A snow-skiing tour product recommendation method and system based on neural collaborative filtering and linear confidence upper bound

Abstract

This invention relates to a method and system for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bounds, belonging to the field of e-commerce product recommendation technology. The invention addresses the exposure bias problem inherent in existing ice and snow tourism product recommendation methods. It fully considers the exposure bias problem in ice and snow tourism product recommendations, where over-recommendation of popular products leads to a long-tail phenomenon, concentrating traffic on high-exposure products and limiting user choices. Simultaneously, low-exposure products are unlikely to enter the user's recommendation list due to exposure bias, resulting in unfair recommendations. This invention combines multi-feature information of users to capture their true preferences and reduce the impact of exposure bias on recommendation results. By designing a neural collaborative network to extract multi-attribute information of users and employing a linear confidence upper bound algorithm to generate reward value features, it improves the ability to explore low-exposure products while maintaining the accuracy and personalization of recommendations. This invention integrates multi-feature information of users with product attribute data and reward value features into the recommendation model to optimize the exposure bias problem in traditional recommendation models and improve the fairness and diversity of recommendations.

Description

Technical Field

[0001] This invention relates to the field of e-commerce product recommendation technology, specifically to a method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound. Technical Background

[0002] In today's digital age, the digital-enabled cultural tourism integration development model, represented by e-commerce, is an important way to achieve the modernization of Chinese tourism, and the e-commerce industry has ushered in unprecedented development opportunities. With the increasing demand for personalized travel experiences, the role of methods for recommending ice and snow tourism products has become particularly important. Especially driven by the popularization of winter sports and the dissemination of ice and snow culture, ice and snow tourism has shown enormous economic potential. However, existing methods for recommending ice and snow tourism products still have certain shortcomings in this field. Current recommendation methods often tend to recommend popular ice and snow tourism products, neglecting users' personalized needs and potential niche markets. This makes it difficult for users to obtain more personalized and targeted ice and snow tourism product choices, thus limiting their access to niche ice and snow tourism products. Furthermore, the insufficient recommendation of long-tail products exacerbates the unfairness of information filtering, affecting not only the diversity of ice and snow tourism products but also limiting the platform's sales strategy optimization in the ice and snow tourism market. Therefore, existing recommendation methods have not effectively solved the following problems: 1) It is difficult to effectively integrate user behavior data and product attribute data, resulting in a lack of personalization in the recommendation results. 2) The accuracy of the preference assessment for new users or new products is not high, which makes the recommendation system perform poorly when facing emerging markets.

[0003] Prior art, document number CN116738066B, discloses a method, device, electronic device, and storage medium for recommending rural tourism services. It includes: acquiring tourism resource and tourist data; constructing a tourism resource profile; performing sentiment analysis on feedback data to obtain sentiment analysis results; determining a rating matrix based on the feedback data and the tourism resource profile; constructing a tourist profile based on basic data and the tourism resource profile, the tourist profile including basic feature tags and feature preference tags; calculating the potential similarity, basic similarity, and preference similarity of a first tourist based on the rating matrix, the basic feature tags, and the feature preference tags of the tourist profile; generating a similarity matrix for the first tourist by weighted fusion of the three similarities; identifying a first tourist similar to a second tourist based on the similarity matrix, and recommending tourism resources to the second tourist. This prior art ensures the reliability and accuracy of tourism resource-related data; and provides tourists with more precise rural tourism resource recommendation services. However, this prior art does not effectively integrate user behavior data and product attribute data.

[0004] The prior art, document CN115438871A, discloses a method and system for recommending ice and snow attractions by integrating preferences to eliminate popularity bias. Its key technical points include: grouping tourists by gender, age, and occupation, and mapping them to values ​​between [0 and 1]; calculating the similarity of tourists' multi-feature preferences based on the mapped values ​​of gender, age, and occupation; calculating tourist preference values ​​based on the similarity of tourist preferences and historical tourist rating data; constructing a tourist-preference value matrix based on the standardized tourist preference values; training a matrix factorization model based on tourist multi-feature preferences using the tourist-preference value matrix and historical tourist rating data to obtain the trained tourist-preference value matrix; and predictively recommending new users based on the trained tourist-preference value matrix. This effectively alleviates the unfavorable situation where low-popularity ice and snow attractions are difficult to recommend to tourists. However, it does not deeply consider the preferences of new users or new products. Summary of the Invention

[0005] In view of the above problems, this invention proposes a method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound, in order to solve the exposure bias problem in existing ice and snow tourism product recommendation methods.

