is impossible to effectively monitor data usage. This further exacerbates visitors' concerns about scenic spot data security.

[0003] Most existing recommendation systems are based on static strategies and are unable to dynamically optimize based on real-time tourist behavior. This lag results in a misalignment between recommended content and actual tourist needs, leading to low marketing conversion rates and inefficient utilization of scenic spots' marketing resources. Some current market solutions have significant limitations. For example, some recommendation systems fail to integrate advanced blockchain technology and cannot provide adequate privacy protection during data sharing. Commercial tools like Google Analytics, while capable of providing a certain level of tourist behavior analysis, remain deficient in secure multi-platform data sharing, making them unable to meet the urgent needs of scenic spots for data privacy protection and collaborative marketing.

[0004] In summary, traditional scenic spot marketing faces multiple challenges, including insufficient precision, a high risk of visitor privacy leakage, and poor dynamic adaptability of recommendation systems. Therefore, an innovative solution combining AI and blockchain technology is urgently needed to simultaneously address the dual needs of accurate recommendations and privacy protection, driving scenic spot marketing towards intelligent and personalized development. Summary of the Invention

[0005] Aiming at the pain points of marketing and data management in smart scenic spots, the present invention innovatively proposes a method for multimodal data fusion and precision marketing of tourist behavior based on AI and blockchain. This method uses federated learning technology to integrate heterogeneous data from multiple platforms. While strictly adhering to the red line of data privacy, it constructs a multimodal feature vector covering tourists' consumption tendencies, action paths, and emotional tendencies, completely breaking down the barriers of cross-platform data. Relying on the hierarchical authority system of the alliance chain technology architecture, it uses cutting-edge encryption technologies such as homomorphic encryption and zero-knowledge proof to ensure the security and compliance of sensitive data sharing, and builds a firewall for the rational use of data. It simultaneously integrates reinforcement learning models and multi-objective optimization algorithms to instantly generate personalized recommendation plans based on the real-time dynamics of tourists, and with the help of online learning mechanisms, it achieves extremely rapid updates and continuous optimization of strategies to ensure the real-time and accurate fit of recommended content. This method tailors a safe, accurate, and flexible marketing decision-making engine for smart scenic spots, driving a wave of innovation in digital marketing for scenic spots.

[0006] The technical solution for achieving the purpose of the present invention is: A tourist behavior multimodal data fusion and precision marketing method based on AI and blockchain, including the following steps: 1) Multimodal data fusion: Federated learning is used to extract cross-platform features of raw data without exposing the original data. 2) Privacy protection and data sharing: Realize data ownership confirmation and controllable sharing based on alliance blockchain technology; 3) Dynamic collaborative recommendation strategy generation and optimization: Reinforcement learning models adjust strategies based on real-time behavioral data to improve conversion rates and visitor experience; The multimodal data fusion step in step 1) specifically includes: 1.1 Collect LBS positioning data, consumption records, public opinion texts and third-party platform data, clean them and store them in a distributed database; 1.2 Federated learning technology is used to fuse multi-source data to generate multimodal feature vectors of tourists (including consumption preferences, path trajectories, and emotional tendencies). Each participant extracts features from local raw data through a training model, and only model parameters, not raw data, are shared during the training process. Global multimodal feature vectors are generated through gradient aggregation to ensure data privacy.

[0007] The steps of privacy protection and data sharing in step 2) specifically include: 2.1 Sensitive data, namely location and identity information, is desensitized to generate an anonymized dataset. Geographic generalization technology is used for location data to convert precise latitude and longitude coordinates into regional ranges. Identity information is hashed and differentially privately processed to ensure that individual identities cannot be reversely inferred. Consumption records are segmented and desensitized, preserving statistical characteristics while eliminating sensitive information. 2.2 Build a data sharing network based on consortium chain technology, where the network nodes include scenic spots, merchants and regulatory agencies; use the Hyperledger Fabric framework to deploy chain code and define the data chain rules of network nodes; and verify data integrity through Merkle trees to ensure that the sharing process cannot be tampered with; 2.3 Smart contracts precisely define data access permissions, and visitors use their private keys to securely authorize third parties to use desensitized data, ensuring the security and compliance of data usage.

