Tourist behavior prediction system and method based on big data

Abstract

The present invention relates to the technical field of big data analysis, and to a tourist behavior prediction system and method based on big data. The system collects relevant photos of a target tourist city from various data sources, and obtains the timestamp of the relevant photos of the target tourist city; pre-processes the collected relevant photos of the target tourist city, thereby classifies the relevant photos of the target tourist city into tourist attractions, and simultaneously analyzes and obtains the rate of change of the number of tourists corresponding to the tourist attractions of the target tourist city; performs saturation quantitative processing on the relevant photos of the tourist attractions corresponding to the target tourist city, and further predicts the intensity of the travel intention of tourists corresponding to the tourist attractions of the target tourist city; and then comprehensively divides the target tourist city into regions, improves the sightseeing experience of tourists and the management efficiency of the scenic spots, can provide a scientific basis for scenic spot management, support data-driven decision-making, and improve the scientificity and effectiveness of management.

Description

Technical Field

[0001] The present invention belongs to the technical field of big data analysis and relates to a tourist behavior prediction system and method based on big data. Background Art

[0002] In recent years, with the rapid development of artificial intelligence, machine learning and other technologies, emerging technical methods such as image recognition, image classification, and image generation have provided good support for the field of tourism research. Social media pictures are an important carrier for tourists to express their tourism perception in a simple, direct and popular way. Studying the content represented by the pictures and the geographic information carried by the pictures is of great significance to understanding the evolution of tourists' perception and behavior of the destination, and provides a new dimension for studying tourist destinations.

[0003] Although the current technologies and methods for predicting tourist behavior have made some progress, there are still some significant defects and drawbacks. First, in terms of data collection and processing, privacy protection and data security issues are becoming increasingly prominent, and tourists may feel uneasy about their personal information being used for behavioral analysis. Second, prediction models often rely on historical data and current trends, which may lead to insufficiently timely or accurate responses to emerging trends and emergencies.

[0004] The current defects and disadvantages of predicting tourist behavior in the future through images mainly include: the timeliness and accuracy of image data are difficult to guarantee, resulting in inaccurate prediction results; in addition, it is difficult to fully capture the behavior patterns of tourists by relying solely on image data, and the lack of comprehensive analysis of tourist and historical behavior data may lead to insufficient generalization ability of the prediction model and inability to effectively respond to emergencies and abnormal behaviors. Summary of the invention

[0005] In view of the above problems existing in the prior art, the present invention provides a tourist behavior prediction system and method based on big data to solve the above technical problems.

[0006] In order to achieve the above purpose and other purposes, the technical solution adopted by the present invention is as follows:

[0007] On one hand, the present invention provides a tourist behavior prediction system based on big data, including a picture collection module, a picture processing module, a picture mining module and a region division module; the above modules are connected by wired and / or wireless connection to achieve data transmission between the modules;

[0008] Picture collection module: collects relevant photos of the target tourist city from various data sources and obtains the timestamps of the relevant photos of the target tourist city;

[0009] Image processing module: pre-processing the collected relevant photos of the target tourist city, thereby classifying the relevant photos of the target tourist city into tourist attractions, and analyzing the change rate of the number of tourists corresponding to each tourist attraction in the target tourist city;

[0010] Image mining module: Quantify the saturation of the relevant photos of each tourist attraction in the target tourist city, and further predict the intensity of tourists' travel intention in each tourist attraction in the target tourist city;

[0011] Regional division module: Comprehensively divide the target tourist cities into regions, where the regions are divided into tourist core areas, tourist secondary core areas and tourist fringe areas.

[0012] Exemplarily, relevant photos of the target tourist city are collected from various data sources. The specific collection process is as follows:

[0013] Filter various image sources within the administrative coordinates of the target tourist city from various data sources and record them as relevant photos;

[0014] In each data source, remove the image sources within the administrative coordinate range of the target tourist city, obtain the text description of each remaining image source in each data source, and extract the text description keywords of each remaining image source in each data source, and then form the text description keyword set E of each remaining image source in each data source. ib , i represents the number of each data source, i=1,2,...j, b represents the number of each remaining image source, b=1,2,...a;

[0015] Synchronously obtain the text description keyword set F corresponding to the preset target tourist city, according to the analysis formula Calculate the matching degree ζ between the text description keyword set of each remaining image source in each data source and the text description keyword set corresponding to the target tourist city ib ;

[0016] The matching degree between the text description keyword set of each remaining picture source in each data source and the text description keyword set corresponding to the target tourist city is compared with the preset text description keyword reference matching degree. If the matching degree between the text description keyword set of a remaining picture source in a data source and the text description keyword set corresponding to the target tourist city is greater than or equal to the preset text description keyword reference matching degree, the remaining picture source in the data source is recorded as a related photo, otherwise the remaining picture source in the data source is removed;

[0017] Based on the above analysis method, relevant photos of the target tourist city are collected.

