Tourism data analysis method and system based on deep learning

Abstract

The invention relates to the field of data analysis, in particular to a tourism data analysis method and system based on deep learning. The method comprises the following steps: acquiring initial tourism data in Internet of Things equipment, performing data preprocessing on the initial tourism data, and performing feature extraction on standard tourism data by using a model architecture combining a convolutional neural network and a recurrent neural network; an analysis model of tourism data is constructed by fusing multiple deep learning algorithms to analyze characteristic tourism data, analysis at least comprises tourism preference analysis of tourists, tourism peak periods, change rules of scenic spot popularity along with time and consumption mode analysis of the tourists in different regions, and the tourists are subdivided by combining a clustering algorithm. And displaying the tourism behavior characteristic group through AR virtual augmented reality, and performing real-time dynamic adjustment according to user demands. Various data sources can be integrated, particularly social media and Internet of Things equipment data are fused, and a more comprehensive tourism data perspective is provided.

Description

Technical Field

[0001] The present invention relates to the field of data analysis, and in particular to a tourism data analysis method and system based on deep learning. Background Art

[0002] With the vigorous development of the global tourism industry, the scale of the tourism market continues to expand, and the diversity and richness of tourism activities are also increasing. Against this background, tourism-related data is showing an explosive growth trend. The data covers many aspects, from basic information of tourists, such as age, gender, region, etc., to detailed information of tourists' travel itinerary, including departure place, destination, and length of stay at various attractions; from tourists' evaluation information on attractions, such as satisfaction with attractions, improvement suggestions, etc., to consumption information during the travel process, such as transportation, accommodation, catering, shopping and other aspects of expenditure. Traditional data analysis methods have obvious limitations when facing such large-scale, high-dimensional, and multimodal tourism data. In terms of analysis efficiency, traditional methods often rely on manual or simple statistical tools. When processing massive data, it takes a lot of manpower and time. To analyze a large amount of tourist itinerary data, manual verification and statistics are required one by one, which makes the entire analysis process slow and unable to meet the needs of rapid decision-making. Summary of the invention

[0003] The purpose of the present invention is to solve the above problems and to design a tourism data analysis method and system based on deep learning.

[0004] To achieve the above object, the technical solution of the present invention is that, further, in the above-mentioned tourism data analysis method based on deep learning, the tourism data analysis method comprises the following steps:

[0005] Acquire initial tourism data from an IoT device, and perform data preprocessing on the initial tourism data to obtain standard tourism data;

[0006] Using a model architecture combining a convolutional neural network and a recurrent neural network to extract features from the standard tourism data, thereby obtaining characteristic tourism data;

[0007] By integrating multiple deep learning algorithms to build an analysis model for tourism data, the characteristic tourism data is analyzed. The analysis at least includes analysis of tourists' tourism preferences, peak tourism periods, changes in the popularity of scenic spots over time, and consumption patterns of tourists in different regions. Tourists are segmented by combining clustering algorithms to obtain groups with tourism behavior characteristics.

[0008] The tourism behavior characteristic groups are displayed through AR virtual augmented reality and adjusted dynamically in real time according to user needs.

[0009] Furthermore, in the above-mentioned tourism data analysis method, the step of obtaining the initial tourism data in the IoT device and performing data preprocessing on the initial tourism data to obtain standard tourism data includes:

[0010] Acquire initial travel data in the IoT device, where the initial travel data includes at least information about booked hotels, travel packages, and scenic spot tickets;

[0011] Extracting basic information of tourists, travel itinerary information and travel consumption information from the initial travel data;

[0012] The initial tourism data also includes collecting travel notes and guides posted by tourists from travel guide websites, obtaining tourists' evaluation information on different attractions, and tourists' detailed itinerary arrangements and personal experiences during the travel process;

[0013] The initial tourism data also includes establishing an interface with the ticket management system of the scenic spot to obtain the tourists' ticket purchase time, ticket quantity, and ticket type information in real time;

[0014] The initial tourism data also includes extracting data such as tourists' entry time and exit time from the ticketing system to calculate the length of time tourists stay in the scenic spot.