[0008] The steps of generating and optimizing the dynamic collaborative recommendation strategy in step 3) specifically include: 3.1 Build a reinforcement learning model that takes multimodal features of tourists as input and outputs personalized recommendation strategies (ticket discounts, route optimization, and product recommendations). The reinforcement learning model specifically includes: State space definition: real-time tourist behavior data (stay duration, click-through rate, spending amount); Action space definition: a set of recommendation strategies (discount strength, push frequency, content priority); Reward function design: Dynamically adjust weights based on conversion rate, visitor ratings, and privacy protection compliance; 3.2 The reinforcement learning model parameters are updated in real time through an online learning mechanism. These parameters are optimized using a dual-center + edge node computing architecture, multi-objective optimization algorithms, and federated learning technology. This ensures that the reinforcement learning model response time does not exceed 1 second, enabling immediate response to changes in tourist behavior and providing tourists with a seamless, personalized recommendation service experience. 3.3 A comprehensive application of A / B testing and a multi-objective evaluation system (covering key indicators such as return on investment (ROI) and visitor satisfaction) is used to deeply optimize recommendation effectiveness. By comparing the actual performance of different recommendation strategies, the optimal strategy combination can be accurately identified. Multi-objective evaluation comprehensively measures the performance of the recommendation system from the dual dimensions of commercial benefits and user experience. Based on the evaluation results, the recommendation system can automatically adjust the parameter configuration of the recommendation algorithm, achieve continuous iteration and upgrade of the recommendation strategy, and ensure that the recommended content is highly consistent with tourist preferences.

[0009] This method aims to propose an innovative precision marketing solution specifically for the efficient integration and in-depth utilization of tourist behavior data in smart scenic spots. This approach cleverly integrates artificial intelligence (AI) and blockchain technologies. Using federated learning, it integrates data resources from multiple platforms. While strictly protecting data privacy, it constructs a multimodal feature vector of tourists, encompassing multiple dimensions such as their consumption preferences, movement trajectories, and emotional tendencies. This effectively breaks down cross-platform data barriers and addresses the problem of data fragmentation. Furthermore, a carefully designed hierarchical permission management system employs consortium blockchain technology. Advanced encryption techniques such as homomorphic encryption and zero-knowledge proofs ensure the security and compliance of sensitive information during sharing, providing a solid foundation for the legitimate use of data. Furthermore, this technical solution combines reinforcement learning models with multi-objective optimization algorithms to dynamically generate personalized recommendation strategies based on real-time tourist behavior. Leveraging online learning mechanisms, it enables extremely low-latency strategy updates and optimization, ensuring the timeliness and accuracy of recommendations. This method creates a secure, accurate, and adaptive marketing decision support system for smart scenic spots, helping them achieve significant breakthroughs in digital marketing. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 is a flow chart of an embodiment; Figure 2 This is a flowchart of multimodal data fusion in an embodiment; Figure 3 This is a flowchart of privacy protection and data sharing in the embodiment; Figure 4 This is a flowchart of dynamic collaborative recommendation strategy generation and optimization in an embodiment. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Example

[0012] Reference Figure 1 , a tourist behavior multimodal data fusion and precision marketing method based on AI and blockchain, including the following steps: 1) Multimodal data fusion: Federated learning is used to extract cross-platform features of raw data without exposing the original data. 2) Privacy protection and data sharing: Realize data ownership confirmation and controllable sharing based on alliance blockchain technology; 3) Dynamic collaborative recommendation strategy generation and optimization: Reinforcement learning models adjust strategies based on real-time behavioral data to improve conversion rates and visitor experience; like Figure 2The multimodal data fusion step in step 1) specifically includes: 1.1 Collect LBS positioning data, consumption records, public opinion texts and third-party platform data, clean them and store them in a distributed database; 1.2 Federated learning technology is used to fuse multi-source data to generate multimodal feature vectors of tourists (including consumption preferences, path trajectories, and emotional tendencies). Participants such as scenic spots and online travel agencies use training models to extract features from local raw data, and the training process only shares model parameters rather than raw data. Global multimodal feature vectors are generated through gradient aggregation to ensure data privacy.

[0013] like Figure 3 The steps of privacy protection and data sharing in step 2) specifically include: 2.1 Sensitive data (location and identity information) is desensitized to generate anonymized datasets. Geographic generalization techniques (such as k-anonymity) are used on location data to convert precise latitude and longitude coordinates into regional ranges. Identity information is hashed and differentially private to ensure that individual identities cannot be reversely inferred. Consumption records are also desensitized in sections (for example, converting specific amounts into interval values), preserving statistical characteristics while eliminating sensitive information. 2.2 Build a data sharing network based on consortium chain technology, where the network nodes include scenic spots, merchants and regulatory agencies; use the Hyperledger Fabric framework to deploy chain code and define the data chain rules of network nodes; and verify data integrity through Merkle trees to ensure that the sharing process cannot be tampered with; 2.3 Smart contracts precisely define data access permissions, and visitors use their private keys to securely authorize third parties to use desensitized data, ensuring the security and compliance of data usage.