[0006] A method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound includes the following steps:

[0007] Step 1: Collect user profile information and historical user rating data. User profile information includes gender, age, and occupation; historical user rating data includes ratings of tourist attractions and browsing, clicking, and purchase records of related e-commerce products.

[0008] Step Two: Collect characteristic information about ice and snow tourism products;

[0009] Furthermore, this includes the ID and type of the ice and snow tourism product, ice and snow tourism product packages (whether the product offers packages for multiple ice and snow activities), winter-themed activities (whether the product includes winter-specific activities such as skiing, snowmobiling, and ice sculpture exhibitions), transportation convenience (the surrounding transportation conditions), guided tours (whether the product provides detailed guided tours or online tours), scenic area signage (whether clear signage is set up within the scenic area), local cuisine (whether local winter cuisine is offered), and accessibility facilities (whether the product is equipped with accessibility facilities).

[0010] Step 3: Construct feature vectors for users and ice and snow tourism products. Using the user attribute information and ice and snow tourism product attribute information collected in Step 1 and Step 2, encode each feature into a low-dimensional dense vector through an embedding layer to form user feature vectors and ice and snow tourism product feature vectors.

[0011] Furthermore, for user characteristics, information such as gender, age, and occupation are encoded into feature vectors; for ice and snow tourism product characteristics, attributes such as product ID, product type, activity type (such as skiing, ice sculpture exhibition, etc.), transportation convenience, and guided tour services are integrated into a single high-dimensional vector representation.

[0012] Step 4: Calculate similarity by combining user behavior characteristics and content characteristics to obtain user preferences;

[0013] Step 5: Introduce the Linear Confidence Upper Bound (LinUCB) algorithm to calculate reward features, and explore and utilize dynamic balance to discover products that users may be interested in but have low exposure rates;

[0014] Furthermore, the LinUCB algorithm calculates the potential reward value for each winter tourism product based on historical user behavior and product attributes. In each recommendation, the system prioritizes products for recommendation based on the upper bound of the confidence interval of the reward value, thereby increasing the recommendation frequency of low-exposure winter tourism products.

[0015] The expected rewards that each user will receive for each ice and snow tourism product are as follows:

[0016]

[0017] This indicates a collection of potential ice and snow tourism products. Indicates the first The experiment involved ice and snow tourism products. Corresponding feature vector, preference vector It is unknown, but its estimated value can be calculated through ridge regression in each round of interaction. Each ice and snow tourism product maintains a [database name missing]. .

[0018] Step 6: Embed the reward value data into the neural collaborative filtering network model to generate reward value embedding vectors for users and products, and combine them with their respective nearest neighbor feature vectors to form the user and product representations required for matrix factorization and multilayer perceptron.

[0019] Step 7: Using the user and ice and snow tourism product representations obtained in Step 6, learn the interaction function between users and ice and snow tourism products to obtain the output vector of the multilayer perceptron. :

[0020]

[0021] in, For activation function, For the first Layer weight matrix, for Layer output, This is the bias vector.

[0022] Step 8: Concatenate the output vectors from the matrix factorization and multilayer perceptron parts, and input them into a fully connected layer to obtain the user's output vector. and ice and snow tourism products and reward value Predicted scores between:

[0023]

[0024] in, The weight vector of the output layer. This represents the bias term of the output layer; This is the Sigmoid function.

[0025] Step 9: The output layer completes the prediction task. Based on the binary cross-entropy loss function, it optimizes the output user's preference probability for ice and snow tourism products. The prediction is transformed into a binary classification problem, and the prediction probability range is set to [0,1].

[0026] Step 10: Sort the final preference probabilities obtained in Step 9 from high to low, and select the top-ranked ones. Recommendations are made for ice and snow tourism products, including... The number of ice and snow tourism products recommended by the representative.

[0027] Step 11: Evaluate the performance of the method of the present invention using a standard dataset, and conduct ablation experiments, comparative experiments, and hyperparameter experiments respectively.