[0015] Furthermore, in the above-mentioned tourism data analysis method, the step of obtaining the initial tourism data in the IoT device and performing data preprocessing on the initial tourism data to obtain standard tourism data includes:

[0016] Acquire initial travel data, filter and delete the data based on preset rules, and obtain first initial travel data;

[0017] Remove duplicate records in the first initial tourism data, use the mean filling method for numerical data, and use the text generation model based on the GAN generative adversarial network to fill the text data, so as to obtain the second initial tourism data;

[0018] The numerical data in the second initial tourism data is processed by normalization, and the categorical data is processed by One-Hot encoding to obtain the third initial tourism data;

[0019] Aligning the travel experience text and picture information posted by tourists on social media in the third initial travel data according to the timestamp and location information in the text and the EXI F information in the picture;

[0020] By using the multimodal fusion technology in deep learning to evaluate the text and related pictures of a certain scenic spot, the attention mechanism will automatically learn the important parts of the text and pictures that describe the characteristics of the scenic spot, fuse the important features, and obtain standard tourism data.

[0021] Furthermore, in the above-mentioned tourism data analysis method, the model architecture combining convolutional neural network and recurrent neural network is used to extract features from the standard tourism data to obtain characteristic tourism data, including:

[0022] Establish a multi-layer CNN architecture, which includes 3-5 convolutional layers, and use 3x3 convolutional kernels of different sizes in the convolutional layers to extract image features of different scales, such as edges, textures, and shapes;

[0023] Add an average pooling layer after the convolutional layer, including a 2x2 maximum pooling window, to reduce the size of the image by half while retaining the most important feature information;

[0024] The RNN with GRU gated recurrent unit structure is used to process tourists’ movement trajectory data. GRU learns the movement patterns of tourists between different attractions.

[0025] The image features extracted by CNN and the sequence data processed by RNN are fused at the input layer, and the image features are used as additional input information, so that the model can use spatial and temporal information for feature extraction;

[0026] According to the importance of data features to tourism analysis, a learnable weight matrix is ​​defined. By multiplying the data features with the attention weight matrix, the attention score of each feature is obtained.

[0027] The original data features are weighted according to the attention scores, so that the model automatically focuses on the data features that have a high impact on tourism analysis and obtains characteristic tourism data.

[0028] Furthermore, in the above-mentioned tourism data analysis method, the analysis model of tourism data is constructed by integrating multiple deep learning algorithms to analyze the characteristic tourism data. The analysis at least includes analysis of tourists' tourism preferences, peak tourism periods, changes in the popularity of scenic spots over time, and consumption patterns of tourists in different regions. The clustering algorithm is combined to segment tourists to obtain tourism behavior characteristic groups, including:

[0029] Establish an autoencoder, and use the autoencoder to perform dimensionality reduction processing on the fused feature data. The autoencoder consists of an encoder and a decoder. The encoder compresses the high-dimensional feature data into a low-dimensional potential representation, and the decoder restores the potential representation to approximate original data. By adjusting the parameters of the autoencoder, the error after restoration is minimized, and the essential feature representation of the data is obtained;

[0030] An analysis model is established based on the GAN generative adversarial network, and the low-dimensional features extracted by the autoencoder are integrated with the characteristic tourism data again and input into the analysis model; the analysis model is used for data expansion and simulation:

[0031] The GAN generative adversarial network at least includes a generator and a discriminator, the generator is responsible for generating false data, and the discriminator is responsible for distinguishing true and false data;

[0032] The analysis includes at least analysis of tourists' travel preferences, peak travel periods, changes in popularity of scenic spots over time, and consumption patterns of tourists from different regions. Tourists are segmented using clustering algorithms to obtain groups with travel behavior characteristics.

[0033] Furthermore, in the above-mentioned tourism data analysis method, the analysis model of tourism data is constructed by integrating multiple deep learning algorithms to analyze the characteristic tourism data, and the analysis at least includes analysis of tourists' tourism preferences, peak tourism periods, changes in the popularity of scenic spots over time, and consumption patterns of tourists in different regions. The clustering algorithm is combined to segment tourists to obtain tourism behavior characteristic groups, and also includes:

[0034] The DBSCAN clustering algorithm is optimized by the GWO Grey Wolf Algorithm to obtain the GWO-DBSCAN algorithm; the clustering is based on the characteristics of tourists' travel behavior, travel consumption, and travel preference information extracted from social media.

[0035] The optimal number of clusters is determined by the elbow rule, and the tourism behavior characteristic groups are input into the GWO-DBSCAN algorithm to obtain different tourist groups, which at least include consumption culture exploration tourists, economical leisure and vacation tourists, family and parent-child tourists, and young backpackers.

