Tourism trend prediction method and system based on multi-source data fusion
Through multi-source data fusion and feature engineering, the accuracy of tourism trend prediction is improved, and the shortcomings of existing models in handling short-term emergencies and weather changes are solved. Combined with the LSTM model to predict, a more comprehensive tourism trend prediction is achieved. .
Patent Information
- Application Number
- CN202411557831.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-04
- Publication Date
- 2024-11-29
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing time series models perform poorly in dealing with tourism trend forecasts, especially short-term emergencies and weather changes, and lack of external factor integration and insufficient data feature extraction, resulting in low prediction accuracy.
Using multi-source data fusion method, the number of historical tourists, hotel reservation and unsubscribe data, meteorological data and holiday information are collected, and the characteristics of daily reservation, unsubscribe, proportion, cumulative and lag are extracted through feature engineering, and Training is performed through the LSTM model, combining grid search and tuning parameters, and forming a time series data set for prediction.
It significantly improves the accuracy and reliability of short-term tourism trend forecasts, enhances the model's perception of emergencies, provides real-time and accurate prediction results, and provides support for tourism management decisions.
Smart Images

Figure CN119047668A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of big data processing, and in particular relates to a tourism trend prediction method and system based on multi-source data fusion. Background Art
[0002] Existing technologies usually use long short-term memory network (LSTM) models or seasonal autoregressive integrated moving average (SARIMA) models for tourism trend forecasting. These models perform well in capturing the trend and periodicity of time series data, however, they have significant limitations in dealing with short-term emergencies.
[0003] The dynamics and complexity of the tourism industry make it difficult to accurately predict the number of tourists by relying solely on traditional time series models. In order to attract tourists, many cities hold various large-scale events regularly or irregularly, such as festivals, concerts, sports events, etc., which will significantly increase the number of tourists in the short term. In addition, weather conditions (such as heavy rain, heavy snow, typhoons, etc.) will also have a significant impact on tourists' travel decisions. Short-term emergencies and weather changes can cause drastic fluctuations in the number of tourists, and traditional LSTM models and seasonal models have the following shortcomings in capturing and processing these short-term changes: Insufficient ability to handle emergencies: LSTM and SARIMA models mainly rely on long-term trends and seasonal patterns in historical data. For emergencies that have never occurred before, these models are difficult to adapt to and predict their impact. This significantly reduces the prediction effect of the model when facing large-scale events or extreme weather.
[0004] Low short-term forecasting accuracy: Although the LSTM model performs well in capturing long-term dependencies, its performance in short-term forecasting is often affected by unexpected events that have not appeared in historical data. Although the SARIMA model can handle seasonal fluctuations, its ability to respond to non-periodic emergencies is limited, resulting in low short-term forecasting accuracy.
[0005] Lack of external factor integration: Existing models usually rely only on time series data itself, while ignoring the impact of external factors (such as meteorological data, social media data, holiday information, etc.) on tourism trends. External factors are often important drivers of emergencies, and incorporating them into the model can significantly improve forecast accuracy.
[0006] Insufficient data feature extraction: When processing data, LSTM and SARIMA models usually only use the original time series data, and fail to fully extract and utilize the various features in the data (such as booking and cancellation volume, cumulative booking volume, lag features, etc.). These features can provide more information about tourist behavior patterns and help the model better predict short-term changes.
[0007] Feature Engineering refers to extracting useful features from raw data and processing them so that the model can better understand and utilize these features, thereby improving the accuracy of prediction or classification. Feature engineering is a very critical step in the process of machine learning and data mining, including steps such as feature extraction, feature selection, feature transformation, and feature construction.
[0008] Feature engineering can be used to construct and capture complex relationships in data and improve the expressiveness of the model. However, there is little public disclosure in the existing technology on improving the existing model training process by performing feature engineering on tourism-related multi-source data. Summary of the invention
[0009] In order to solve the problems in the background technology, a tourism trend prediction method and system based on multi-source data fusion are proposed.
[0010] In one embodiment, the present invention proposes a tourism trend prediction method based on multi-source data fusion, comprising the following steps: Collect multi-source data related to tourism, the multi-source data at least including: historical number of tourists, number of hotel reservations and cancellations, number of non-refundable hotels, number of ticket reservations and cancellations, number of non-refundable tickets, meteorological data, and holiday types; Extract features from booking and cancellation data, including: daily bookings and cancellations, booking and cancellation ratio, non-refundable bookings, cumulative bookings and cancellations in the past 7 days and 30 days; lagged bookings and cancellations in the previous 1 day, 3 days and 7 days, to obtain a first feature data set; fuse the first feature data set with non-first feature data sets in the multi-source data, and remove redundant or irrelevant features to form a time series data set; use the time series data set to train an LSTM model, perform parameter tuning through a grid search method, and determine the optimal model parameters; use the trained model to perform short-term predictions to predict tourism trends in the future.