[0028] The beneficial technical effects of this invention are:

[0029] This invention effectively solves the exposure bias problem in current ice and snow tourism product recommendation methods by combining neural collaborative filtering and linear confidence upper bound algorithms, and introducing reward value features. By acquiring user characteristics such as gender, age, and occupation, as well as historical rating data, and mapping this information to values ​​between [0 and 1], a user-multi-feature matrix is ​​constructed to better capture users' true preferences, thereby reducing the impact of exposure bias on recommendation results.

[0030] This invention fully considers the exposure bias problem in the recommendation of ice and snow tourism products. Over-recommending popular products leads to a long-tail phenomenon in recommendations, concentrating traffic on high-exposure products and limiting users' choices. Simultaneously, low-exposure products are unlikely to enter users' recommendation lists due to exposure bias, resulting in unfair recommendations. This invention combines multi-feature information of users to capture their true preferences and reduce the impact of exposure bias on recommendation results. By designing a neural collaborative network to extract multi-attribute information of users and using a linear confidence upper bound algorithm to generate reward value features, this invention improves the ability to explore low-exposure products while maintaining the accuracy and personalization of recommendations. This invention integrates multi-feature information of users with product attribute data and reward value features into the recommendation model to optimize the exposure bias problem in traditional recommendation models and improve the fairness and diversity of recommendations. Ultimately, this invention not only improves the overall performance of the recommendation system but also enhances the user experience, providing a more scientific and accurate recommendation solution for tourism e-commerce platforms.

[0031] This invention designs a neural collaborative network to extract multi-attribute information from users and combines it with attribute information from ice and snow tourism products, including product ID, type, package deals, winter-themed activities, transportation convenience, guided tours, scenic area signage, specialty restaurants, and accessibility facilities. This attribute information is converted into binary sparse vectors using one-hot encoding and then fed into an embedding layer to generate embeddings. These feature vectors are then fed into a multilayer perceptron module and a matrix factorization module, and finally, the element-wise product is output through a fully connected layer and a softmax output layer. Finally, the output layer completes the prediction task, outputting the probability of user preference for ice and snow tourism products. A binary cross-entropy loss function is used for optimization, addressing the problem of poor performance of recommendation systems when facing new users or new products. It optimizes the deep mining of user behavior and product characteristics, improves the recommendation ability for low-exposure products, effectively alleviates the unfairness of information filtering for long-tail ice and snow tourism products, and enhances the fairness and diversity of recommendations. This allows the recommendation system to better meet users' personalized needs and promote the development of the tourism e-commerce industry.

[0032] This invention effectively integrates user behavior data and product attribute data, enabling personalized recommendations. It also achieves accurate assessment of preferences for new users or new products, making the recommendation system highly adaptable to emerging markets. Attached Figure Description

[0033] Figure 1 A schematic flowchart of the present invention is shown;

[0034] Figure 2 This is a comparison chart showing that the present invention demonstrates the best fairness and algorithm performance in the recommendation phase (it can be seen that the present invention has the smallest Gini coefficient compared with other methods).

[0035] Figure 3 Another comparison chart shows the optimal fairness and best algorithm performance of this invention in the recommendation phase (it can be seen that this invention has the smallest Gini coefficient compared with other methods).

[0036] Figure 4 This is a graph showing the experimental results of comparing the performance of different learning rates on the dataset. Detailed Implementation

[0037] Referring to the diagrams, the specific steps of the ice and snow tourism product recommendation method based on neural collaborative filtering and linear confidence upper bound of the present invention include:

[0038] The first step is to collect users' historical ratings of ice and snow tourism products, as well as their gender, age, and occupation information provided during registration. Based on this information, users are initially categorized, with users of the same gender, age, or occupation grouped together. The users' gender, age, and occupation characteristics are mapped to the range [0~1].

[0039] The second step is to collect feature information about ice and snow tourism products. The Ice and Snow Tourism Product ID serves as a unique identifier for each product, helping the system accurately track and analyze product exposure and user interaction. Category information, as an important factor for users choosing ice and snow tourism products, can improve the relevance of recommendations and user satisfaction.