[0036] Furthermore, in the above tourism data analysis method, the DBSCAN clustering algorithm is optimized by the GWO Grey Wolf Algorithm to obtain the GWO-DBSCAN algorithm, which also includes:

[0037] Setting a maximum number of iterations, a population size, and an initial vector value of the GWO algorithm, and generating an initial solution set according to the maximum number of iterations, the population size, and the initial vector value;

[0038] Substitute the elements in the initial solution set into the DBSCAN algorithm one by one, cluster the original user data set, and calculate and store the silhouette index of each clustering result;

[0039] Sort the initial solution set and the corresponding silhouette values ​​in ascending order to obtain the top three best solutions and top three best silhouette values ​​in the initial solution set;

[0040] Calculate the parameters of the coefficient vector, update the position of each element in the initial solution set, and re-substitute it into the DBSCAN algorithm to cluster the data set and calculate the silhouette value;

[0041] It is determined whether the maximum number of iterations of the GWO algorithm has been reached. If so, the maximum number of iterations, the population size and the initial vector value are output, and the maximum number of iterations, the population size and the initial vector value are set as initial parameters of the DBSCAN algorithm to obtain the GWO-DBSCAN algorithm.

[0042] Furthermore, in the above-mentioned tourism data analysis system based on deep learning, the tourism data analysis system includes:

[0043] A data acquisition module is used to acquire initial tourism data from an IoT device and perform data preprocessing on the initial tourism data to obtain standard tourism data;

[0044] A feature extraction module, used to extract features from the standard tourism data using a model architecture combining a convolutional neural network and a recurrent neural network to obtain feature tourism data;

[0045] A data analysis module is used to analyze the characteristic tourism data by building an analysis model of tourism data by integrating multiple deep learning algorithms. The analysis includes at least analysis of tourists' tourism preferences, peak tourism periods, changes in the popularity of scenic spots over time, and consumption patterns of tourists in different regions. The clustering algorithm is used to segment tourists and obtain groups with tourism behavior characteristics.

[0046] The data display module is used to display the tourism behavior characteristic groups through AR virtual augmented reality and make real-time dynamic adjustments according to user needs.

[0047] Furthermore, in the above-mentioned tourism data analysis system based on deep learning, the data acquisition module includes the following submodules:

[0048] A screening submodule is used to obtain initial travel data, screen and delete the data based on preset rules, and obtain first initial travel data;

[0049] A filling submodule is used to remove duplicate records in the first initial tourism data, use a mean filling method for numerical data, and use a text generation model based on a GAN generative adversarial network to fill text data, so as to obtain second initial tourism data;

[0050] A processing submodule, used for processing the numerical data in the second initial tourism data by normalization and processing the categorical data by One-Hot encoding to obtain the third initial tourism data;

[0051] An alignment submodule, for aligning the travel experience text and picture information posted by tourists on social media in the third initial travel data according to the timestamp and location information in the text and the EXI F information in the picture;

[0052] The fusion submodule is used to use the multimodal fusion technology in deep learning to evaluate the text and related pictures of a certain scenic spot. The attention mechanism will automatically learn the important parts of the description of the scenic spot features in the text and pictures, fuse the important features, and obtain standard tourism data.

[0053] Furthermore, in the above-mentioned tourism data analysis system based on deep learning, the feature extraction module includes the following submodules:

[0054] Establish a submodule for establishing a multi-layer CNN architecture, wherein the multi-layer CNN architecture includes 3-5 convolutional layers, and uses 3x3 convolutional kernels of different sizes in the convolutional layers to extract image features of different scales, such as edges, textures, and shapes;

[0055] Added a submodule for adding an average pooling layer after the convolutional layer, including a 2x2 max pooling window to reduce the size of the image by half while retaining the most important feature information;

[0056] The learning submodule is used to process tourists’ movement trajectory data using an RNN with a GRU gated recurrent unit structure. The GRU learns the movement patterns of tourists between different attractions.

[0057] The extraction submodule is used to fuse the image features extracted by CNN and the sequence data processed by RNN at the input layer, taking the image features as additional input information so that the model can extract features using spatial and temporal information;

[0058] The weight submodule is used to define a learnable weight matrix according to the importance of the data features to tourism analysis, and obtain the attention score of each feature by multiplying the data features with the attention weight matrix;

[0059] The analysis submodule is used to perform weighted processing on the original data features according to the attention score, so that the model automatically focuses on the data features with high impact on tourism analysis and obtains characteristic tourism data.