[0018] For example, the relevant photos of the target tourist city are classified into tourist attractions. The specific classification process is as follows:

[0019] Resize the relevant photos of the target tourist city to the set standard photo size, and normalize the relevant photos of the target tourist city. The specific processing formula is as follows:

[0020]

[0021] Normalization helps the training process of deep learning models because a smaller value range can improve numerical stability and speed up convergence;

[0022] Collect and preprocess the photo datasets of tourist attractions in the target tourist cities. Use the Places365-CNNS model to extract features from the preprocessed photo datasets of tourist attractions in the target tourist cities. Map the extracted features to tourist attraction categories through the fully connected layer of the Places365-CNNS model.

[0023] The normalized relevant photos are imported into the Places365-CNNS model to obtain the distribution probability of the relevant photos of the target tourist city corresponding to each tourist attraction. The specific recognition formula of the Places365-CNNS model can be abstracted as:

[0024] P(y|X)=softmax(W β ×f β-1 (...f2(f1(X))));

[0025] X is the input related image, f1,f2,...f β-1 is the function of the convolutional layer and pooling layer in the Places365-CNNS model, responsible for feature extraction, W β is the weight of the fully connected layer, P(y|X) is the distribution probability of tourist attraction y predicted by the Places365-CNNS model given the input related image X;

[0026] The distribution probabilities of the tourist attractions corresponding to the relevant photos of the target tourist city are arranged in descending order, and then the tourist attractions with the highest distribution probability are selected as the tourist attractions of the relevant photos of the target tourist city, thereby classifying the relevant photos of the target tourist city into tourist attractions.

[0027] For example, the travel intention intensity of tourists in each tourist attraction in the target tourist city is predicted, and the prediction process is as follows:

[0028] Obtain the mean value and variance of pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city;

[0029] The specific calculation process of the mean pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city is as follows: the saturation value of each pixel in each relevant photo of each tourist attraction in each time period of the target tourist city is summed up, and the result of the summation is averaged to obtain the mean pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city;

[0030] The specific calculation process of the pixel saturation variance in each relevant photo of each tourist attraction in each time period of the target tourist city is as follows: the saturation value of each pixel in each relevant photo of each tourist attraction in each time period of the target tourist city is subtracted from the mean value of the pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city, and the result of the subtraction calculation is squared, and the result after the square operation is summed, and the result after the sum operation is divided by the total number of pixels in each relevant photo of each tourist attraction in each time period of the target tourist city -1, so as to obtain the pixel saturation variance in each relevant photo of each tourist attraction in each time period of the target tourist city.

[0031] The pixel saturation mean and pixel saturation variance in each relevant photo of each tourist attraction in each time period of the target tourist city are combined into the feature vector H of each tourist attraction in the target tourist city. c ;

[0032] The training model is then used to predict the travel intention intensity Y of tourists in the target tourist city corresponding to each tourist attraction. c =g(H c ), g is the trained model.

[0033] For example, the target tourist cities are divided into regions in a comprehensive manner, and the specific division process is as follows:

[0034] Combined with the change rate of the number of tourists in each tourist attraction in the target tourist city and the intensity of tourists' travel intention, the expected number of tourists in each tourist attraction in the target tourist city is predicted;

[0035] The estimated number of visitors to each tourist attraction corresponding to the target tourist city is compared with the interval of the number of visitors in each area. If the estimated number of visitors to a tourist attraction corresponding to the target tourist city is within the interval of the number of visitors in the tourist core area, a circle is drawn with the tourist attraction corresponding to the target tourist city as the center and the radius of the preset core area as the radius. The resulting area is recorded as the tourist core area;

[0036] If the estimated number of visitors to a certain tourist attraction in the target tourist city is within the predetermined number range of the tourist secondary core area, a circle is constructed with the target tourist city corresponding to the tourist attraction as the center and the preset secondary core area radiation radius as the radius, and the resulting area is recorded as the tourist secondary core area;

[0037] If the estimated number of visitors to a certain tourist attraction in the target tourist city is within the predetermined number range of the tourist edge area, a circle is constructed with the target tourist city corresponding to the tourist attraction as the center and the preset edge area radiation radius as the radius, and the resulting area is recorded as the tourist edge area;

[0038] The target tourist cities are divided into regions in this way.