[0011] In another embodiment, the present invention also proposes a tourism trend prediction system based on multi-source data fusion, and includes the following modules: A collection module, used to collect multi-source data related to tourism, wherein the multi-source data includes at least: the number of historical tourists, the number of hotel reservations and cancellations, the number of non-refundable hotels, the number of ticket reservations and cancellations, the number of non-refundable tickets, meteorological data, and holiday types; The feature engineering module is used to extract features from the booking and cancellation data, including: daily booking and cancellation volume, booking and cancellation ratio, non-refundable volume, cumulative booking and cancellation volume in the past 7 days and 30 days; lagged booking and cancellation volume in the previous 1 day, 3 days and 7 days, to obtain the first feature data set; A fusion module, used to fuse the first feature data set with non-first feature data sets in the multi-source data, and remove redundant or irrelevant features to form a time series data set; A training module is used to train the LSTM model using the time series data set, and to perform parameter tuning through a grid search method to determine the optimal model parameters; The prediction module is used to make short-term predictions using the trained model to predict tourism trends in the future.
[0012] The present invention can produce the following beneficial effects through the above technical solution: By extracting and utilizing multiple features (such as daily bookings and cancellations, booking-to-cancellation ratio, cumulative bookings and cancellations, and lagged bookings and cancellations), the model can more comprehensively capture short-term changes in tourist behavior and improve the accuracy of short-term forecasts.
[0013] By collecting external factors such as meteorological data and holiday types and incorporating them into the feature data set, the model’s ability to perceive unexpected events is enhanced. For example, data on large-scale events and extreme weather can help the model better capture and predict short-term dramatic fluctuations in the number of tourists.
[0014] Through systematic feature engineering steps, we extract features such as daily bookings and cancellations, booking and cancellation ratios, non-refundable bookings, cumulative bookings and cancellations, and delayed bookings and cancellations from booking and cancellation data to form a feature data set. Then, we perform feature selection and standardization to optimize data quality and model performance.
[0015] Through the data fusion step, multi-source data such as the historical number of tourists, hotel bookings and cancellations, the number of non-refundable hotels and tickets, meteorological data and holiday types are integrated to provide more comprehensive background information and help the model better understand and predict tourism trends.
[0016] This solution significantly improves the accuracy and reliability of tourism trend forecasting through systematic data collection, feature engineering, and data fusion. By dealing with short-term emergencies and external factors, optimizing model structure and hyperparameter configuration, it provides real-time and accurate tourism trend forecasting results, providing strong support for tourism management and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0018] Figure 1 is a flow chart of the method of the present invention. DETAILED DESCRIPTION
[0019] Below, the invention is preferably described in conjunction with the accompanying drawings and specific implementation methods.
[0020] This embodiment solves the above problem through the following steps: In one embodiment, reference Figure 1 The present invention provides a tourism trend prediction method based on multi-source data fusion, comprising the following steps: Collect multi-source data related to tourism, the multi-source data at least including: historical number of tourists, number of hotel reservations and cancellations, number of non-refundable hotels, number of ticket reservations and cancellations, number of non-refundable tickets, meteorological data, and holiday types.
[0021] In order to improve the accuracy and reliability of tourism trend prediction, it is necessary to collect multi-source data that can fully reflect the various influencing factors of tourism activities. The multi-source data includes at least the following categories: Historical number of tourists: Historical number of tourists refers to the total number of tourists who visited a certain scenic spot or tourist destination in a certain period of time in the past. These data are usually recorded by day, week or month. By analyzing the historical number of tourists data, we can identify the peak and off-season of tourism, and help the model capture the long-term trend and seasonal changes in the time series. This is of great reference value for predicting the number of tourists in the future.
[0022] Hotel booking and cancellation numbers: Hotel booking numbers refer to the total number of times tourists book hotels within a certain period of time, while cancellation numbers refer to the total number of times they cancel reservations. These data record the status of refundable and non-refundable hotel reservations. Hotel booking and cancellation data reflect tourists' travel plans and actual travel situations. Through these data, tourists' booking behaviors and changing trends can be analyzed, and the number of tourists in the short term can be predicted.