[0060] The beneficial effect is that by obtaining the initial tourism data in the IoT device, the initial tourism data is preprocessed to obtain standard tourism data; the standard tourism data is feature extracted by using a model architecture combining a convolutional neural network and a recurrent neural network to obtain characteristic tourism data; the characteristic tourism data is analyzed by building an analysis model for tourism data by integrating multiple deep learning algorithms, the analysis at least includes analysis of tourists' tourism preferences, peak tourism hours, the changing laws of scenic spot popularity over time, and consumption patterns of tourists in different regions, and the tourists are segmented by combining a clustering algorithm to obtain tourism behavior characteristic groups; the tourism behavior characteristic groups are displayed through AR virtual augmented reality, and are adjusted dynamically in real time according to user needs. Multiple data sources can be integrated, especially social media and IoT device data, to provide a more comprehensive perspective on tourism data. Combined with multiple advanced technologies such as convolutional neural networks, recurrent neural networks, and attention mechanisms, the accuracy and efficiency of feature extraction are improved. Tourists are segmented by clustering algorithms to provide tourism companies with a more accurate market positioning and a basis for personalized services. Augmented reality and interactive dynamic reports are used to make the analysis results more intuitive, easier to understand and use. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

[0062] Figure 1 Schematic diagram of a first embodiment of a tourism data analysis method based on deep learning in an embodiment of the present invention;

[0063] Figure 2 Schematic diagram of a second embodiment of a tourism data analysis method based on deep learning in an embodiment of the present invention;

[0064] Figure 3 This is a schematic diagram of a first embodiment of a tourism data analysis system based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with 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.

[0066] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "said" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0067] The present invention will be described in detail below in conjunction with the accompanying drawings. Figure 1 As shown, a tourism data analysis method based on deep learning, the tourism data analysis method includes the following steps:

[0068] Step 101: Acquire initial tourism data in an IoT device, perform data preprocessing on the initial tourism data, and obtain standard tourism data;

[0069] Specifically, in this embodiment, initial tourism data is obtained from the IoT device, and the initial tourism data at least includes information on booked hotels, travel packages, and scenic spot tickets; basic information of tourists, travel itinerary information, and travel consumption information are extracted from the initial tourism data; the initial tourism data also includes collecting travel notes and guides posted by tourists from travel guide websites, obtaining tourists' evaluation information on different scenic spots, and tourists' detailed itinerary arrangements and personal experiences during the tour; the initial tourism data also includes establishing an interface with the ticketing management system of the scenic spot to obtain tourists' ticket purchase time, ticket quantity, and ticket type information in real time; the initial tourism data also includes extracting tourists' entry time, exit time, and other data from the ticketing system to calculate the length of time tourists stay in the scenic spot.

[0070] Initial tourism data is obtained, and data is screened and deleted based on preset rules to obtain the first initial tourism data; duplicate records in the first initial tourism data are removed, and the mean filling method is used for numerical data, and the text generation model based on the GAN generative adversarial network is used to fill the text data to obtain the second initial tourism data; the numerical data in the second initial tourism data is normalized, and the categorical data is processed using One-Hot encoding to obtain the third initial tourism data; the tourism experience text and picture information posted by tourists on social media in the third initial tourism data are aligned according to the timestamp, location information in the text and the EXI F information of the picture; the multimodal fusion technology in deep learning is used for the evaluation text and related pictures of a certain scenic spot, and the attention mechanism will automatically learn the important parts of the description of the scenic spot features in the text and pictures, and the important features are fused to obtain standard tourism data.

[0071] Step 102: extract features from standard tourism data using a model architecture combining a convolutional neural network and a recurrent neural network to obtain feature tourism data;

[0072] Specifically, in this embodiment, a multi-layer CNN architecture is established, and the multi-layer CNN architecture includes 3-5 convolutional layers, and 3x3 convolutional kernels of different sizes are used in the convolutional layers to extract image features of different scales, such as edges, textures, and shapes;

[0073] Add an average pooling layer after the convolutional layer, including a 2x2 maximum pooling window, to reduce the size of the image by half while retaining the most important feature information;

[0074] The RNN with GRU gated recurrent unit structure is used to process the movement trajectory data of tourists. GRU learns the movement patterns of tourists between different scenic spots. The image features extracted by CNN and the sequence data processed by RNN are fused in the input layer. The image features are used as additional input information, so that the model can use spatial and temporal information for feature extraction. According to the importance of data features to tourism analysis, a learnable weight matrix is ​​defined, and the attention score of each feature is obtained by multiplying the data features with the attention weight matrix. The original data features are weighted according to the attention score, so that the model automatically focuses on the data features that have a high impact on tourism analysis, and obtains characteristic tourism data.

[0075] Step 103: Analyze the characteristic tourism data by building an analysis model of tourism data by integrating multiple deep learning algorithms. The analysis includes at least analysis of tourists' tourism preferences, peak tourism periods, changes in the popularity of scenic spots over time, and consumption patterns of tourists in different regions. The tourists are segmented by combining a clustering algorithm to obtain groups with tourism behavior characteristics.