[0039] For example, the estimated number of visitors to each tourist attraction in the target tourist city is predicted, and the specific prediction process is as follows:

[0040] Obtain various factors that affect the number of tourists, and number them in sequence as 1, 2, ..., w;

[0041] Arrange the estimated number of tourists in each time period in each tourist attraction of the target tourist city in descending order according to the number of the time period, and obtain the last time period in the time period arrangement, which is recorded as the end time period;

[0042] Using regression model

[0043] , the estimated number of visitors to each tourist attraction in the target tourist city is predicted to be θ c , γ0 is the intercept term, γ1, γ2, γ3, ...γ(w-3) are model parameters, representing the weight of each variable on the prediction of the number of tourists, γ′ is the error term, representing random disturbance, Represents the estimated number of tourists in the target tourist city corresponding to the end time period of the cth tourist attraction, It represents the impact value of the wth changing factor affecting the number of tourists.

[0044] Another aspect of the present invention provides a method for predicting tourist behavior based on big data, comprising:

[0045] Image collection: Collect relevant photos of the target tourist city from various data sources, and obtain the timestamps of the relevant photos of the target tourist city;

[0046] Image processing: pre-process the collected photos of the target tourist city, classify the photos into tourist attractions, and analyze the change rate of the number of tourists in each tourist attraction of the target tourist city;

[0047] Image mining: Quantify the saturation of the relevant photos of each tourist attraction in the target tourist city, and further predict the intensity of tourists' travel intention in each tourist attraction in the target tourist city;

[0048] Regional division: Comprehensively divide the target tourist cities into regions, among which the regions are divided into tourism core areas, tourism secondary core areas and tourism fringe areas.

[0049] As described above, the tourist behavior prediction system and method based on big data provided by the present invention have at least the following beneficial effects:

[0050] The tourist behavior prediction system and method based on big data provided by the present invention collects relevant photos of the target tourist city from various data sources, and obtains the timestamp of the relevant photos corresponding to the target tourist city; pre-processes the collected relevant photos of the target tourist city, thereby classifies the relevant photos of the target tourist city into tourist attractions, and analyzes and obtains the rate of change of the number of tourists corresponding to each tourist attraction in the target tourist city; saturation quantization is performed on the relevant photos of each tourist attraction corresponding to the target tourist city, and further predicts the intensity of the tourists' travel intention corresponding to each tourist attraction in the target tourist city; and then comprehensively divides the target tourist city into regions, thereby providing more accurate tourist behavior prediction and scenic area management information. By combining timestamps and photos, the rate of change of the number of tourists can be dynamically monitored, helping scenic area managers to adjust resource allocation and management strategies in a timely manner to avoid overcrowding or waste of resources; by quantifying the saturation of photos, the intensity of tourists' travel intention can be better predicted, helping scenic area managers to understand tourists' interests and preferences, thereby optimizing the services and facilities of the scenic area; by comprehensively analyzing the changes in the number of tourists and travel intentions in each scenic area, scientific regional divisions can be made, the layout of scenic areas can be optimized, and the tourists' sightseeing experience and the management efficiency of scenic areas can be improved; it can provide a scientific basis for scenic area management, support data-driven decision-making, and improve the scientificity and effectiveness of management; through accurate prediction and management, the overall tourism experience can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0052] Figure 1 It is a schematic diagram of the connection of various modules of the system of the present invention.

[0053] Figure 2 It is a schematic diagram of the connection of each step of the method of the present invention. DETAILED DESCRIPTION

[0054] The above contents in combination with the implementation of the present invention are merely examples and explanations of the concept of the present invention. The technical personnel in the relevant technical field may make various modifications or supplements to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the claims, they shall all fall within the protection scope of the present invention.