[0023] Number of non-refundable hotels: The number of non-refundable hotels refers to the total number of times tourists have booked non-refundable hotels within a certain period of time. A higher number of non-refundable hotels indicates that tourists are more confident about their trips and reflects their certainty of plans. This is important for predicting changes in the number of tourists in the short term, because non-refundable bookings usually better reflect actual travel intentions.
[0024] Ticket booking and cancellation numbers: Ticket booking numbers refer to the total number of times tourists booked tickets for attractions within a certain period of time, while cancellation numbers refer to the total number of times ticket bookings were canceled. These data record refundable and non-refundable ticket bookings respectively. Ticket booking and cancellation data are similar to hotel booking data and can reflect tourists' travel plans and actual behaviors. Through these data, we can analyze tourists' interest and demand changes for specific attractions, thereby more accurately predicting tourism trends.
[0025] Number of non-refundable tickets: The number of non-refundable tickets refers to the total number of times tourists have booked non-refundable tickets for attractions within a certain period of time. The number of non-refundable tickets reflects the tourists' certainty and confidence in their itineraries. By analyzing the number of non-refundable tickets, we can further understand tourists' travel intentions and the certainty of their plans, and help predict changes in the number of tourists in the short term.
[0026] Meteorological data: Meteorological data includes records of weather conditions such as temperature, precipitation, wind speed, humidity, etc. Weather conditions are an important factor affecting tourists' travel decisions. Through meteorological data, it is possible to analyze the impact of weather changes on the number of tourists, especially the short-term impact of extreme weather events (such as heavy rain, heavy snow, typhoons, etc.) on tourism activities. This helps to improve the accuracy of forecasts, especially in the short term.
[0027] Holiday type: Holiday types include information about statutory holidays, major events, and local festivals. Holidays and major events usually have a significant impact on the number of tourists. By collecting holiday type data, the model can be helped to identify the impact of these special dates on the number of tourists and improve the accuracy of the forecast. In addition, this information can help identify potential peak tourist seasons and allocate and manage resources accordingly.
[0028] Among the above data, historical tourist number data can be obtained from official agencies such as tourism management departments and statistics bureaus. Hotel-related data can be obtained directly from the hotel's reservation system or third-party platforms such as Alitrip and Qunar. Cooperate with ticket sales platforms (such as Mafengwo and Ctrip) to obtain ticket reservation and cancellation data.
[0029] Features are extracted from the booking and cancellation data, including: daily booking and cancellation volume, booking and cancellation ratio, non-refundable volume, cumulative booking and cancellation volume in the past 7 days and 30 days; lagged booking and cancellation volume in the previous 1 day, 3 days and 7 days, to obtain the first feature data set.
[0030] Extracting features from booking and cancellation data can provide more useful information for the model and help improve the accuracy and reliability of tourism trend predictions.
[0031] Daily bookings and cancellations: Daily bookings refer to the total number of hotel or attraction bookings on a given day. Daily cancellations refer to the total number of hotel or attraction cancellations on a given day. Directly reflects daily bookings and cancellations, helping the model capture tourists' booking behaviors and changing trends. It can be used to identify short-term fluctuations, such as a surge in bookings or cancellations before and after an event.
[0032] Booking-to-Cancellation Ratio: The booking-to-cancellation ratio is the ratio of the number of daily bookings to the number of cancellations. It reflects the relative changes in bookings and cancellations, helping the model understand the stability or volatility of tourist behavior. Changes in the booking-to-cancellation ratio during a specific period (such as before and after a large event) can indicate whether the model needs to adjust its prediction of the number of tourists.
[0033] Non-refundable bookings: refers to the total number of non-refundable hotel or ticket bookings on a certain day. A higher non-refundable booking volume indicates that tourists are more confident about their trips and reflects their deterministic plans. It helps the model identify more stable and certain travel demands, thereby improving the accuracy of forecasts.
[0034] Cumulative bookings and cancellations: Cumulative bookings in the past 7 days: The sum of all bookings in the past 7 days.
[0035] Cumulative bookings in the past 30 days: The sum of all bookings in the past 30 days.
[0036] Cumulative cancellations in the past 7 days: The sum of all cancellations in the past 7 days.
[0037] Cumulative cancellations in the past 30 days: The sum of all cancellations in the past 30 days.