[0076] Specifically, in this embodiment, an autoencoder is established, and the autoencoder is used to perform dimensionality reduction processing on the fused feature data. The autoencoder is composed of an encoder and a decoder. The encoder compresses high-dimensional feature data into low-dimensional potential representation, and the decoder restores the potential representation to approximate original data; by adjusting the parameters of the autoencoder, the error after restoration is minimized to obtain the essential feature representation of the data; an analysis model is established based on the GAN generative adversarial network, and the low-dimensional features extracted by the autoencoder are fused with the characteristic tourism data again and input into the analysis model; the analysis model is used to expand and simulate data: the GAN generative adversarial network includes at least a generator and a discriminator, the generator is responsible for generating false data, and the discriminator is responsible for distinguishing true and false data; the analysis includes at least analysis of tourists' travel preferences, peak travel periods, changes in the popularity of attractions over time, and consumption patterns of tourists in different regions, and tourists are segmented in combination with a clustering algorithm to obtain groups of tourism behavior characteristics.

[0077] The DBSCAN clustering algorithm is optimized by the GWO Grey Wolf Algorithm to obtain the GWO-DBSCAN algorithm; the clustering is based on multiple characteristics such as tourists' travel behavior characteristics, travel consumption characteristics and travel preference information extracted from social media; the optimal number of clusters is determined by the elbow rule, and the travel behavior characteristic groups are input into the GWO-DBSCAN algorithm to obtain different tourist groups, which include at least consumer culture exploration tourists, economical leisure and vacation tourists, family and parent-child tourists, and young backpackers.

[0078] The DBSCAN clustering algorithm is optimized by the GWO grey wolf algorithm to obtain the GWO-DBSCAN algorithm, which also includes: setting the maximum number of iterations, population size and initial vector value of the GWO algorithm, generating an initial solution set according to the maximum number of iterations, population size and initial vector value; substituting the elements in the initial solution set into the DBSCAN algorithm in sequence, clustering the original user data set, calculating and storing the silhouette index of each clustering result; sorting the initial solution set and the corresponding silhouette values ​​in ascending order, obtaining the optimal first three solutions and the optimal first three silhouette values ​​in the initial solution set; calculating the parameters of the coefficient vector, updating the position of each element in the initial solution set, and re-substituting it into the DBSCAN algorithm to cluster the data set and calculate the silhouette value; judging whether the maximum number of iterations of the GWO algorithm is reached, and if it is satisfied, outputting the maximum number of iterations, population size and initial vector value, setting the maximum number of iterations, population size and initial vector value as the initial parameters of the DBSCAN algorithm, and obtaining the GWO-DBSCAN algorithm.

[0079] Step 104: Display the tourism behavior characteristic groups through AR virtual augmented reality, and make real-time dynamic adjustments based on user needs.

[0080] Specifically, in this embodiment

[0081] Augmented Reality (AR) Display:

[0082] Developing mobile-based AR applications: The application can run on a smartphone or tablet and use the device's camera and sensors to obtain real-world scene information.

[0083] Data visualization: Combine tourism data analysis results with real-life scenarios for visualization. For example, when tourists open an AR application in a scenic spot, the scenic spot’s popularity information (such as hot spots of different colors to indicate the popularity of the scenic spot), tourist flow information (such as the number of people on the road), and the best tour route (indicated by arrows or lines) will be superimposed on the scenic spot screen seen through the camera.

[0084] Interactive function: The AR application provides interactive functions. Visitors can click on the hot spots on the screen to obtain more detailed information, such as attraction introductions, visitor reviews, etc.; they can also adjust the type and transparency of displayed information according to their needs.

[0085] Interactive dynamic report output:

[0086] Report generation: Use data visualization tools (such as Tableau, PowerBI, etc.) to generate dynamic reports. The report contains various charts (such as bar charts, line charts, pie charts, etc.) and data tables to show tourists' travel preferences, the changes in the popularity of attractions over time, and the consumption patterns of tourists in different regions.

[0087] Interactive design: Set interactive elements in the report, such as drop-down menus, filter buttons, sliders, etc. Users can select different analysis dimensions (such as different time periods, different attractions, different tourist groups, etc.) through interactive elements, and update the data and charts in the report in real time. For example, the tourism management department can select different seasons through the drop-down menu to view the flow distribution and consumption of tourists in the scenic area in that season; tourism companies can select specific tourist groups through the filter button to analyze tourists' travel needs and consumption potential.