[0055] Example 1

[0056] See also Figure 1 As shown, the tourist behavior prediction system based on big data includes a picture collection module, a picture processing module, a picture mining module and a region division module; the above modules are connected by wired and / or wireless connection to realize data transmission between the modules;

[0057] Picture collection module: collects relevant photos of the target tourist city from various data sources and obtains the timestamps of the relevant photos of the target tourist city;

[0058] The relevant photos include promotional photos and tourist photos, and the data sources include but are not limited to travel websites, social media platforms and official websites of tourist attractions;

[0059] It should be added that the Python crawler technology is used to call Flickr's API (Application Programming Interface), and the longitude and latitude range of the target tourist city is used as the constraint condition of the bbox. Since Flickr's API limits the number of data returned to 5,000 per request for personal accounts, in order to ensure the integrity of the data, the period from 2010-01-01 to 2019-12-31 is divided into days, and the timestamps of the start and end times of each day are used as the constraints of the min taken date and max taken date. After the collection is completed, the data is counted and summarized by day. If the summary data is greater than 5,000, the time constraints within the time range are divided twice, and so on, calling the flickr.photos.search method. Finally, we obtained the source data of 1,308,908 pictures within the coordinate range shared by tourists on the social media website Flickr from 2010 to 2019, including picture ID (id), user name (ownername), user ID (owner), longitude (latitude), latitude (longitude), shooting time (datetaken), theme (title), tags (tags), and the secret key (server, secret) required for picture downloading, etc. Among them, the user ID is used to screen inbound tourists, the longitude and latitude are used to distinguish regional correlation and analyze the changes in tourists' perception in different areas of the study area, the shooting time is used for the temporal tracking of tourists' footprints, and the secret key required for the download link is to download the link information of the picture data according to the picture ID combination, and further analyze the content represented by the picture.

[0060] Due to the limitations of Flickr, and the fact that inbound tourists can more objectively reflect the international influence of the target tourist city and avoid the impact of special periods on inbound tourism, and the tourism industry in the target tourist city requires coordinated development among various provinces and regions, the research data selected pictures taken by inbound tourists in the target tourist city from 2010 to 2019.

[0061] Based on the above embodiment, relevant photos of the target tourist city are collected from various data sources. The specific collection process is as follows:

[0062] Filter various image sources within the administrative coordinates of the target tourist city from various data sources and record them as relevant photos;

[0063] In each data source, remove the image sources within the administrative coordinate range of the target tourist city, obtain the text description of each remaining image source in each data source, and extract the text description keywords of each remaining image source in each data source, and then form the text description keyword set E of each remaining image source in each data source. ib , i represents the number of each data source, i=1,2,...j, b represents the number of each remaining image source, b=1,2,...a;

[0064] Synchronously obtain the text description keyword set F corresponding to the preset target tourist city, according to the analysis formula Calculate the matching degree ζ between the text description keyword set of each remaining image source in each data source and the text description keyword set corresponding to the target tourist city ib ;

[0065] The matching degree between the text description keyword set of each remaining picture source in each data source and the text description keyword set corresponding to the target tourist city is compared with the preset text description keyword reference matching degree. If the matching degree between the text description keyword set of a remaining picture source in a data source and the text description keyword set corresponding to the target tourist city is greater than or equal to the preset text description keyword reference matching degree, the remaining picture source in the data source is recorded as a related photo, otherwise the remaining picture source in the data source is removed;

[0066] Based on the above analysis method, relevant photos of the target tourist city are collected.

[0067] Image processing module: pre-processing the collected relevant photos of the target tourist city, thereby classifying the relevant photos of the target tourist city into tourist attractions, and analyzing the change rate of the number of tourists corresponding to each tourist attraction in the target tourist city;

[0068] On the basis of the above embodiment, the relevant photos of the target tourist city are classified into tourist attractions. The specific classification process is as follows:

[0069] Resize the relevant photos of the target tourist city to the set standard photo size, and normalize the relevant photos of the target tourist city. The specific processing formula is as follows:

[0070]

[0071] This helps the training process of deep learning models because a smaller value range can improve numerical stability and speed up convergence;

[0072] Collect and preprocess the photo datasets of tourist attractions in the target tourist cities. Use the Places365-CNNS model to extract features from the preprocessed photo datasets of tourist attractions in the target tourist cities. Map the extracted features to tourist attraction categories through the fully connected layer of the Places365-CNNS model.

[0073] Convolutional Neural Network (CNN) is a deep learning model used for processing two-dimensional or three-dimensional data such as images, voice and video. Its core idea is to use convolution operations and pooling operations to extract features and reduce dimensions of data, so as to achieve tasks such as image recognition, voice recognition, and target detection. It mainly consists of four parts: convolutional layer, pooling layer, activation function, and fully connected layer. CNN can optimize model parameters through the back propagation algorithm so that it can better fit the training data and obtain better generalization ability on the test data. In recent years, CNN has achieved great success in image recognition, target detection, natural language processing and other fields, and has become one of the important research directions in the field of deep learning. This study uses the deep learning model of Places-CNNS scene recognition in 2017. Places365-CNNS is a convolutional neural network model for scene recognition and visual general deep scene features trained on the open source database Places. The design of the Places dataset follows the principles of human visual cognition. Its goal is to build a core of visual knowledge that can be used to train artificial systems to perform high-level visual understanding tasks such as scene context, object recognition, action and event prediction, and theory of mind reasoning. The semantic categories of Places are labels representing environments, and the dataset has different categories of bedrooms or streets, etc., because they behave differently and have different predictions about what might happen next. Places contains over 10 million images in total, including more than 400 unique scene categories. The dataset has 5,000 to 30,000 training images per class, consistent with the frequency of occurrence in the real world. Using convolutional neural networks (CNNs), the Places dataset allows deep scene features to be learned for a variety of scene recognition tasks, with the goal of establishing new state-of-the-art performance on scene-centric benchmarks.