[0040] Furthermore, the product's feature information also includes the ID and type of the ice and snow tourism product, ice and snow tourism product packages (whether the product offers packages for multiple ice and snow activities to meet users' comprehensive experience needs), winter-themed activities (whether the product includes winter-specific activities such as skiing, snowmobiling, and ice sculpture exhibitions to enrich users' ice and snow tourism experience), transportation convenience (the surrounding transportation situation of the product, such as whether direct or convenient shuttle transportation is provided to facilitate users' access to ice and snow tourism scenic spots), guided tours (whether the product provides detailed guided tours or online tours to help users gain a deeper understanding of the culture and characteristics of ice and snow tourism scenic spots), scenic spot signage (whether clear signage is set up within the scenic area to help users tour smoothly in the ice and snow environment), specialty dining (whether local specialty winter cuisine is provided to meet users' needs for local delicacies), and accessibility facilities (whether the product is equipped with accessibility facilities to provide an accessible tourism experience for users with mobility impairments), etc.

[0041] The third step is to use the user attribute information and ice and snow tourism product attribute information collected in steps one and two to construct feature vectors for users and ice and snow tourism products. Each feature is encoded into a low-dimensional dense vector through an embedding layer, forming the user feature vector and the ice and snow tourism product feature vector.

[0042] Furthermore, for user features, information such as gender, age, and occupation are encoded into feature vectors. The user's occupation information is mapped through a predefined occupation dictionary and transformed into a one-hot encoded vector. The user's age is directly input into the model as a continuous value feature. For ice and snow tourism product features, attributes such as product ID, type, activity type (such as skiing, ice sculpture exhibition, etc.), transportation convenience, and guided tour services are integrated into a single high-dimensional vector representation.

[0043] In the third step, after obtaining the feature vectors of users and ice and snow tourism products, considering the need to optimize model performance and maintain reasonable computational efficiency, we choose to reduce the dimensionality of this vector through a fully connected layer and process it through the ReLU activation function to obtain the final feature representation of the product.

[0044] The fourth step is to calculate similarity by combining user behavior characteristics and content characteristics to obtain user preferences;

[0045] The fifth step is to introduce the Linear Confidence Upper Bound (LinUCB) algorithm to calculate reward features, and through dynamic balance exploration and utilization, to discover products that users may be interested in but have low exposure rates;

[0046] Furthermore, the LinUCB algorithm calculates the potential reward value for each winter tourism product based on historical user behavior and product attributes. In each recommendation, the system prioritizes products for recommendation based on the upper bound of the confidence interval of the reward value, thereby increasing the recommendation frequency of low-exposure winter tourism products. The expected reward for each user for each winter tourism product is as follows:

[0047]

[0048] This indicates a collection of potential ice and snow tourism products. Indicates the first The experiment involved ice and snow tourism products. Corresponding feature vector, preference vector It is unknown, but its estimated value can be calculated through ridge regression in each round of interaction. Each ice and snow tourism product maintains a [database name missing]. The system will ultimately select the snow and ice tourism product with the highest upper bound of the confidence interval for recommendation.

[0049]

[0050] in . It is composed of historical information matrix, It is the dimension of the context feature vector. This represents the input training data. Indicates the estimated parameters, To balance the exploration and utilization of hyperparameters.

[0051] The historical information mainly includes user interaction records with items (such as ratings), contextual features (age and occupation), and a feature set consisting of parameter estimates from the model training process.

[0052] Furthermore, the ridge regression method is used to calculate the optimal reward value, and the optimization objective is as follows:

[0053]

[0054] in, yes The results of this experiment will define the reward value accordingly. Ultimately, the system will record the item that yielded the highest reward value. and These represent the user's acceptance and rejection of the recommendation, respectively. yes The eigenvector matrix observed in this experiment, where each row corresponds to an eigenvector. ,Right now Regularization, in which It is a diagonal matrix, which can effectively avoid overfitting.

[0055] right Taking the derivative, we get:

[0056]

[0057] If the first I've already recommended a few ice and snow tourism products. Then, the probability of this product receiving a reward can be calculated as follows:

[0058]

[0059] For any Both exist:

[0060]

[0061] in, It is the true probability of the product. These are hyperparameters. Therefore, the selection mechanism of the LinUCB algorithm satisfies:

[0062]

[0063] in, .

[0064] The sixth step is to embed the reward value data into the neural collaborative filtering network model to generate reward value embedding vectors for users and products, and combine them with their respective nearest neighbor feature vectors to form the user and product representations required for matrix factorization and multilayer perceptron.

[0065] The seventh step is to obtain the matrix factorization and the corresponding user and ice and snow tourism product representations for the multilayer perceptron. The lowest input layer consists of seven features: user ID, user gender, user age, user occupation, ice and snow tourism product ID, ice and snow tourism product type, and reward value. The interaction function between the user and the ice and snow tourism product is learned to obtain the output vector of the multilayer perceptron. :

[0066]

[0067] in, For activation function, For the first Layer weight matrix, for Layer output, This is the bias vector.

