Multi-channel room reservation method and system for hotels

Abstract

The application provides a multi-channel room booking method and system for a hotel, comprising: acquiring different booking channel room booking history data, the room booking history data comprising room price level, price fluctuation and room type inventory data, and dividing the data according to customer booking behavior, length of stay, and historical order average room price key attributes to obtain a room type data set; removing outliers and missing values in the room type data set, extracting customer age range, preferred room type, price sensitivity, and historical stay frequency as key features, normalizing the key features to form a standardized feature vector describing the demand characteristics and price sensitivity of each segmented customer group; and dynamically shutting down the sales of part of the room types in combination with real-time room status data of the hotel, room type demand prediction results and price sensitivity thresholds.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and in particular to a multi-channel room booking method and system for a hotel. BACKGROUND

[0002] In the multi-channel booking management of a hotel, there is a problem of inconsistent customer behavior data of different channels. How to standardize these multi-dimensional data and accurately extract customer demand characteristics and price sensitivity is an important problem. The hotel needs to classify customers according to these characteristics and predict the room type demand of various customer groups in different time periods in the future. However, how to combine these prediction results with real-time room state management to maximize revenue while avoiding the loss of potential customers due to incorrect room closing operations remains a challenge. Building a dynamic pricing mechanism to make price adjustments consistent with customer price sensitivity and form a closed-loop feedback with demand prediction for continuous optimization is an important challenge to improve hotel management efficiency. SUMMARY

[0003] To solve the above problems, the present application provides a multi-channel room booking method for a hotel, mainly comprising:

[0004] Obtain different booking channel booking history data, including room price level, price fluctuation and room type inventory data, and divide the data according to customer booking behavior, length of stay, and historical order average room price key attributes to obtain a room type data set;

[0005] Remove outliers and missing values in the room type data set, extract customer age range, preferred room type, price sensitivity, and historical stay frequency as key features, normalize the key features, and form a standardized feature vector describing the demand characteristics and price sensitivity of each segmented customer group;

[0006] According to the standardized feature vector describing the demand characteristics and price sensitivity of each segmented customer group, a clustering algorithm is used to divide the customers into different demand preference groups, and for each demand preference group, a room type demand standardized feature vector data set and a price sensitivity threshold data set are established under different channels and different time granularities;

[0007] For the room type demand feature vector data set of each demand preference group, a time series prediction model is used to predict the room type demand of different channels and different groups in the future by dynamically adjusting the time granularity of the training data, and according to the price sensitivity threshold data set, the upper limit of the price adjustment range of different groups is determined;

[0008] Combine the hotel real-time room state data, room type demand prediction results and price sensitivity threshold to dynamically adjust the price and stop the sale of part of the room types;

[0009] According to the price information after dynamic pricing and the price sensitivity threshold, in combination with the historical check-in frequency, evaluation score and repeat purchase rate of different demand preference groups, an optimal room type combination is matched from a hotel room type library by using a collaborative filtering algorithm, and personalized room type recommendations are pushed to target customers at different time nodes in different room type publishing channels;

[0010] If the customer prefers a room type and the reservation price is within the price sensitivity threshold, the reservation of the room type is locked, and the room state association strategy of the room type is dynamically adjusted according to the customer reservation time and the length of stay attribute; if the customer does not prefer a room type or the reservation price exceeds the price sensitivity threshold, a substitute room type is dynamically recommended within the price sensitivity threshold according to the customer demand characteristics and the hotel room type inventory.

[0011] The application also provides a multi-channel room reservation system for a hotel, comprising:

[0012] A data acquisition and preprocessing module is configured to obtain historical room reservation data of different reservation channels, the historical room reservation data including room price level, price fluctuation and room type inventory data, and to divide the data according to customer reservation behavior, length of stay and historical order average room price key attributes to obtain a room type data set; abnormal values and missing values in the room type data set are removed, customer age range, preferred room type, price sensitivity and historical check-in frequency are extracted as key features, the key features are normalized to form a standardized feature vector describing demand characteristics and price sensitivity of each segmented customer group;

[0013] A customer grouping module is configured to divide customers into different demand preference groups by using a clustering algorithm according to the standardized feature vector describing demand characteristics and price sensitivity of each segmented customer group, and to establish a room type demand standardized feature vector data set and a price sensitivity threshold data set of each demand preference group at different channels and different time granularities;

[0014] A demand prediction module is configured to predict room type demand of different channels and different groups in the future by dynamically adjusting the time granularity of training data for the room type demand feature vector data set of each demand preference group by using a time series prediction model, and to determine the upper limit of price adjustment amplitude of different groups according to the price sensitivity threshold data set;

[0015] A dynamic pricing module is configured to dynamically price and stop selling part of the room types in combination with real-time room state data of the hotel, room type demand prediction results and price sensitivity thresholds;

[0016] The personalized recommendation module is used for matching the optimal room type combination from the hotel room type library according to the price information after dynamic price adjustment and the price sensitivity threshold, combining the historical check-in frequency, evaluation score and repurchase rate of key attributes of different demand preference groups, and pushing personalized room type recommendations to target customers at different time nodes in different room state release channels.