[0074] The normalized relevant photos are imported into the Places365-CNNS model to obtain the distribution probability of the relevant photos of the target tourist city corresponding to each tourist attraction. The specific recognition formula of the Places365-CNNS model can be abstracted as:

[0075] P(y|X)=softmax(W β ×f β-1 (...f2(f1(X))));

[0076] X is the input related image, f1,f2,...f β-1 is the function of the convolutional layer and pooling layer in the Places365-CNNS model, responsible for feature extraction, W β is the weight of the fully connected layer, P(y|X) is the distribution probability of tourist attraction y predicted by the Places365-CNNS model given the input related image X;

[0077] The specific implementation code is as follows:

[0078]

[0079]

[0080]

[0081]

[0082] The load_model() function downloads and loads the weights of the Places365-CNN model.

[0083] Image preprocessing: The preprocess_image() function normalizes the input image to meet the input requirements of the model.

[0084] Inference: The infer_image() function performs inference on the preprocessed image to obtain the probability distribution of each category.

[0085] Get category labels: The load_labels() function loads the category label file of Places365.

[0086] Main function: The main() function combines the above steps, processes the input image and outputs the top prediction result and its probability.

[0087] The distribution probabilities of the tourist attractions corresponding to the relevant photos of the target tourist city are arranged in descending order, and then the tourist attractions with the highest distribution probability are selected as the tourist attractions of the relevant photos of the target tourist city, thereby classifying the relevant photos of the target tourist city into tourist attractions.

[0088] Image mining module: Quantify the saturation of the relevant photos of each tourist attraction in the target tourist city, and further predict the intensity of tourists' travel intention in each tourist attraction in the target tourist city;

[0089] Based on the above embodiment, the travel intention intensity of tourists corresponding to each tourist attraction in the target tourist city is predicted, and the prediction process is as follows:

[0090] Obtain the mean value and variance of pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city;

[0091] The specific calculation process of the mean pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city is as follows: the saturation value of each pixel in each relevant photo of each tourist attraction in each time period of the target tourist city is summed up, and the result of the summation is averaged to obtain the mean pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city;

[0092] The specific calculation process of the pixel saturation variance in each relevant photo of each tourist attraction in each time period of the target tourist city is as follows: the saturation value of each pixel in each relevant photo of each tourist attraction in each time period of the target tourist city is subtracted from the mean value of the pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city, and the result of the subtraction calculation is squared, and the result after the square operation is summed, and the result after the sum operation is divided by the total number of pixels in each relevant photo of each tourist attraction in each time period of the target tourist city -1, so as to obtain the pixel saturation variance in each relevant photo of each tourist attraction in each time period of the target tourist city.

[0093] The pixel saturation mean and pixel saturation variance in each relevant photo of each tourist attraction in each time period of the target tourist city are combined into the feature vector H of each tourist attraction in the target tourist city. c ;

[0094] The training model is then used to predict the travel intention intensity Y of tourists in the target tourist city corresponding to each tourist attraction. c =g(H c ), g is the trained model.

[0095] It should be added that the specific process of training the model is:

[0096] Image data: Collect a large number of images of tourist attractions in the target tourist cities, ensuring that the images have different saturation levels;

[0097] Tourist behavior data: Collect data on tourists’ travel intentions, which can be obtained through questionnaire surveys, social media data, and click-through rates of travel websites;

[0098] Image processing: Use image processing tools (such as OpenCV, PIL, etc.) to extract the saturation data of the image. Calculate the average saturation value of each image;

[0099] Data cleaning: Process missing values ​​and outliers to ensure data quality.

[0100] Feature engineering: Match the saturation data of the image with the tourists’ travel intention data to form a training data set.