[0088] The beneficial effect is that by obtaining the initial tourism data in the IoT device, the initial tourism data is preprocessed to obtain standard tourism data; the standard tourism data is feature extracted by using the model architecture combining convolutional neural network and recurrent neural network to obtain characteristic tourism data; the characteristic tourism data is analyzed by building an analysis model of tourism data by integrating multiple deep learning algorithms, and the analysis includes at least the analysis of tourists' travel preferences, peak travel time, the change of scenic spot popularity over time, and the consumption pattern analysis of tourists in different regions, and the tourists are segmented by combining clustering algorithms to obtain tourism behavior characteristic groups; the tourism behavior characteristic groups are displayed through AR virtual augmented reality, and are adjusted in real time according to user needs. It can integrate multiple data sources, especially the integration of social media and IoT device data, to provide a more comprehensive perspective on tourism data. It combines multiple advanced technologies such as convolutional neural network, recurrent neural network, attention mechanism, etc. to improve the accuracy and efficiency of feature extraction. The clustering algorithm is used to segment tourists, providing a basis for more accurate market positioning and personalized services for tourism enterprises. The use of augmented reality and interactive dynamic reports makes the analysis results more intuitive, easier to understand and use.

[0089] In this embodiment, please refer to Figure 2In a second embodiment of a tourism data analysis method based on deep learning in an embodiment of the present invention, obtaining initial tourism data in an IoT device, performing data preprocessing on the initial tourism data, and obtaining standard tourism data include the following steps:

[0090] Step 201: Acquire initial travel data, filter and delete the data based on preset rules, and obtain first initial travel data;

[0091] Step 202: remove duplicate records in the first initial tourism data, use the mean filling method for numerical data, and use the text generation model based on the GAN generative adversarial network to fill the text data, so as to obtain the second initial tourism data;

[0092] Step 203: normalize the numerical data in the second initial tourism data, and process the categorical data using One-Hot encoding to obtain the third initial tourism data;

[0093] Step 204: align the travel experience text and picture information posted by the tourists on the social media in the third initial travel data according to the timestamp and location information in the text and the EX IF information of the picture;

[0094] Step 205: Utilize the multimodal fusion technology in deep learning to evaluate the text and related pictures of a certain scenic spot. The attention mechanism will automatically learn the important parts of the description of the scenic spot features in the text and pictures, fuse the important features, and obtain standard tourism data.

[0095] The above describes a tourism data analysis method based on deep learning provided by an embodiment of the present invention. The following describes a tourism data analysis system based on deep learning provided by an embodiment of the present invention. Figure 3 In one embodiment of the present invention, a tourism data analysis system includes:

[0096] A data acquisition module is used to acquire initial tourism data from an IoT device and perform data preprocessing on the initial tourism data to obtain standard tourism data;

[0097] A feature extraction module, used to extract features from the standard tourism data using a model architecture combining a convolutional neural network and a recurrent neural network to obtain feature tourism data;

[0098] A data analysis module is used to analyze the characteristic tourism data by building an analysis model of tourism data by integrating multiple deep learning algorithms. The analysis includes at least analysis of tourists' tourism preferences, peak tourism periods, changes in the popularity of scenic spots over time, and consumption patterns of tourists in different regions. The clustering algorithm is used to segment tourists and obtain groups with tourism behavior characteristics.

[0099] The data display module is used to display the tourism behavior characteristic groups through AR virtual augmented reality and make real-time dynamic adjustments according to user needs.

[0100] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A tourism data analysis method based on deep learning, characterized in that: The tourism data analysis method comprises the following steps: Acquire initial tourism data from an IoT device, and perform data preprocessing on the initial tourism data to obtain standard tourism data; Using a model architecture combining a convolutional neural network and a recurrent neural network to extract features from the standard tourism data, thereby obtaining characteristic tourism data; By integrating multiple deep learning algorithms to build an analysis model for tourism data, the characteristic tourism data is analyzed. The analysis at least includes analysis of tourists' tourism preferences, peak tourism periods, changes in the popularity of scenic spots over time, and consumption patterns of tourists in different regions. Tourists are segmented by combining clustering algorithms to obtain groups with tourism behavior characteristics. The tourism behavior characteristic groups are displayed through AR virtual augmented reality and adjusted dynamically in real time according to user needs.