[0017] The technical scheme provided by the embodiment of the application can include the following beneficial effects:

[0018] The application discloses a multi-channel room booking method. In view of the problems of inaccurate demand prediction and inflexible pricing strategy in hotel room type booking, the application obtains room price levels, price fluctuations and room type inventory data in a multi-granularity time dimension, and combines customer booking behavior, check-in duration, historical order average room price and other key attributes to split and clean the data, extracts customer age range, preferred room type, price sensitivity and historical check-in frequency as key features, and divides customers into different demand preference groups by using a clustering algorithm. On this basis, the application establishes a standardized feature vector describing the demand characteristics and price sensitivity of each segmented customer group, and uses a time series prediction model to predict room type demand of different channels and groups in the future, and determine the upper limit of price adjustment range of different groups. Finally, the application dynamically adjusts room type prices and stops room type sales according to room type demand prediction results and price sensitivity thresholds, combines real-time room state data of the hotel, and uses a collaborative filtering algorithm to push personalized room type recommendations to target customers. The application can improve the accuracy of hotel room type demand prediction, realize dynamic pricing and personalized recommendation, thereby optimizing hotel room type inventory management and improving hotel operation efficiency and revenue level. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A flowchart of a multi-channel room booking method for a hotel according to the application.

[0020] Figure 2 A block diagram of a multi-channel room booking system for a hotel according to the application. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical scheme and advantages of the application clearer, the application will be described in detail below with reference to the drawings and specific embodiments.

[0022] As Figure 1 , the multi-channel room booking method for a hotel according to the embodiment specifically can include:

[0023] In step S101, booking history data of different booking channels is obtained, the booking history data includes room price level, price fluctuation and room type inventory data, and the data is divided according to customer booking behavior, length of stay, and historical order average room price key attributes to obtain a room type data set.

[0024] Raw data carrying a booking channel identifier is obtained, the raw data being sent by different booking channels; the raw data is preprocessed to obtain a feature matrix, the feature matrix containing multiple time granularity features; the feature matrix is subjected to clustering processing to obtain clustering results of different room types; a room type sub-data set is constructed for each clustering result; a time series decomposition instruction is executed on the room type sub-data set, the time series decomposition instruction being used to separate trend, seasonality and random fluctuation components; a price fluctuation index is calculated according to the time series decomposition result, the price fluctuation index containing multiple time granularities; and the price fluctuation index is combined with historical data to obtain a room type data set.

[0025] Specifically, raw data from different booking channels is obtained, a dataset is constructed for each booking channel separately, data cleaning methods are used to remove outliers and missing values, data standardization is used to unify the data format of different channels, and data is aggregated by hour, day, week, month, etc. according to the timestamp to obtain multi-dimensional time series data containing room price level, price volatility and room type inventory. Feature engineering is performed on the aggregated multi-dimensional time series data, the groupby and agg functions of the pandas library are used to calculate the average booking lead time, booking frequency and stay length of each customer, customer booking behavior features are extracted, the average stay length of each customer is calculated, the average room price of historical orders is calculated, and for classification features such as room type, the get_dummies function of pandas is used for one-hot encoding, the PCA function of sklearn library is used for principal component analysis, and 95% of the variance is set as the threshold to generate a feature matrix containing multiple time granularities. Based on the generated feature matrix, the KMeans function of sklearn library is used to cluster room types, the optimal number of clusters is selected by the silhouette coefficient, the Euclidean distance from the sample to each cluster center is calculated, and the center point is iteratively optimized until it is stable, obtaining the clustering result of different room types. For each clustering result, a sub-dataset containing features of this type of room is constructed separately. For each room type sub-dataset, the seasonal_decompose function of the statsmodels library is used for time series decomposition to separate trend, seasonality and random fluctuation components, the SimpleExpSmoothing function is used to predict the trend component, the prediction period is set to 30 days, and the price standard deviation of the past 30 days is calculated as the price volatility indicator. Combine the prediction results and historical data to calculate the price volatility indicator at each time granularity, and use the merge function of pandas to combine all the processing results to generate the final room type dataset containing multiple time granularities. When obtaining booking channel data, raw data is collected from multiple online travel platforms and hotel direct sales channels, including booking date, check-in date, check-out date, room type, price, etc. During data cleaning, records with negative or abnormally high prices are detected and deleted, and missing room type information is filled. During data standardization, the date format of different channels is unified to YYYY-MM-DD, and the price unit is unified to RMB. When aggregating by time, the average room price and remaining inventory are calculated by hour, the booking quantity and total revenue are aggregated by day, and the occupancy rate is calculated by week and month. In the feature engineering stage, the average booking lead time (the average of the difference between the booking date and the check-in date), the booking frequency (the number of bookings per month), and the average stay length (the average of the difference between the check-out date and the check-in date) are calculated for each customer. Room types are one-hot encoded, and "standard room", "double room", etc. are converted to binary features. Principal component analysis retains 95% of the variance, which may reduce the original 20+ features to about 10.For room type clustering, the K-means algorithm is used, and the silhouette coefficient is calculated for different K values from 2 to 10. The K value with the maximum silhouette coefficient is selected as the optimal clustering number. Assuming the optimal clustering number is 4, four room types are obtained, including economy, business, luxury, and suite. Time series decomposition decomposes the daily average price of each room type into trend, seasonality, and random components. For example, the trend of economy rooms may show a slow rise, and the seasonality shows a 15% price increase on weekends. When forecasting, the SimpleExpSmoothing model is used, with a setting of a = 0.3, to predict the trend for the next 30 days. The price volatility index is calculated as the standard deviation of the past 30 days' prices. For example, the standard deviation of a certain type of room is 50 yuan, indicating that the price fluctuation is large. The final multi-time granularity room type dataset contains information such as the average price, predicted price, price volatility index, and inventory of each room type at the hour, day, week, and month scales.