[0101] Data exploration: Use statistical analysis tools (such as Pandas, NumPy, etc.) to conduct preliminary exploration of the data to understand the distribution, correlation, etc. of the data.

[0102] Data visualization: Use visualization tools (such as Matplotlib, seaborn, etc.) to draw scatter plots, histograms, heat maps, etc. to intuitively show the relationship between image saturation and tourists' travel intentions

[0103] Model selection: If travel intention is a continuous variable, you can choose linear regression, ridge regression, Lasso regression, etc. If travel intention is a discrete variable (such as "willing" and "unwilling"), you can choose logistic regression, random forest, support vector machine (SVM), etc.

[0104] Data splitting: Divide the dataset into training and testing sets (e.g. 80% training, 20% testing).

[0105] Model training: Use the training set to train the model. This can be achieved using the machine learning library. The specific code is shown below:

[0106]

[0107]

[0108] Regional division module: Comprehensively divide the target tourist cities into regions, where the regions are divided into tourist core areas, tourist secondary core areas and tourist fringe areas.

[0109] Based on the above embodiment, the target tourist city is divided into regions in a comprehensive manner. The specific division process is as follows:

[0110] Combined with the change rate of the number of tourists in each tourist attraction in the target tourist city and the intensity of tourists' travel intention, the expected number of tourists in each tourist attraction in the target tourist city is predicted;

[0111] The estimated number of visitors to each tourist attraction corresponding to the target tourist city is compared with the interval of the number of visitors in each area. If the estimated number of visitors to a tourist attraction corresponding to the target tourist city is within the interval of the number of visitors in the tourist core area, a circle is drawn with the tourist attraction corresponding to the target tourist city as the center and the radius of the preset core area as the radius. The resulting area is recorded as the tourist core area;

[0112] If the estimated number of visitors to a certain tourist attraction in the target tourist city is within the predetermined number range of the tourist secondary core area, a circle is constructed with the target tourist city corresponding to the tourist attraction as the center and the preset secondary core area radiation radius as the radius, and the resulting area is recorded as the tourist secondary core area;

[0113] If the estimated number of visitors to a certain tourist attraction in the target tourist city is within the predetermined number range of the tourist edge area, a circle is constructed with the target tourist city corresponding to the tourist attraction as the center and the preset edge area radiation radius as the radius, and the resulting area is recorded as the tourist edge area;

[0114] The target tourist cities are divided into regions in this way.

[0115] Based on the above embodiment, the estimated number of visitors to each tourist attraction in the target tourist city is predicted. The specific prediction process is as follows:

[0116] Obtain various factors that affect the number of tourists, and number the factors that affect the number of tourists as 1, 2, ..., w in sequence; the factors include but are not limited to seasonal factors, weather conditions, socioeconomic factors, transportation convenience, scenic spot attractiveness, social media popularity or travel booking data;

[0117] Arrange the estimated number of tourists in each time period in each tourist attraction of the target tourist city in descending order according to the number of the time period, and obtain the last time period in the time period arrangement, which is recorded as the end time period;

[0118] Using regression model , the estimated number of visitors to each tourist attraction in the target tourist city is predicted to be θ c , γ0 is the intercept term, γ1, γ2, γ3, ...γ(w-3) are model parameters, representing the weight of each variable on the prediction of the number of tourists, γ′ is the error term, representing random disturbance, Represents the estimated number of tourists in the target tourist city corresponding to the end time period of the cth tourist attraction, It represents the impact value of the wth changing factor affecting the number of tourists.

[0119] Example 2

[0120] See also Figure 2 As shown in the figure, the tourist behavior prediction method based on big data includes:

[0121] Image collection: Collect relevant photos of the target tourist city from various data sources, and obtain the timestamps of the relevant photos of the target tourist city;

[0122] Image processing: pre-process the collected photos of the target tourist city, classify the photos into tourist attractions, and analyze the change rate of the number of tourists in each tourist attraction of the target tourist city;

[0123] Image mining: Quantify the saturation of the relevant photos of each tourist attraction in the target tourist city, and further predict the intensity of tourists' travel intention in each tourist attraction in the target tourist city;

[0124] Regional division: Comprehensively divide the target tourist cities into regions, among which the regions are divided into tourism core areas, tourism secondary core areas and tourism fringe areas.

[0125] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0126] It should be understood that determining B based on A does not mean determining B only based on A. B can also be determined based on A and / or other information.