[0068] The eighth step is to concatenate the output vectors from the matrix factorization and multilayer perceptron parts, and then input them into a fully connected layer to obtain the user's output. and ice and snow tourism products and reward value Predicted scores between:

[0069]

[0070] in, The weight vector of the output layer. This represents the bias term of the output layer; This is the Sigmoid function.

[0071] The ninth step is for the output layer to complete the prediction task, outputting the probability of user preference for ice and snow tourism products. A binary cross-entropy loss function is used for optimization. Positive samples indicate interaction between the user and the ice and snow tourism product, while negative samples indicate no interaction. The prediction is transformed into a binary classification problem, with the prediction probability range set to [0,1].

[0072]

[0073] in, Represents the set of positive samples. This represents the set of negative samples.

[0074] The tenth step is to sort the final preference probabilities obtained in the ninth step from high to low, and select the top-ranked ones. Recommendations are made for ice and snow tourism products, including... The number of ice and snow tourism products recommended by the representative.

[0075] The eleventh step is to evaluate the performance of the method of the present invention using a standard dataset, and to conduct ablation experiments, comparative experiments and hyperparameter experiments respectively.

[0076] Furthermore, the ablation experiment was conducted to verify the effectiveness of each component in the method of the present invention. The neural collaborative filtering component and the reward value feature fusion component in the method were removed respectively, and ablation experiments were performed on the MovieLens-100K and MovieLens-1M datasets. The results are shown in Table 1.

[0077] Table 1

[0078]

[0079] Furthermore, by comparing the traditional NCF, EBPR, FaiRIR_RL, FaiRIR_Sim and the method of this invention, the recommendation performance of these five models was evaluated using fairness and accuracy metrics. The results show that the exposure performance of the five models varies on different datasets, but all of them help to reduce bias.

[0080] The NCF model had the lowest exposure across all recommended numbers; the EBPR model showed relatively stable and improved exposure, demonstrating its ability to mitigate exposure bias; the FaiRIR_Sim model showed significantly improved exposure at 10 and 20 recommended numbers, demonstrating its initial advantage; the FaiRIR_RL model had high overall exposure, reflecting the effectiveness of optimizing exposure balance; the method of this invention had the highest exposure across all recommended numbers, improving by approximately 60% compared to the NCF model, effectively reducing exposure bias.

[0081] Furthermore, the fairness of the models was evaluated based on the Gini coefficient (the lower the Gini coefficient, the higher the fairness). In the MovieLens-100K dataset, the NCF model had the lowest fairness and the highest Gini coefficient; the EBPR, FaiRIR_Sim, and FaiRIR_RL models performed well with lower numbers of recommendations, especially the FaiRIR_RL model; the method of this invention exhibited optimal fairness in all recommendation stages. See the results below. Figure 2 , 3 ;

[0082] Furthermore, during model training, the learning rate (lr), as a key hyperparameter, has a decisive impact on the model's convergence speed and final recommendation performance. To precisely explore the effect of the learning rate on model performance, the method of this invention employs a grid search strategy to optimize within a set of predefined learning rate values ​​{0.00001, 0.0001, 0.0005, 0.001, 0.005}. While keeping other parameters constant, experiments are conducted to compare the performance of the model on the dataset under different learning rates. The experimental results are as follows: Figure 4 As shown.

[0083] Experiments have verified that the neural collaborative filtering model incorporating reward features outperforms the original model in mitigating exposure bias and improving recommendation accuracy. While enhancing the ability to explore low-exposure products, it maintains the accuracy and diversity of recommendations, providing more comprehensive coverage of recommendation options, reducing biased recommendations for high-exposure products, and significantly alleviating the exposure bias problem, thereby achieving fairness in the recommendation results.

[0084] As can be seen, this invention fully considers the exposure bias problem in the recommendation of ice and snow tourism products. Over-recommending popular products leads to a long-tail phenomenon in recommendations, concentrating traffic on high-exposure products and limiting users' choices. Simultaneously, low-exposure products are unlikely to enter users' recommendation lists due to exposure bias, resulting in unfair recommendations. This invention combines multi-feature information of users to capture their true preferences and reduce the impact of exposure bias on recommendation results. By designing a neural collaborative network to extract multi-attribute information of users and employing a linear confidence upper bound algorithm to generate reward value features, it improves the ability to explore low-exposure products while maintaining the accuracy and personalization of recommendations.