[0026] In step S102, the abnormal values and missing values in the room type dataset are removed, and the customer age range, preferred room type, price sensitivity, and historical occupancy frequency are extracted as key features. The key features are normalized to form a standardized feature vector describing the demand characteristics and price sensitivity of each segmented customer group.

[0027] The obtained room type dataset is preprocessed, including deleting abnormal values and filling missing values, to obtain a complete dataset. From the complete dataset, the customer age range, preferred room type, price sensitivity, and historical occupancy frequency are extracted as key features. The price sensitivity is determined by calculating the standard deviation of the booking price in the past predefined period divided by the average price. The extracted key features are normalized, and the minimum-maximum scaling method is used for continuous numerical features to map the feature values to a predefined interval. The processed features are combined to obtain a standardized feature vector describing the demand characteristics and price sensitivity of each segmented customer group.

[0028] Specifically, after obtaining the multi-time granularity room type dataset, the interquartile range method is used to detect outliers, and the interquartile range of each feature is calculated. If a data point is more than 1.5 times the interquartile range above the upper quartile or less than 1.5 times the interquartile range below the lower quartile, it is determined to be an outlier and the record is deleted, obtaining the cleaned dataset. For missing values in the cleaned dataset, the missing rate of each feature is calculated, and different processing methods are selected according to the data type: for numerical features, if the missing rate is less than 5%, use the median to fill, if the missing rate is between 5% and 20%, use the MICE algorithm in multiple imputation method to fill; for categorical features, use the mode to fill. If the missing rate exceeds 20%, delete the feature, thus obtaining a complete dataset. From the complete dataset, extract customer age range, preferred room type, price sensitivity, historical stay frequency as key features, determine price sensitivity by calculating the standard deviation of the booking price in the past year divided by the average price, calculate the booking frequency of each room type and select the highest proportion to determine the preferred room type, calculate the age according to the customer's birth date and divide it into predefined age ranges 18-25, 26-35, 36-45, etc., and count the number of stays in the past year to get the historical stay frequency. Normalize the extracted key features, for continuous numerical features such as price sensitivity and historical stay frequency, use the min-max scaling method to map the feature values to the [0, 1] interval, the calculation method is (x-min(x)) / (max(x)-min(x)), where x is the original feature value, min(x) and max(x) are the minimum and maximum values of the feature respectively; for discrete features such as age range and preferred room type, use one-hot encoding to convert them into binary feature vectors. Finally, combine all the processed features to obtain a standardized feature vector that describes the demand characteristics and price sensitivity of each segmented customer group. When processing the multi-time granularity room type dataset, first apply the interquartile range method to each feature to detect outliers. Taking room price as an example, the first quartile Q1 is 300 yuan, the third quartile Q3 is 800 yuan, and the interquartile range IQR is 500 yuan. Set the outlier threshold to Q1-1.5IQR(-450 yuan) and Q3+1.5IQR(1550 yuan), delete records with room price below-450 yuan or above 1550 yuan. For missing values, calculate the missing rate of each feature, such as the missing rate of customer age being 3%, use the median 35 years old to fill; the missing rate of room type preference is 15%, use the MICE algorithm for multiple imputation. When extracting key features, calculate the price sensitivity; the standard deviation of the booking price of a certain customer in the past year is 200 yuan, and the average price is 1000 yuan, so the price sensitivity is 0.2. The preferred room type is determined by calculating the booking proportion, such as a customer's standard room, double room, and suite booking proportions are 60%, 30%, and 10% respectively, so the preferred room type is standard room.The age is calculated according to the customer's birth date, for example, a customer born in 1990 is 34 years old, which falls into the 26-35 age group. The historical stay frequency is directly counted as the number of stays in the past year, for example, 12 times. During normalization, the price sensitivity is scaled by minimum-maximum scaling, assuming that the minimum value is 0.1 and the maximum value is 0.5, then the original value of 0.2 is normalized to 0.25. The age group and preferred room type are converted by one-hot encoding, for example, the 26-35 age group is converted to [0, 1, 0, 0], and the preference for a standard room is converted to [1, 0, 0]. Finally, all the processed features are merged to form a standardized feature vector, which describes the demand characteristics and price sensitivity of the segmented customer groups.

[0029] In step S103, according to the standardized feature vector describing the demand characteristics and price sensitivity of each segmented customer group, a clustering algorithm is used to divide the customers into different demand preference groups, and for each demand preference group, a standardized feature vector dataset of room type demand and a price sensitivity threshold dataset under different channels and different time granularities are established.

[0030] Receiving a standardized feature vector carrying customer demand characteristics and price sensitivity; according to the standardized feature vector, a clustering algorithm is used to group customers, the clustering algorithm detects and processes outliers by quartile range method; obtain the silhouette coefficient of different K values, determine the optimal clustering number K, and get the customer division result; for each clustering group, obtain the order information of different booking channels from the historical order database, and construct a feature vector reflecting the preference of the group in different channels; aggregate the historical order data by time granularity to obtain a standardized vector reflecting the demand characteristics of the group under different time granularities; using a random forest regression model, input the feature vector and the standardized vector, calculate the price sensitivity threshold of each demand preference group under different channels and time granularities.