[0014] like Figure 4 The steps of generating and optimizing the dynamic collaborative recommendation strategy in step 3) specifically include: 3.1 Build a reinforcement learning model that takes multimodal features of tourists as input and outputs personalized recommendation strategies (ticket discounts, route optimization, and product recommendations). The reinforcement learning model specifically includes: State space definition: real-time tourist behavior data (stay duration, click-through rate, spending amount); Action space definition: a set of recommendation strategies (discount strength, push frequency, content priority); Reward function design: Dynamically adjust weights based on conversion rate, visitor ratings, and privacy protection compliance; 3.2 The reinforcement learning model parameters are updated in real time through an online learning mechanism. These parameters are optimized using a dual-center + edge node computing architecture, a multi-objective optimization algorithm, and federated learning technology. The dual-center + edge node computing architecture builds a primary data center and a backup data center within the reinforcement learning model, and combines this architecture with node connectivity at the edge of the primary and backup data centers to ensure that the reinforcement learning model response time does not exceed 1 second, enabling immediate response to changes in tourist behavior and providing tourists with a seamless, personalized recommendation service experience. 3.3 A comprehensive A / B testing and multi-objective evaluation system (covering key metrics such as return on investment (ROI) and visitor satisfaction) is used to deeply optimize recommendation effectiveness. By comparing the actual performance of different recommendation strategies, the optimal strategy combination is accurately identified. Multi-objective evaluation comprehensively measures the performance of the recommendation system from the dual perspectives of commercial benefits and user experience. Based on the evaluation results, the recommendation system can automatically adjust the parameter configuration of the recommendation algorithm, achieving continuous iteration and upgrading of the recommendation strategy, ensuring that the recommended content is highly consistent with visitor preferences.

Claims

1. A tourist behavior multimodal data fusion and precision marketing method based on AI and blockchain, characterized by: The following steps are involved: 1) Multimodal data fusion: Federated learning is used to extract cross-platform features of raw data without exposing the original data. 2) Privacy protection and data sharing: Realize data ownership confirmation and controllable sharing based on alliance blockchain technology; 3) Dynamic collaborative recommendation strategy generation and optimization: The reinforcement learning model adjusts strategies based on real-time behavioral data to improve conversion rates and visitor experience.

2. The method for multimodal data fusion and precision marketing of tourist behavior based on AI and blockchain according to claim 1 is characterized in that: The multimodal data fusion steps in step 1) specifically include: 1.1 Collect LBS positioning data, consumption records, public opinion texts and third-party platform data, clean them and store them in a distributed database; 1.2 Federated learning technology is used to fuse multi-source data to generate multimodal feature vectors of tourists. These feature vectors include consumption preferences, path trajectories, and emotional tendencies. Each participant extracts features from local raw data through a training model, and the training process only shares model parameters rather than raw data. Global multimodal feature vectors are generated through gradient aggregation to ensure data privacy.

3. The method for multimodal data fusion and precision marketing of tourist behavior based on AI and blockchain according to claim 1 is characterized in that: The steps of privacy protection and data sharing in step 2) specifically include: 2.1 Sensitive data, namely location and identity information, is desensitized to generate an anonymized dataset. Geographic generalization technology is used on location data to convert precise latitude and longitude coordinates into regional ranges. Identity information is hashed and differentially privately processed to ensure that individual identities cannot be reversely inferred. Consumption records are also desensitized in sections, preserving statistical features while removing sensitive information. 2.2 Build a data sharing network based on consortium chain technology, where the network nodes include scenic spots, merchants and regulatory agencies; use the Hyperledger Fabric framework to deploy chain code and define the data chain rules of network nodes; and verify data integrity through Merkle trees to ensure that the sharing process cannot be tampered with; 2.3 Smart contracts precisely define data access permissions, and visitors use their private keys to securely authorize third parties to use desensitized data, ensuring the security and compliance of data usage.

4. The method for multimodal data fusion and precision marketing of tourist behavior based on AI and blockchain according to claim 1 is characterized in that: The steps of generating and optimizing the dynamic collaborative recommendation strategy in step 3) specifically include: 3.1 Build a reinforcement learning model, with multimodal features of tourists as input and personalized recommendation strategies as output. The reinforcement learning model specifically includes: State space definition: real-time tourist behavior data, including length of stay, click-through rate, and spending amount; Action space definition: a set of recommendation strategies, i.e., discount intensity, push frequency, and content priority; Reward function design: Dynamically adjust weights based on conversion rate, visitor ratings, and privacy protection compliance; 3.2 The reinforcement learning model parameters are updated in real time through an online learning mechanism. These parameters are optimized using a dual-center + edge node computing architecture, multi-objective optimization algorithms, and federated learning technology. This ensures that the reinforcement learning model response time does not exceed 1 second, enabling immediate response to changes in tourist behavior and providing tourists with a seamless, personalized recommendation service experience. 3.3 A comprehensive application of A / B testing and a multi-objective evaluation system is used to deeply optimize recommendation effectiveness. By comparing the actual performance of different recommendation strategies, the optimal strategy combination can be accurately identified. Multi-objective evaluation comprehensively measures the performance of the recommendation system from the dual dimensions of commercial benefits and user experience. Based on the evaluation results, the recommendation system can automatically adjust the parameter configuration of the recommendation algorithm, realize the continuous iteration and upgrade of the recommendation strategy, and ensure that the recommended content is highly consistent with tourist preferences.

Images