2. A tourism data analysis method based on deep learning as claimed in claim 1, characterized in that: The obtaining of the initial tourism data in the IoT device and performing data preprocessing on the initial tourism data to obtain standard tourism data include: Acquire initial travel data in the IoT device, where the initial travel data includes at least information about booked hotels, travel packages, and scenic spot tickets; Extracting basic information of tourists, travel itinerary information and travel consumption information from the initial travel data; The initial tourism data also includes collecting travel notes and guides posted by tourists from travel guide websites, obtaining tourists' evaluation information on different attractions, and tourists' detailed itinerary arrangements and personal experiences during the travel process; The initial tourism data also includes establishing an interface with the ticket management system of the scenic spot to obtain the tourists' ticket purchase time, ticket quantity, and ticket type information in real time; The initial tourism data also includes extracting data such as tourists' entry time and exit time from the ticketing system to calculate the length of time tourists stay in the scenic spot.

3. A tourism data analysis method based on deep learning as claimed in claim 1, characterized in that: The obtaining of the initial tourism data in the IoT device and performing data preprocessing on the initial tourism data to obtain standard tourism data include: Acquire initial travel data, filter and delete the data based on preset rules, and obtain first initial travel data; Remove duplicate records in the first initial tourism data, use the mean filling method for numerical data, and use the text generation model based on the GAN generative adversarial network to fill the text data, so as to obtain the second initial tourism data; The numerical data in the second initial tourism data is processed by normalization, and the categorical data is processed by One-Hot encoding to obtain the third initial tourism data; Aligning the travel experience text and picture information posted by tourists on social media in the third initial travel data according to the timestamp and location information in the text and the EXIF ​​information of the picture; By using the multimodal fusion technology in deep learning to evaluate the text and related pictures of a certain scenic spot, the attention mechanism will automatically learn the important parts of the text and pictures that describe the characteristics of the scenic spot, fuse the important features, and obtain standard tourism data.

4. A tourism data analysis method based on deep learning as claimed in claim 1, characterized in that: The model architecture combining convolutional neural network and recurrent neural network is used to extract features from the standard tourism data to obtain characteristic tourism data, including: Establish a multi-layer CNN architecture, which includes 3-5 convolutional layers, and use 3x3 convolution kernels of different sizes in the convolutional layers to extract image features of different scales, such as edges, textures, and shapes; Add an average pooling layer after the convolutional layer, including a 2x2 maximum pooling window, to reduce the size of the image by half while retaining the most important feature information; The RNN with GRU gated recurrent unit structure is used to process tourists’ movement trajectory data. GRU learns the movement patterns of tourists between different attractions. The image features extracted by CNN and the sequence data processed by RNN are fused at the input layer, and the image features are used as additional input information, so that the model can use spatial and temporal information for feature extraction; Define a learnable weight matrix to obtain the attention score of each feature by multiplying the data feature with the attention weight matrix; The original data features are weighted according to the attention scores, so that the model automatically focuses on the data features that have a high impact on tourism analysis and obtains characteristic tourism data.

5. The tourism data analysis method based on deep learning as claimed in claim 1, characterized in that: The analysis model of tourism data is constructed by integrating multiple deep learning algorithms to analyze the characteristic tourism data. The analysis at least includes analysis of tourists' tourism preferences, peak tourism periods, changes in the popularity of scenic spots over time, and consumption patterns of tourists in different regions. Tourists are segmented by combining clustering algorithms to obtain tourism behavior characteristic groups, including: Establish an autoencoder, and use the autoencoder to perform dimensionality reduction processing on the fused feature data. The autoencoder consists of an encoder and a decoder. The encoder compresses the high-dimensional feature data into a low-dimensional potential representation, and the decoder restores the potential representation to approximate original data. By adjusting the parameters of the autoencoder, the essential feature representation of the data is obtained; An analysis model is established based on the GAN generative adversarial network, and the low-dimensional features extracted by the autoencoder are integrated with the characteristic tourism data again and input into the analysis model; the analysis model is used for data expansion and simulation: The GAN generative adversarial network at least includes a generator and a discriminator, the generator is responsible for generating false data, and the discriminator is responsible for distinguishing true and false data; The analysis includes at least analysis of tourists' travel preferences, peak travel periods, changes in popularity of scenic spots over time, and consumption patterns of tourists from different regions. Tourists are segmented using clustering algorithms to obtain groups with travel behavior characteristics.

6. A tourism data analysis method based on deep learning as claimed in claim 1, characterized in that: The analysis model of tourism data is constructed by integrating multiple deep learning algorithms to analyze the characteristic tourism data. The analysis includes at least analysis of tourists' tourism preferences, peak tourism periods, changes in the popularity of scenic spots over time, and consumption patterns of tourists in different regions. The clustering algorithm is combined to segment tourists to obtain tourism behavior characteristic groups, and also includes: The DBSCAN clustering algorithm is optimized by the GWO Grey Wolf Algorithm to obtain the GWO-DBSCAN algorithm; the clustering is based on the characteristics of tourists' travel behavior, travel consumption, and travel preference information extracted from social media. The optimal number of clusters is determined by the elbow rule, and the tourism behavior characteristic groups are input into the GWO-DBSCAN algorithm to obtain different tourist groups, which at least include consumption culture exploration tourists, economical leisure and vacation tourists, family and parent-child tourists, and young backpackers.