[0031] Specifically, according to the standardized feature vectors describing the demand characteristics and price sensitivity of each segmented customer group, the K-means clustering algorithm is used to group customers. First, the interquartile range method is used to detect outliers, and the interquartile range IQR of each feature is calculated. Data points exceeding the range of Q1-1.5IQR and Q3+1.5IQR are considered outliers and are replaced by the median. Then, the optimal cluster number K is determined by calculating the silhouette coefficient for different K values. Set the random seed to ensure the result is reproducible. Iterate until the cluster center converges or the maximum iteration number is reached to get the customer division result. For each demand preference group obtained from each cluster, extract the historical order data of different booking channels such as the official website, OTA platform, and telephone booking. Calculate the order proportion, average booking lead time, and average stay length of each channel. Normalize these features using Min-Max normalization to scale the values to the [0,1] interval. Then concatenate the normalized features into a vector to construct a feature vector that reflects the channel preference of the group. For each demand preference group, based on historical order data, aggregate order information by hour, day, week, month, and other time granularities. Calculate the average order quantity, average room price, and room type distribution at each time granularity. Similarly, normalize these features using Min-Max normalization and concatenate them into a vector to construct a standardized vector that reflects the demand characteristics of the group at different time granularities. Combine the feature vectors of each demand preference group at different channels and different time granularities to calculate the price sensitivity threshold. Use a random forest regression model with input features including price variation percentage, time factors such as weekends / weekdays, holidays, and seasonal factors, and output as order quantity change percentage. Set a threshold for order quantity change, such as -10%, to determine the critical point of price variation that leads to significant changes in order quantity. Set this critical point as the price sensitivity threshold of the group at a specific channel and time granularity to form a complete dataset. When processing customer group data, first apply the interquartile range method to the standardized feature vector to detect outliers. For example, calculate Q1 as 0.2, Q3 as 0.8, and IQR as 0.6. Replace values less than -0.7 or greater than 1.7 with the median 0.5. Then use the K-means algorithm to group customers. Calculate the silhouette coefficient for K values from 2 to 10, and find that when K=5, the silhouette coefficient is highest at 0.68, so the optimal cluster number is 5. For each cluster group, extract order data from different channels, such as the order proportion of a group on the official website, OTA platform, and telephone booking, which are 50%, 30%, and 20% respectively. The average booking lead time is 7 days, 5 days, and 3 days, and the average stay length is 3 nights, 2 nights, and 4 nights. These features are normalized to [1,0.5,0], [1,0.5,0], and [0.5,0,1] respectively, and concatenated into a feature vector of [1,0.5,0,1,0.5,0,0.5,0,1].The time granularity data is also processed similarly, such as the average order quantity of 10 orders at 1 a.m., the average room price of 200 yuan, and the standard room proportion of 70%. These data are also normalized and spliced into the feature vector. Finally, the random forest regression model is used to analyze the price sensitivity, and the input includes factors such as 10% price increase, weekend, summer vacation, etc. The model output is a 15% decrease in order quantity. Set the order quantity change threshold to -10%, and the price increase amplitude that makes the order quantity decrease by 10% is 6.7%, which is set as the price sensitivity threshold of the group under specific conditions. Repeat this process to obtain the complete group feature and price sensitivity dataset.

[0032] Step S104, for the room type demand feature vector dataset of each demand preference group, a time series prediction model is used to predict the room type demand of different channels and different groups in the future by dynamically adjusting the time granularity of the training data. At the same time, according to the price sensitivity threshold dataset, the upper limit of the price adjustment amplitude of different groups is determined.

[0033] Obtain time series data, and adjust the time granularity of the training data according to the historical span of the time series data and the prediction target. If the prediction target is short-term, use hourly data; if the prediction target is medium-term, use daily data; if the prediction target is long-term, use weekly or monthly data. Use the adjusted time granularity data to predict the room type demand of different channels for the demand preference group; judge whether the average absolute percentage error of the prediction result is lower than the preset threshold. Obtain the price sensitivity data from the price sensitivity threshold dataset; calculate the price sensitivity distribution of different channels and time granularities for the demand preference group; take the preset quantile of the price sensitivity distribution as the upper limit of the price adjustment amplitude. Use linear programming method to optimize the price adjustment strategy; take the maximum expected revenue as the objective function, and take the upper limit of the price adjustment amplitude as the constraint condition to obtain the optimal price adjustment scheme.