[0085] This invention integrates user multi-feature information with product attribute data and reward value features into the recommendation model to optimize the exposure bias problem in traditional recommendation models and improve the fairness and diversity of recommendations.

[0086] Verification has shown that the method proposed in this invention solves the technical problem raised in this invention, and practical application has verified the technical effects and practicality claimed in this invention.

[0087] The method described in this invention has been verified through simulation experiments and practical applications, demonstrating the technical effects claimed by this invention.

[0088] The algorithm (method) proposed in this invention is the underlying technical core of this invention, and various products can be derived based on the algorithm.

[0089] Based on the algorithm (method) proposed in this invention, a snow and ice tourism product recommendation system based on neural collaborative filtering and linear confidence upper bound is developed using a programming language. This system has program modules corresponding to the steps of the above technical solution, and executes the steps in the above-mentioned method for recommending snow and ice tourism products based on neural collaborative filtering and linear confidence upper bound when running.

[0090] The developed system (software) computer program is stored on a computer-readable storage medium. The computer program is configured to implement the steps of the above-described method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bounds when invoked by a processor. In other words, the invention is materialized on a carrier, becoming a computer program product.

[0091] A device for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bounds is disclosed. The device includes at least one processor and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which are then executed to enable the at least one processor to perform the aforementioned method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bounds, thereby achieving scientific and effective recommendations for ice and snow tourism products on the Internet.

[0092] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, application-specific integrated circuits (ASICs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0093] The computational programs (also referred to as programs, software, software applications, or code) of this invention include machine instructions of a programmable processor and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device PLD) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0094] It should be understood that the various processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this application can be achieved, they are all within the protection scope of this invention.

Claims

1. A method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound, characterized in that, Includes the following steps: Step 1: Collect user characteristic information and user historical rating data: User characteristic information covers gender, age, and occupation; user historical rating data includes ratings of tourist attractions and browsing, clicking, and purchase records of related e-commerce products; Step Two: Collect characteristic information about ice and snow tourism products; Step 3: Construct feature vectors for users and ice and snow tourism products. Using the user attribute information and ice and snow tourism product attribute information collected in Step 1 and Step 2, encode each feature into a low-dimensional dense vector through an embedding layer to form user feature vectors and ice and snow tourism product feature vectors. Step 4: Calculate similarity by combining user behavior characteristics and content characteristics to obtain user preferences; Step 5: Introduce the Linear Confidence Upper Bound (LinUCB) algorithm to calculate reward features, and explore and utilize dynamic balance to discover products that users may be interested in but have low exposure rates; Step 6: Embed the reward value data into the neural collaborative filtering network model to generate reward value embedding vectors for users and products, and combine them with their respective nearest neighbor feature vectors to form the user and product representations required for matrix factorization and multilayer perceptron. Step 7: Using the user and ice and snow tourism product representations obtained in Step 6, learn the interaction function between users and ice and snow tourism products to obtain the output vector of the multilayer perceptron; Step 8: Concatenate the output vectors from the matrix factorization and multilayer perceptron parts, and input them into a fully connected layer to obtain the predicted score between the user and the ice and snow tourism product and reward value: Step 9: The output layer completes the prediction task. Based on the binary cross-entropy loss function, it optimizes the output user's preference probability for ice and snow tourism products. The prediction is transformed into a binary classification problem, and the prediction probability range is set to [0,1]. Step 10: Sort the final preference probabilities obtained in Step 9 from high to low, and select the top-ranked ones. Recommendations are made for ice and snow tourism products, including... The number of ice and snow tourism products recommended by the representative.

2. The method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound as described in claim 1, characterized in that, In step one, users are initially classified, and users with the same gender, age or occupation are grouped together. The gender, age and occupation characteristics of users are mapped to the range [0~1].

3. A method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bounds as described in claim 1 or 2, characterized in that, In step two, The ice and snow tourism product ID serves as a unique identifier for each product, helping the system accurately track and analyze the exposure and user interaction of ice and snow tourism products. Type information, as an important factor for users to choose ice and snow tourism products, can improve the relevance of recommendations and user satisfaction. The characteristic information of ice and snow tourism products includes the product ID, type, ice and snow tourism package, winter-themed activities, transportation convenience, guided tours, scenic area signage, specialty restaurants, and barrier-free facilities.