[0038] Providing booking and cancellation trends in the short and medium term helps the model capture cyclical and trend characteristics. The changes in cumulative volume can reveal some potential trends, such as continued growth or decline, and provide a long-term perspective for the model's predictions.
[0039] Delayed bookings and cancellations: Previous day's lagged bookings: The number of bookings from the previous day.
[0040] 3-Day Lag Bookings: The number of bookings made in the previous three days.
[0041] 7-Day Lag Bookings: The number of bookings made in the previous seven days.
[0042] Delayed cancellations from the previous day: The number of cancellations from the previous day.
[0043] Late cancellations in the last 3 days: The number of cancellations in the last three days.
[0044] Late cancellations in the last 7 days: The number of cancellations in the last seven days.
[0045] The lag feature helps the model capture the dynamic changes and autocorrelation characteristics in the time series. By introducing lags of different time spans, the model can be more sensitive to short-term and medium-term changes and improve the accuracy of predictions.
[0046] The first feature data set is merged and redundant or irrelevant features are removed to form a time series data set; Data fusion is the process of integrating data from different data sources or different characteristics into a unified dataset. Here, it refers to the fusion of the first characteristic dataset with other data (such as meteorological data, holiday information, etc.). Data fusion can combine information from different sources to provide more comprehensive background information, helping the model to better understand and predict tourism trends.
[0047] A time series dataset is a dataset that is arranged in chronological order and contains all useful features after feature selection and data fusion. A time series dataset is the basic data for training time series prediction models (such as LSTM).
[0048] Integrate the first feature dataset with data from other data sources. For example, align and merge meteorological data (temperature, precipitation, wind speed, etc.) and holiday information (whether it is a holiday, date of major events, etc.) with the first feature dataset by time index. Generate a comprehensive dataset that contains all the features that may affect tourism trends.
[0049] Furthermore, statistical methods or machine learning algorithms are used for feature selection. Common methods include correlation analysis, principal component analysis (PCA), Lasso regression, etc. By calculating the correlation between each feature and the target variable, the features with high correlation are screened out, and the features with low correlation or irrelevant are removed. PCA and other dimensionality reduction methods are used to reduce the number of features, retain the main components, and simplify the model.
[0050] Furthermore, the data is cleaned, missing values and outliers are processed to ensure data quality. All features are standardized and feature values are scaled to the same range (such as [0,1]) to ensure the stability and fast convergence of model training.
[0051] In this step, by fusing multi-source data, the model can use more comprehensive information, capture complex trends and influencing factors, and improve prediction accuracy. Feature selection can reduce the number of input features of the model, simplify the model structure, improve training and prediction efficiency, and reduce the risk of overfitting.
[0052] The LSTM model is trained using the time series dataset, and the parameters are tuned using the grid search method to determine the optimal model parameters.
[0053] Long Short-Term Memory (LSTM) is a special type of recurrent neural network (RNN) that can effectively capture and memorize dependencies over long time spans. LSTM solves the gradient vanishing and exploding problems in traditional RNNs by introducing a gating mechanism. The LSTM model is used to process and predict time series data and is suitable for complex data with long-term dependencies.
[0054] This embodiment selects an LSTM model, whose structure includes at least an input layer, one or more LSTM hidden layers, and an output layer. This embodiment does not specifically limit the number of network layers and the number of neurons in each layer. Those skilled in the art can perform experiments and determine them according to the amount of actual data when implementing this solution.
[0055] Determine the hyperparameters that need to be tuned, such as learning rate, number of LSTM layers, number of neurons per layer, batch size, number of training iterations, etc. Set the possible value range for each hyperparameter.
[0056] Perform model training and evaluation for each set of hyperparameter combinations, and record the model performance of each training (such as the mean square error on the validation set). Use the cross-validation method to evaluate the generalization ability of the model and ensure the reliability of the parameter tuning results.
[0057] Select the parameter combination that makes the model perform best on the validation set as the final hyperparameter of the LSTM model. Retrain the model using all training data and the best hyperparameters.
[0058] In this step, the grid search method is used to find the best hyperparameter combination, which can significantly improve the prediction accuracy of the LSTM model. Parameter tuning can optimize the structure of the LSTM model to better adapt it to the characteristics of time series data.
[0059] Use the trained model to make short-term forecasts and predict tourism trends in the future.
[0060] Collect the latest real-time data and ensure that the data is consistent with the data structure and format used when the model is trained. Perform the same preprocessing and feature extraction operations on the real-time data, including normalization and time series alignment.