7. A tourism data analysis method based on deep learning as claimed in claim 6, characterized in that: The method of optimizing the DBSCAN clustering algorithm by the GWO Grey Wolf Algorithm to obtain the GWO-DBSCAN algorithm further includes: Setting a maximum number of iterations, a population size, and an initial vector value of the GWO algorithm, and generating an initial solution set according to the maximum number of iterations, the population size, and the initial vector value; Substitute the elements in the initial solution set into the DBSCAN algorithm one by one, cluster the original user data set, and calculate and store the silhouette index of each clustering result; Sort the initial solution set and the corresponding silhouette values ​​in ascending order to obtain the top three best solutions and the top three best silhouette values ​​in the initial solution set; Calculate the parameters of the coefficient vector, update the position of each element in the initial solution set, and re-substitute it into the DBSCAN algorithm to cluster the data set and calculate the silhouette value; It is determined whether the maximum number of iterations of the GWO algorithm has been reached. If so, the maximum number of iterations, the population size and the initial vector value are output, and the maximum number of iterations, the population size and the initial vector value are set as initial parameters of the DBSCAN algorithm to obtain the GWO-DBSCAN algorithm.

8. A tourism data analysis system based on deep learning, characterized in that: The tourism data analysis system includes the following modules: A data acquisition module is used to acquire initial tourism data from an IoT device and perform data preprocessing on the initial tourism data to obtain standard tourism data; A feature extraction module, used to extract features from the standard tourism data using a model architecture combining a convolutional neural network and a recurrent neural network to obtain feature tourism data; A data analysis module is used to analyze the characteristic tourism data by building an analysis model of tourism data by integrating multiple deep learning algorithms. The analysis includes at least analysis of tourists' tourism preferences, peak tourism periods, changes in the popularity of scenic spots over time, and consumption patterns of tourists in different regions. The clustering algorithm is used to segment tourists and obtain groups with tourism behavior characteristics. The data display module is used to display the tourism behavior characteristic groups through AR virtual augmented reality and make real-time dynamic adjustments according to user needs.

9. A tourism data analysis method based on deep learning as claimed in claim 8, characterized in that: The data acquisition module includes the following submodules: A screening submodule is used to obtain initial travel data, screen and delete the data based on preset rules, and obtain first initial travel data; A filling submodule is used to remove duplicate records in the first initial tourism data, use a mean filling method for numerical data, and use a text generation model based on a GAN generative adversarial network to fill text data, so as to obtain second initial tourism data; A processing submodule, used for processing the numerical data in the second initial tourism data by normalization and processing the categorical data by One-Hot encoding to obtain the third initial tourism data; An alignment submodule, used to align the travel experience text and picture information posted by tourists on social media in the third initial travel data according to the timestamp, location information in the text and EXIF ​​information of the picture; The fusion submodule is used to use the multimodal fusion technology in deep learning to evaluate the text and related pictures of a certain scenic spot. The attention mechanism will automatically learn the important parts of the description of the scenic spot features in the text and pictures, fuse the important features, and obtain standard tourism data.

10. A tourism data analysis method based on deep learning as claimed in claim 8, characterized in that: The feature extraction module, Includes the following submodules: Establish a submodule for establishing a multi-layer CNN architecture, wherein the multi-layer CNN architecture includes 3-5 convolutional layers, and uses 3x3 convolutional kernels of different sizes in the convolutional layers to extract image features of different scales, such as edges, textures, and shapes; Added a submodule for adding an average pooling layer after the convolutional layer, including a 2x2 max pooling window to reduce the size of the image by half while retaining the most important feature information; The learning submodule is used to process tourists’ movement trajectory data using an RNN with a GRU gated recurrent unit structure. The GRU learns the movement patterns of tourists between different attractions. The extraction submodule is used to fuse the image features extracted by CNN and the sequence data processed by RNN at the input layer, taking the image features as additional input information so that the model can extract features using spatial and temporal information; The weight submodule is used to define a learnable weight matrix according to the importance of the data features to tourism analysis, and obtain the attention score of each feature by multiplying the data features with the attention weight matrix; The analysis submodule is used to perform weighted processing on the original data features according to the attention score, so that the model automatically focuses on the data features with high impact on tourism analysis and obtains characteristic tourism data.

Images