4. The method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound as described in claim 3, characterized in that, In step three, For user characteristics, gender, age, and occupation information are encoded into feature vectors; for ice and snow tourism product characteristics, product ID, type, activity type, transportation convenience, and tour guide service attributes are integrated into a single high-dimensional vector representation. After obtaining the feature vectors of users and ice and snow tourism products in the third step, considering the need to optimize model performance and maintain reasonable computational efficiency, we choose to reduce the dimensionality of this vector through a fully connected layer and process it through the ReLU activation function to obtain the final feature representation of the product.

5. The method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound as described in claim 4, characterized in that, In step five, the LinUCB algorithm calculates the potential reward value for each winter tourism product based on historical user behavior and product attributes. In each recommendation, the system prioritizes products for recommendation based on the upper bound of the confidence interval of the reward value, thereby increasing the recommendation frequency of low-exposure winter tourism products. The expected rewards that each user will receive for each ice and snow tourism product are as follows: This indicates a collection of potential ice and snow tourism products. Indicates the first The experiment involved ice and snow tourism products. Corresponding feature vector, preference vector It is unknown, but its estimated value is calculated using ridge regression in each round of interaction. Each ice and snow tourism product maintains a [value / value]. ; The system will ultimately select the snow and ice tourism product with the highest upper bound of the confidence interval for recommendation: in , It is composed of historical information matrix, It is the dimension of the context feature vector. This represents the input training data. Indicates the estimated parameters, To balance the exploration-utilization of hyperparameters; The historical information includes a feature set consisting of user and item interaction records, contextual features, and parameter estimates from the model training process. Furthermore, the ridge regression method is used to calculate the optimal reward value, and the optimization objective is as follows: in, yes The results of this experiment will define the reward value accordingly. Ultimately, the system will record the item that yielded the highest reward value. and These represent the user's acceptance and rejection of the recommendation, respectively. yes The eigenvector matrix observed in this experiment, where each row corresponds to an eigenvector; ,Right now Regularization, in which As a diagonal matrix, it can effectively avoid overfitting. right Taking the derivative, we get: If the first I've already recommended a few ice and snow tourism products. Then, the probability of this product receiving a reward can be calculated as follows: For any Both exist: in, It is the true probability of the product. Since it is a hyperparameter, the selection mechanism of the LinUCB algorithm satisfies: in, .

6. The method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound as described in claim 5, characterized in that, The output vector in step seven is: Step seven involves obtaining the matrix factorization and the corresponding user and snow tourism product representations for the multilayer perceptron. The lowest input layer consists of seven features: user ID, user gender, user age, user occupation, snow tourism product ID, snow tourism product type, and reward value. The interaction function between the user and the snow tourism product is learned to obtain the output vector of the multilayer perceptron. : in, For activation function, For the first Layer weight matrix, for Layer output, This is the bias vector.

7. The method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound as described in claim 6, characterized in that, In step eight, the output vectors from the matrix factorization and multilayer perceptron parts are concatenated and input into a fully connected layer to obtain the user's output vector. and ice and snow tourism products and reward value Predicted scores between: in, The weight vector of the output layer. Indicates the bias term of the output layer; For the Sigmoid function; This is the output vector of matrix decomposition.

8. The method for recommending ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound as described in claim 7, characterized in that, Step nine involves the output layer completing the prediction task, outputting the probability of user preference for ice and snow tourism products. A binary cross-entropy loss function is used for optimization. Positive samples indicate interaction between the user and the ice and snow tourism product, while negative samples indicate no interaction. The prediction is transformed into a binary classification problem, with the prediction probability range set to [0,1]. in, Represents the set of positive samples. This represents the set of negative samples.

9. A recommendation system for ice and snow tourism products based on neural collaborative filtering and linear confidence upper bound, characterized in that: The system has a program module corresponding to the steps of any one of the claims 1-8 above, and executes the steps in the ice and snow tourism product recommendation method based on neural collaborative filtering and linear confidence upper bound when it is run.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program configured to, when invoked by a processor, implement the steps of the ice and snow tourism product recommendation method based on neural collaborative filtering and linear confidence upper bound as described in any one of claims 1-8.

Images