[0061] Construct a data set suitable for the model input format, arrange the latest feature data in time series, and prepare to input it into the trained LSTM model. Use the trained LSTM model to predict the constructed input data and obtain the tourism trend prediction results for a period of time in the future.
[0062] Analyze forecast results to identify future peak periods and potential problems. Develop corresponding management and operation strategies based on forecast results, such as increasing service staff, adjusting opening hours, optimizing marketing activities, etc.
[0063] In the above scheme, by extracting and utilizing multiple features (such as daily bookings and cancellations, booking-to-cancellation ratio, cumulative bookings and cancellations, and lagged bookings and cancellations), the model can more comprehensively capture the short-term changes in tourist behavior and improve the accuracy of short-term predictions.
[0064] By collecting external factors such as meteorological data and holiday types and incorporating them into the feature data set, the model’s ability to perceive unexpected events is enhanced. For example, data on large-scale events and extreme weather can help the model better capture and predict short-term dramatic fluctuations in the number of tourists.
[0065] Through systematic feature engineering steps, we extract features such as daily bookings and cancellations, booking and cancellation ratios, non-refundable bookings, cumulative bookings and cancellations, and delayed bookings and cancellations from booking and cancellation data to form a feature data set. Then, we perform feature selection and standardization to optimize data quality and model performance.
[0066] Through the data fusion step, multi-source data such as the historical number of tourists, hotel bookings and cancellations, the number of non-refundable hotels and tickets, meteorological data and holiday types are integrated to provide more comprehensive background information and help the model better understand and predict tourism trends.
[0067] This solution significantly improves the accuracy and reliability of tourism trend forecasting through systematic data collection, feature engineering, and data fusion. By dealing with short-term emergencies and external factors, optimizing model structure and hyperparameter configuration, it provides real-time and accurate tourism trend forecasting results, providing strong support for tourism management and decision-making.
[0068] On the other hand, the present invention also provides a tourism trend prediction system based on multi-source data fusion, comprising the following modules: A collection module, used to collect multi-source data related to tourism, wherein the multi-source data includes at least: the number of historical tourists, the number of hotel reservations and cancellations, the number of non-refundable hotels, the number of ticket reservations and cancellations, the number of non-refundable tickets, meteorological data, and holiday types; The feature engineering module is used to extract features from the booking and cancellation data, including: daily booking and cancellation volume, booking and cancellation ratio, non-refundable volume, cumulative booking and cancellation volume in the past 7 days and 30 days; lagged booking and cancellation volume in the previous 1 day, 3 days and 7 days, to obtain the first feature data set; A fusion module, used to fuse the first feature data set with non-first feature data sets in the multi-source data, and remove redundant or irrelevant features to form a time series data set; A training module is used to train the LSTM model using the time series data set, and to perform parameter tuning through a grid search method to determine the optimal model parameters; The prediction module is used to make short-term predictions using the trained model to predict tourism trends in the future.
[0069] Furthermore, the specific implementation methods of the above-mentioned tourism trend prediction system based on multi-source data fusion are the same as a tourism trend prediction method based on multi-source data fusion, and all further technical solutions in a tourism trend prediction method based on multi-source data fusion are fully introduced into a tourism trend prediction system based on multi-source data fusion.
[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
[0071] The present invention does not particularly specify the structure of some modules, which shall be subject to the contents recorded in the prior art. The prior art mentioned in the above background technology section and the specific embodiment section of the present invention can be used as part of the present invention to understand the meaning of some technical features or parameters. The scope of protection of the present invention shall be subject to the contents actually recorded in the claims.
Claims
1. A tourism trend prediction method based on multi-source data fusion, characterized in that: The following steps are involved: Collect multi-source data related to tourism, the multi-source data at least including: historical number of tourists, number of hotel reservations and cancellations, number of non-refundable hotels, number of ticket reservations and cancellations, number of non-refundable tickets, meteorological data, and holiday types; Extract features from booking and cancellation data, including: daily booking volume and cancellation volume, booking and cancellation ratio, non-refundable volume, cumulative booking volume and cancellation volume in the past 7 days and 30 days; lagged booking volume and cancellation volume in the previous 1 day, 3 days, and 7 days, to obtain the first feature data set; The first feature data set is merged with non-first feature data sets in the multi-source data, and redundant or irrelevant features are removed to form a time series data set; Use the time series data set to train the LSTM model, perform parameter tuning through a grid search method, and determine the optimal model parameters; Use the trained model to make short-term forecasts and predict tourism trends in the future.