[0034] Specifically, for each demand preference group's room type demand feature vector dataset, ARIMA and Prophet time series prediction models are used. The autocorrelation function and partial autocorrelation function are calculated to determine the value range of ARIMA model parameters p, d, and q. The grid search method is used to traverse different parameter combinations, and the parameter combination with the smallest AIC information criterion is selected as the optimal model parameter. At the same time, the cross-validation method is used to evaluate the performance of the two models, and the better-performing model is selected as the room type demand prediction model. According to the time span of historical data and the prediction target, the time granularity of training data is dynamically adjusted. Hourly data is used for short-term prediction of 1-7 days, daily data is used for medium-term prediction of 8-30 days, and weekly or monthly data is used for long-term prediction of 31-90 days. The data is divided into training and validation sets using the sliding window method. Using the adjusted time granularity data, the room type demand of each demand preference group in different channels is predicted, and the prediction results are compared with the actual data. The mean absolute percentage error (MAPE) and root mean square error (RMSE) are calculated. When MAPE is less than 20% and RMSE is less than 0.5 times the standard deviation of historical data, the model performance is good, otherwise the time granularity is adjusted. Based on the price sensitivity threshold dataset, the price sensitivity distribution of each demand preference group in different channels and time granularities is calculated, and the 95th percentile is taken as the upper limit of the price adjustment range for that group. This upper limit value is used as a constraint condition in the demand prediction model to ensure that the price change is considered in the prediction. The linear programming method is used to optimize the price adjustment strategy to maximize the expected revenue, forming a complete demand prediction and price adjustment decision support system. In dealing with the room type demand prediction of a certain business demand preference group, historical data analysis shows that the order data of this group in OTA channels presents obvious periodicity and trend. By comparing ARIMA and Prophet models, the optimal parameters of the ARIMA model are calculated as (2, 1, 2) with an AIC value of 876.5. The Prophet model automatically captures the annual, weekly, and holiday effects of the data. Using 5-fold cross-validation to evaluate the two models, the average MAPE of ARIMA is 15.3%, and that of Prophet is 13.8%. Therefore, the Prophet model is selected for subsequent prediction. For short-term prediction within 7 days, hourly data is used; for medium-term prediction of 8-30 days, daily data is used; for long-term prediction of 31-90 days, weekly data is used. For example, in the prediction of the next 30 days, the first 7 days use 24-hour granularity, and the next 23 days use daily granularity. The prediction results show that the MAPE is 16.5%, and the RMSE is 0.48 times the standard deviation of historical data, meeting the preset threshold requirements. According to the price sensitivity data of this group, the 95th percentile is calculated as 8.2%, which is the upper limit of the price adjustment range.The constraint condition is incorporated into the prediction model to ensure that the price change is not more than 8.2% when predicting. Finally, the linear programming method is used to optimize the price strategy to maximize the expected 30-day revenue, and the optimal price adjustment scheme is obtained, which is to increase the price by 3.5% on weekdays and 6.8% on weekends, which is expected to bring a 12.4% increase in revenue.

[0035] Step S105, combined with hotel real-time room status data, room type demand prediction results and price sensitivity threshold, dynamically adjust the price and close part of the room type sales.

[0036] Obtain a room type demand prediction model and room status data, calculate the demand prediction value of different room types in the future according to the room type demand prediction model and the room status data; calculate the optimal pricing strategy using a genetic algorithm, the optimal pricing strategy is based on the demand prediction value and the current room status data; receive real-time booking data after price adjustment, calculate the booking quantity and occupancy rate of each room type; if the booking quantity of a room type exceeds a preset threshold or the occupancy rate reaches a predetermined level, trigger a gradual closing mechanism; the gradual closing mechanism includes limiting sales through some channels and completely closing the sales of the room type; transmit the actual booking data, occupancy rate data and customer feedback information after price adjustment and room closing to the room type demand prediction model; adjust the room type demand prediction model parameters in real time according to the actual booking data, occupancy rate data and customer feedback information.

[0037] Specifically, according to the room type demand prediction model and the price sensitivity threshold, combined with real-time room status data of the hotel, an integrated prediction model is constructed by using ARIMA and Prophet algorithms, external data such as weather and holidays are introduced, and the demand of different room types in the next 7 days, 30 days and 90 days is predicted. The prediction results are updated daily through the rolling prediction method to obtain the room type demand prediction value under each time granularity. Based on the predicted demand and the current room status data, the optimal pricing strategy is calculated by using the genetic algorithm, the nonlinear relationship between price and demand is considered, and the differential pricing of different room types and different time periods is carried out under the constraint of the price sensitivity threshold. The price sensitivity threshold is determined by analyzing the minimum percentage of demand change caused by price change in the past 90 days, and the adjusted price is updated to each sales channel in real time. By monitoring the real-time booking data after price adjustment, the booking quantity and occupancy rate of each room type are calculated, and dynamic thresholds are set: the booking quantity threshold is 85% of the predicted demand, and the occupancy rate threshold is 90%. When the booking quantity of a room type exceeds the preset threshold or the occupancy rate reaches the predetermined level, the gradual shutdown mechanism is triggered, and the sales of part of the channels is limited first, and then the sales of the room type is completely closed. The actual booking data, occupancy rate data and customer feedback information after price adjustment and room closing are fed back to the demand prediction model, and the Online Gradient Descent online learning algorithm is used to update the model parameters in real time to improve the prediction accuracy. At the same time, the sliding window method is used to retain the data of the latest 90 days, and the price sensitivity threshold is dynamically adjusted to form a closed-loop dynamic optimization process.

[0038] In the process of implementing room type demand prediction and dynamic pricing, first, an integrated prediction model is built using ARIMA and Prophet algorithms. For example, for a luxury suite, the ARIMA model parameters are (2, 1, 2), and the Prophet model captures the obvious weekend effect and summer peak. The introduction of weather data shows that when the temperature exceeds 30°C, the demand for this room type increases by 15%. Combining these factors, the model predicts that the average daily demand for this room type in the next 7 days will be 20, in the next 30 days will be 18, and in the next 90 days will be 15. Based on these prediction results, the genetic algorithm calculates the optimal pricing strategy: the weekday price remains at 1200 yuan / night, the weekend price increases to 1500 yuan / night, and the summer peak period can reach a maximum of 1800 yuan / night. Historical data analysis shows that a price change of more than 8% will cause a significant change in demand, so the price sensitivity threshold is set to 8%. Real-time monitoring shows that on a certain Friday, the booking quantity of this room type reached 17, exceeding the 85% threshold of the predicted demand of 20, so the system automatically limits sales on some channels. When the occupancy rate reaches 92%, the sales of this room type are completely stopped. After price adjustment and room closure, the actual occupancy is 19, which is slightly lower than the predicted value. Using the OnlineGradient Descent algorithm, the model parameters are updated with a learning rate of 0.01, and the prediction accuracy is improved by 2%. At the same time, based on the data of the last 90 days, the price sensitivity threshold is adjusted to 7.8%, which is used for the next round of pricing decisions.