[0127] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0128] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Tourist behavior prediction system based on big data, characterized by: It includes an image collection module, an image processing module, an image mining module and an area division module; the above modules are connected by wired and / or wireless connection to achieve data transmission between the modules; Picture collection module: collects relevant photos of the target tourist city from various data sources and obtains the timestamps of the relevant photos of the target tourist city; Collect relevant photos of the target tourist city from various data sources. The specific collection process is as follows: Filter various image sources within the administrative coordinates of the target tourist city from various data sources and record them as relevant photos; Eliminate the image sources within the administrative coordinates of the target tourist city from each data source, obtain the text descriptions of the remaining image sources in each data source, and extract the text description keywords of the remaining image sources in each data source, and then form a text description keyword set of the remaining image sources in each data source. , i represents the number of each data source, i=1,2,...j, b represents the number of each remaining image source, b=1,2,...a; Synchronously obtain the text description keyword set F corresponding to the preset target tourist city, according to the analysis formula , calculate the matching degree between the text description keyword set of each remaining image source in each data source and the text description keyword set corresponding to the target tourist city ; The matching degree between the text description keyword set of each remaining picture source in each data source and the text description keyword set corresponding to the target tourist city is compared with the preset text description keyword reference matching degree. If the matching degree between the text description keyword set of a remaining picture source in a data source and the text description keyword set corresponding to the target tourist city is greater than or equal to the preset text description keyword reference matching degree, the remaining picture source in the data source is recorded as a related photo, otherwise the remaining picture source in the data source is removed; Then, relevant photos of the target tourist city are collected; Image processing module: pre-processing the collected relevant photos of the target tourist city, thereby classifying the relevant photos of the target tourist city into tourist attractions, and analyzing the change rate of the number of tourists in each tourist attraction of the target tourist city; Classify the relevant photos of the target tourist city into tourist attractions. The specific classification process is as follows: Resize the relevant photos of the target tourist city to the set standard photo size, and normalize the relevant photos of the target tourist city. The specific processing formula is as follows: Collect and preprocess the photo datasets of tourist attractions in the target tourist cities. Use the Places365-CNNS model to extract features from the preprocessed photo datasets of tourist attractions in the target tourist cities. Map the extracted features to tourist attraction categories through the fully connected layer of the Places365-CNNS model. The normalized relevant photos are imported into the Places365-CNNS model to obtain the distribution probability of the relevant photos of the target tourist city corresponding to each tourist attraction. The specific recognition formula of the Places365-CNNS model is abstracted as follows: ； X is the input related image, It is the function of the convolutional layer and pooling layer in the Places365-CNNS model, responsible for feature extraction. is the weight of the fully connected layer, is the distribution probability of tourist attraction y predicted by the Places365-CNNS model given the input related image X; Arrange the distribution probabilities of the tourist attractions corresponding to the relevant photos of the target tourist city in descending order, and then select the tourist attractions with the highest distribution probability as the tourist attractions of the relevant photos of the target tourist city, thereby classifying the relevant photos of the target tourist city into tourist attractions; The change rate of the number of tourists in each tourist attraction in the target tourist city is obtained through analysis. The analysis process is as follows: According to the classification categories of tourist attractions corresponding to the target tourist city, obtain the timestamps of the relevant photos in the tourist attractions corresponding to the target tourist city; Divide the time periods into time periods according to the timestamps of the relevant photos in the tourist attractions corresponding to the target tourist city according to the preset timestamp intervals, thereby obtaining the relevant photos of the relevant photos in the tourist attractions corresponding to the target tourist city in each time period; Through spatial analysis, the estimated density of tourists in each tourist attraction in each time period in the target tourist city is obtained. , where s is the number of each related photo, s=1,2,...N, h is the bandwidth, K is the kernel function, is the total number of relevant photos in the vth time period in the cth tourist attraction of the target tourist city, c is the number of each tourist attraction, v is the number of each time period, v=1,2,...p, is the geographic coordinates of the sth related photo, The geographical coordinates of the cth tourist attraction corresponding to the target tourist city; Get the area of ​​each tourist attraction corresponding to the target tourist city , calculate the estimated number of tourists in each tourist attraction in each time period in the target tourist city ; For tourist attraction c, in time periods v and The change in the number of tourists between , is the preset timestamp interval; The change rate of the number of tourists in each tourist attraction in the target tourist city is calculated in this way. ; Image mining module: Quantify the saturation of the relevant photos of each tourist attraction in the target tourist city, and further predict the intensity of tourists' travel intention in each tourist attraction in the target tourist city; The prediction process is as follows: Obtain the mean value and variance of pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city; The specific calculation process of the mean pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city is as follows: the saturation value of each pixel in each relevant photo of each tourist attraction in each time period of the target tourist city is summed up, and the result of the summation is averaged to obtain the mean pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city; The specific calculation process of the pixel saturation variance in each relevant photo of each tourist attraction in each time period of the target tourist city is as follows: the saturation value of each pixel in each relevant photo of each tourist attraction in each time period of the target tourist city is calculated by difference with the mean value of the pixel saturation in each relevant photo of each tourist attraction in each time period of the target tourist city, and the result of the difference calculation is squared, and the result after the square calculation is summed, and the result after the sum calculation is divided by the total number of each pixel in each relevant photo of each tourist attraction in each time period of the target tourist city -1, to obtain the pixel saturation variance in each relevant photo of each tourist attraction in each time period of the target tourist city; The pixel saturation mean and pixel saturation variance in each relevant photo of each tourist attraction in each time period of the target tourist city are combined into the feature vector of each tourist attraction in the target tourist city. ; Train the model, and then use the model to predict the intensity of tourists' travel intentions in the target tourist city corresponding to each tourist attraction , g is the trained model; Regional division module: comprehensively divide the target tourist city into regions, where the regions are divided into tourist core regions, tourist secondary core regions and tourist marginal regions; The target tourist cities are divided into regions in a comprehensive manner. The specific division process is as follows: Combined with the change rate of the number of tourists in each tourist attraction in the target tourist city and the intensity of tourists' travel intention, the expected number of tourists in each tourist attraction in the target tourist city is predicted; The estimated number of visitors to each tourist attraction corresponding to the target tourist city is compared with the interval of the number of visitors in each area. If the estimated number of visitors to a tourist attraction corresponding to the target tourist city is within the interval of the number of visitors in the tourist core area, a circle is drawn with the tourist attraction corresponding to the target tourist city as the center and the radius of the preset core area as the radius. The resulting area is recorded as the tourist core area; If the estimated number of visitors to a certain tourist attraction in the target tourist city is within the predetermined number range of the tourist secondary core area, a circle is constructed with the target tourist city corresponding to the tourist attraction as the center and the preset secondary core area radiation radius as the radius, and the resulting area is recorded as the tourist secondary core area; If the estimated number of visitors to a certain tourist attraction in the target tourist city is within the predetermined number range of the tourist edge area, a circle is constructed with the target tourist city corresponding to the tourist attraction as the center and the preset edge area radiation radius as the radius, and the resulting area is recorded as the tourist edge area; Use this division method to divide the target tourist cities into regions; The estimated number of visitors to each tourist attraction in the target tourist city is predicted. The specific prediction process is as follows: Obtain various factors that affect the number of tourists, and number the factors that affect the number of tourists as 1, 2, ... w in sequence; the factors include but are not limited to seasonal factors, weather conditions, socio-economic factors, transportation convenience, scenic spot attractiveness, social media popularity or travel booking data; Arrange the estimated number of tourists in each time period in each tourist attraction of the target tourist city in descending order according to the number of the time period, and obtain the last time period in the time period arrangement, which is recorded as the end time period; Using regression model , predict the expected number of visitors to each tourist attraction in the target tourist city , is the intercept term, are model parameters, representing the weight of each variable on the prediction of tourist numbers, is the error term, representing random disturbance, Represents the estimated number of tourists in the target tourist city corresponding to the end time period of the cth tourist attraction, It represents the impact value of the wth changing factor affecting the number of tourists.

2. A tourist behavior prediction method based on big data, characterized by: The tourist behavior prediction system based on big data described in claim 1 is implemented, comprising: Image collection: Collect relevant photos of the target tourist city from various data sources, and obtain the timestamps of the relevant photos of the target tourist city; Image processing: pre-process the collected photos of the target tourist city, classify the photos into tourist attractions, and analyze the change rate of the number of tourists in each tourist attraction of the target tourist city; Image mining: Quantify the saturation of the relevant photos of each tourist attraction in the target tourist city, and further predict the intensity of tourists' travel intention in each tourist attraction in the target tourist city; Regional division: Comprehensively divide the target tourist cities into regions, among which the regions are divided into tourism core areas, tourism secondary core areas and tourism fringe areas.

3. An electronic device, characterized in that , comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the visitor behavior prediction system based on big data as claimed in claim 1 by calling the computer program stored in the memory.

4. A computer program product stored on a computer readable medium, characterized in that : It includes a computer-readable program, which, when executed on an electronic device, provides a user input interface to implement the tourist behavior prediction system based on big data as described in claim 1.

Images