2. The tourism trend prediction method based on multi-source data fusion according to claim 1 is characterized in that The delayed booking volume and cancellation volume in the previous 1 day, 3 days and 7 days are as follows: Previous day's delayed bookings: the number of bookings from the previous day; 3-day delayed bookings: the number of bookings in the previous three days; 7-day delayed bookings: the number of bookings made in the previous seven days; Delayed cancellations of the previous day: the number of cancellations of the previous day; Delayed cancellations in the previous three days: the number of cancellations in the previous three days; Late cancellations in the last 7 days: The number of cancellations in the last seven days.
3. The tourism trend prediction method based on multi-source data fusion according to claim 1 is characterized in that Removing redundant or irrelevant features includes: Correlation analysis, principal component analysis, or Lasso regression calculates the correlation between each feature and the target variable, screens out features with high correlation, and removes features with low correlation or irrelevant features.
4. The tourism trend prediction method based on multi-source data fusion according to claim 3 is characterized in that After removing redundant or irrelevant features, it also includes: The PCA dimensionality reduction method is used to reduce the number of features and retain the main components.
5. The tourism trend prediction method based on multi-source data fusion according to claim 1 is characterized in that Before using the trained model for short-term forecasting, collect the latest real-time data, ensure that the data is consistent with the data structure and format used when the model is trained, and perform the same preprocessing and feature extraction operations on the real-time data, including standardization and time series alignment.
6. A tourism trend prediction system based on multi-source data fusion, characterized in that: Includes the following modules: A collection module, used to collect multi-source data related to tourism, wherein the multi-source data includes at least: the number of historical tourists, the number of hotel reservations and cancellations, the number of non-refundable hotels, the number of ticket reservations and cancellations, the number of non-refundable tickets, meteorological data, and holiday types; The feature engineering module is used to extract features from the booking and cancellation data, including: daily booking and cancellation volume, booking and cancellation ratio, non-refundable volume, cumulative booking and cancellation volume in the past 7 days and 30 days; lagged booking and cancellation volume in the previous 1 day, 3 days and 7 days, to obtain the first feature data set; A fusion module, used to fuse the first feature data set with non-first feature data sets in the multi-source data, and remove redundant or irrelevant features to form a time series data set; A training module is used to train the LSTM model using the time series data set, perform parameter tuning through a grid search method, and determine the optimal model parameters; The prediction module is used to make short-term predictions using the trained model to predict tourism trends in the future.
7. The tourism trend prediction system based on multi-source data fusion according to claim 6 is characterized in that The delayed booking volume and cancellation volume in the previous 1 day, 3 days and 7 days are as follows: Previous day's delayed bookings: the number of bookings from the previous day; 3-day delayed bookings: the number of bookings in the previous three days; 7-day delayed bookings: the number of bookings made in the previous seven days; Delayed cancellations of the previous day: the number of cancellations of the previous day; Delayed cancellations in the previous three days: the number of cancellations in the previous three days; Late cancellations in the last 7 days: The number of cancellations in the last seven days.
8. The tourism trend prediction system based on multi-source data fusion according to claim 6 is characterized in that Removing redundant or irrelevant features includes: Correlation analysis, principal component analysis, or Lasso regression calculates the correlation between each feature and the target variable, screens out features with high correlation, and removes features with low correlation or irrelevant features.
9. The tourism trend prediction system based on multi-source data fusion according to claim 8 is characterized in that After removing redundant or irrelevant features, it also includes: The PCA dimensionality reduction method is used to reduce the number of features and retain the main components.
10. The tourism trend prediction system based on multi-source data fusion according to claim 6 is characterized in that Before using the trained model for short-term forecasting, collect the latest real-time data, ensure that the data is consistent with the data structure and format used when the model is trained, and perform the same preprocessing and feature extraction operations on the real-time data, including standardization and time series alignment.
Citation Information
Patent Citations
User retrieving method and system based on user figure
CN107730019A
Travel passenger flow volume prediction method based on seasonal long short-term memory network model
CN108446759A
A tourist flow prediction method based on machine learning
CN109034469A
Daily tourist volume prediction method based on K-means and LSTM
CN111652444A
Scenic spot people flow control system and method
CN113255563A
Cited By
Disaster early warning method and system based on key target detection and image segmentation
CN119649034A
Tourism prediction method based on multi-source data fusion
CN120893623A
Tourism demand prediction method and device based on TFT model, and storage medium
CN121146179A