[0039] Based on the real-time room status data of the hotel, different room types are classified according to the current reservation status. For room types that are not reserved, the probability of future reservations and the demand for room types in a certain period of time are evaluated; for room types that are already reserved and on hold, it is determined whether there is overbooking or sudden demand increase, and whether early release or price increase locking is needed.

[0040] Receive real-time room status data carrying room type status identifier, the room status data is generated by the hotel management system in real time; according to the room status data, the room type is divided into three states of unbooked, booked and reserved, and the room number proportion of each state is calculated;Random forest algorithm is used to evaluate the booking probability of unbooked room type, the input features of the random forest algorithm include historical booking data, seasonal factors, weekends / weekdays and current room price;The future room demand is predicted by ARIMA model, the ARIMA model is trained based on historical booking data;Judge whether there is overbooking in the booked and reserved room type, if the ratio of current booking quantity to total room quantity exceeds the preset threshold, it is determined that there is overbooking;Calculate the probability of reserved room type converting to actual booking, the conversion probability is obtained based on historical reservation data statistics;According to the overbooking judgment result, the demand surge judgment result, the booking probability, the demand prediction and the reservation conversion probability, the dynamic programming algorithm is used to calculate the optimal room allocation scheme;Determine whether it is necessary to release the reserved room in advance or to increase the price to lock part of the room, and generate the corresponding operation suggestion list.

[0041] Specifically, according to the real-time room status data of the hotel, the room types are directly divided into three states of unbooked, booked and reserved according to the labels of the room status data, and the overall booking status of the current hotel is obtained by calculating the proportion of the number of rooms in each state. For the room types in the unbooked state, the random forest algorithm is used to evaluate the booking probability in the future period of time, the input features include historical booking data, seasonal factors, weekends / weekdays, current room price, etc., and the output is a booking probability value between 0 and 1, while the ARIMA model is used to predict the room demand in the next 7 days, 14 days and 30 days. For the room types in the booked and reserved states, by comparing the ratio of the current booking quantity to the total quantity of the room types, it is determined whether there is an overbooking situation, and the preset threshold is determined based on the average booking proportion when overbooking occurs in the past 90 days. The moving average method is used to calculate the recent demand growth rate, and when the growth rate exceeds the threshold value obtained from historical data analysis, it is determined as a demand surge. For the room types in the reserved state, the probability of converting reserved to actual booking is calculated and included in the subsequent decision-making process. Based on the judgment results of overbooking and demand surge, combined with the booking probability, demand prediction and reservation conversion probability of the room types, a dynamic programming algorithm is used to calculate the optimal room allocation scheme. The state is defined as the available number of each type of room, the action is the number of released or locked rooms, the reward function considers the expected revenue and overbooking risk, and the optimal strategy is calculated by backward iteration to determine whether to release reserved rooms in advance or lock some rooms at a higher price, and the corresponding operation suggestion list is generated. In the implementation of the room status management process, first, the 100 rooms of a certain hotel are divided into three states of unbooked, booked and reserved through real-time room status data, resulting in 50 unbooked, 40 booked and 10 reserved. For the 50 unbooked room types, the random forest algorithm is used to evaluate the booking probability, the input features include the average daily booking rate of the past 30 days 85%, the current season is summer, weekend, room price 800 yuan / night, etc., and the output booking probability is 0.75. At the same time, the ARIMA(2,1,2) model is used to predict the daily demand in the next 7 days as 45 rooms, 14 days as 42 rooms, and 30 days as 40 rooms. For the 40 booked and 10 reserved room types, the current booking proportion is 50%, which is lower than the average booking proportion of 80% when overbooking occurs in the past 90 days, and it is determined that there is no overbooking. Through the 5-day moving average method, the demand growth rate is calculated as 15%, which exceeds the 10% threshold value obtained from historical data analysis, and it is determined as a demand surge. For the 10 reserved room types, the probability of converting reserved to actual booking is calculated as 0.6 based on historical data. Based on the above data, the dynamic programming algorithm is used to calculate the optimal room allocation scheme, the current state is 50 available rooms, the action space is to release 0-10 reserved rooms or lock 0-20 unbooked rooms, and the reward function considers the expected revenue and overbooking risk. After 1000 iterations, the optimal strategy is to release 5 reserved rooms and lock 10 unbooked rooms, which can increase the revenue by 8% and control the overbooking risk within 3%.

[0042] According to the state of different room types and the future demand prediction results, a room closing strategy is formulated. For room types with high future demand but no current booking, a dynamic pricing strategy is preferred to increase the price to control demand; for room types with sustained low demand, the room is temporarily closed or the room sales channel is reopened. At the same time, the room state is updated in real time combined with the customer booking behavior pattern to ensure the synchronization and accuracy of the room state information and the room demand prediction model.

[0043] The current booking rate, future predicted demand, and historical average occupancy rate are obtained, and a random forest algorithm is used to build a room closing strategy model to obtain a room closing suggestion. According to the room closing suggestion, it is determined whether dynamic pricing or temporary room closing is needed. If dynamic pricing is needed, the price elasticity value is calculated using the demand elasticity formula; according to the price elasticity value, the pricing range is determined by a linear programming method, which includes setting a target function of maximizing revenue and constraint conditions of room quantity, maximum and minimum price. The sales channel is analyzed to identify inefficient channels. Real-time analysis of booking data is performed using a distributed message queue, and an online gradient descent algorithm is used to optimize the demand prediction model to update the room state information.

[0044] Specifically, according to the real-time state and future demand prediction results of different room types, a room closing strategy model is constructed using the random forest algorithm. The input features include the current booking rate, future 7-day predicted demand, historical average occupancy rate, etc. The output is a room closing recommendation to determine whether to take dynamic pricing or temporary room closing measures. For room types with high future demand but no current booking, the price elasticity is calculated using the demand elasticity formula E = (ΔQ / Q) / (ΔP / P). E is the price elasticity coefficient of demand, which represents the percentage change in demand when the price changes by 1%. ΔQ is the change in demand, which is the difference between the new demand and the original demand. Q is the original demand level before the price change. ΔP is the price change, which is the difference between the new price and the original price. P is the original price before the demand change. The specific pricing range is determined by setting the objective function (maximize revenue) and constraint conditions (room quantity, maximum and minimum price) using linear programming methods to achieve precise control of demand. A price threshold is set to prevent excessive pricing from affecting hotel reputation. For room types with persistent low demand, a demand slump judgment standard is set, such as a 7-day booking rate below 30%, triggering a temporary room closing mechanism, and starting a sales channel reevaluation program. The existing sales channels are analyzed using the K-means clustering algorithm to identify inefficient channels and suspend cooperation, optimizing the channel structure. Combining customer booking behavior patterns and sales channel analysis results, a real-time room state update mechanism is constructed to analyze booking data in real-time using a distributed message queue. The demand prediction model is continuously optimized using an online gradient descent algorithm to ensure the synchronization and accuracy of room state information and demand prediction models, achieving dynamic room state management. During the implementation of dynamic room state management, the room closing strategy model is first constructed using the random forest algorithm. The input features include the current booking rate of 75%, the future 7-day predicted demand of 90 rooms, the historical average occupancy rate of 85%, etc. After training with 1000 decision trees, the model outputs a recommendation to take dynamic pricing measures for the luxury suite. Subsequently, the price elasticity for the luxury suite is calculated. When the price is raised from 1000 yuan to 1100 yuan, the demand decreases from 20 rooms to 18 rooms, resulting in a price elasticity of -0.9. Based on this elasticity value, the optimal pricing range is determined to be 8% through linear programming, i.e., the price is adjusted to 1080 yuan. At the same time, the standard room has a 7-day booking rate below 25%, triggering a temporary room closing mechanism. The K-means clustering algorithm is started, dividing the 10 sales channels into 3 categories, identifying 2 inefficient channels, and suspending their room allocation.Combined with customer booking behavior and channel analysis results, 1000 booking data per second is processed by using a distributed message queue, the demand prediction model parameters are updated in real time by using an online gradient descent algorithm, the learning rate is set to 0.01, the model is updated once every 100 data, the synchronization of room state information and demand prediction model is improved to 99.5%, the accuracy is improved by 3 percentage points, and precise dynamic room state management is realized.

[0045] In step S106, according to the price information after dynamic pricing and the price sensitivity threshold, combined with the historical occupancy frequency, evaluation score and repeat purchase rate of different demand preference groups, the optimal room type combination is matched from the hotel room type library by using a collaborative filtering algorithm, and personalized room type recommendations are pushed to target customers in different room state publishing channels and different time nodes.

[0046] The customer and room type matrix is obtained, which includes historical occupancy frequency, evaluation score and repeat purchase rate; the customer and room type matrix is standardized to obtain standardized customer and room type data. The similarity between customers is calculated by using a user-based collaborative filtering algorithm; the recommendation score of each room type for the target customer is calculated according to the historical preference of the similar customer group and the current room type inventory status, combined with the current price of the room type, the price sensitivity of the customer and the historical evaluation; the recommendation score is sorted, and the K room types with the highest score are selected as the candidate recommendation result; the candidate recommendation result is filtered according to the price sensitivity threshold to obtain the final recommended room type. The customer historical behavior data is received; the best push time is predicted by using the customer historical behavior data; the push strategy of different channels is dynamically optimized, and the room type recommendation information is sent to the target customer through the optimal channel at the best push time.

[0047] Specifically, according to the dynamic pricing information and the price sensitivity threshold, a customer-room type matrix is constructed, the matrix elements contain the historical stay frequency, evaluation score and repeat purchase rate of customers to room types, the Z-score standardization method is used to process the data, and the attributes of different dimensions are converted to a unified scale. The similarity between customers is calculated using the user-based collaborative filtering algorithm, and the cosine similarity and Pearson correlation coefficient are used as the measurement indicators. For each target customer, find 5% of the total number of customers, not less than 10 similar customers, form a similar customer group. Based on the historical preferences of the similar customer group, combined with the current room type inventory status, the weighted factor decomposition machine model is used to calculate the recommendation score of each room type for the target customer, and the score calculation considers the current price of the room type, the price sensitivity of the customer, the historical evaluation and other factors. According to the recommendation score ranking, select the top K room types as the recommendation result, and use the price sensitivity threshold to filter the recommendation result to ensure that the recommended room type price does not exceed the customer's acceptable range. Combined with the characteristics of different room state publishing channels and the customer's active time, time series analysis is introduced, and the best push time is predicted based on the customer's historical behavior data. Recurrent neural network is used to dynamically optimize the push strategy of different channels, and at the appropriate time node, the target customer is pushed room type recommendation information through the optimal channel. In the process of implementing personalized room type recommendation, first, a matrix containing 1000 customers and 50 room types is constructed, each element contains historical stay frequency 0-20 times, evaluation score 1-5 points and repeat purchase rate 0-100%. The Z-score standardization method is used to process the data, and the three attributes are converted to a standard normal distribution with a mean of 0 and a standard deviation of 1. Then, the collaborative filtering algorithm is used to calculate the customer similarity, and the cosine similarity and Pearson correlation coefficient are considered, and the weighted average of the two is taken as the final similarity, with weights of 0.6 and 0.4. For each target customer, select the top 50 customers with the highest similarity as the similar customer group. Next, the weighted factor decomposition machine model is used to calculate the recommendation score, the input features include the current room type price 500-5000 yuan, the customer price sensitivity 0.1-2.0, the historical evaluation 1-5 points, etc. 10 dimensions, and the final model is obtained through 100 iterations of training. According to the recommendation score output by the model, select the top 5 room types with the highest score, and use the price sensitivity threshold, such as 1.5, to filter, to ensure that the recommended room type price does not exceed 1.5 times the customer's historical average consumption. Finally, the long short-term memory network (LSTM) is used to analyze the customer's behavior data in the past 90 days to predict the best push time. At the same time, the recursive neural network dynamically optimizes the selection strategy of the 5 push channels, such as APP, SMS, email, etc., and at the predicted best time point, the customer is pushed personalized room type recommendation through the optimal channel.

[0048] Further, if the customer prefers a room type and the reservation price is within the price sensitivity threshold, the reservation of the room type is locked, and the room state association strategy of the room type is dynamically adjusted according to the customer reservation time and the length of stay attribute. If the customer does not prefer the room type or the reservation price exceeds the price sensitivity threshold, the alternative room type is dynamically recommended within the price sensitivity threshold according to the customer demand characteristics and the hotel room type inventory.

[0049] The reservation request information carrying the unique customer identification number is received, and the reservation request information is issued by the customer. The preferred room type list and the price sensitivity threshold information are obtained from the customer portrait database according to the unique customer identification number. The real-time price of the room type in the reservation request information is obtained, and it is judged whether the real-time price is lower than the price sensitivity threshold. If the real-time price is lower than the price sensitivity threshold, the room type is marked as a reserved state by triggering the room type locking mechanism. The decision tree algorithm is called according to the reserved state to dynamically adjust the availability and price of the associated room type. If the real-time price is higher than the price sensitivity threshold, the alternative room type recommendation process is started, and the alternative room type list is generated. The feedback information of the customer to the alternative room type list is received, and the customer preference information in the customer portrait database is updated according to the feedback information.

[0050] Specifically, after receiving the customer reservation request, the preferred room type list and price sensitivity threshold information of the customer are extracted from the customer portrait database. The database structure includes customer ID, preferred room type list, price sensitivity threshold, etc. fields. The relevant information is obtained using SQL query language, and the real-time price of the current reservation room type is obtained. By comparing the reservation price with the price sensitivity threshold, it is determined whether the reservation conditions are met. If the customer reserves the preferred room type and the reservation price is within the price sensitivity threshold, the room type locking mechanism is triggered, the room type is marked as reserved, and the customer reservation time and stay duration information are recorded. According to these attributes, the decision tree algorithm is called to dynamically adjust the availability and price of the associated room type, and the associated strategy of the room type is updated. If the customer reserves a non-preferred room type or the reservation price exceeds the price sensitivity threshold, the alternative room type recommendation process is started. Combined with collaborative filtering algorithm and content-based recommendation method, the room type characteristics such as area, facilities and customer preferences are considered, and a weighted hybrid strategy is used to generate an alternative room type list that meets the price sensitivity threshold. The alternative room type list is pushed to the customer through the customer's preferred communication channel, and a real-time monitoring mechanism is started. Real-time customer feedback is received using a distributed message queue, and real-time data analysis and updating are performed using a stream processing tool. The customer's feedback and selection of the recommended room type are tracked, and the data is returned to the customer portrait database for subsequent personalized recommendation optimization and dynamic adjustment of room state association strategy. In the implementation of the dynamic room type recommendation and reservation process, first, the customer ID is received as 10086, and the customer information is immediately extracted from the customer portrait database. The SQL query shows that the preferred room type is "luxury sea view room" and the price sensitivity threshold is 1500 yuan / night. The real-time price of the current reservation "luxury sea view room" is 1200 yuan / night, which is lower than the threshold, and it is determined that the reservation conditions are met. The room type locking mechanism is triggered, the room type is marked as reserved, the reservation time is recorded as 2023-06-15 20:30, and the stay duration is 3 nights. The decision tree algorithm adjusts the prices of the 5 adjacent "luxury sea view rooms" to 1296 yuan / night, and the prices of the 10 "standard sea view rooms" to 945 yuan / night. If the customer chooses to reserve a "standard double room" (1000 yuan / night), which exceeds the price sensitivity threshold, the alternative room type recommendation is started. The collaborative filtering algorithm recommends "superior double room" based on the historical choices of similar customers, and the content-based recommendation method considers the characteristics of the room type, such as area 80 square meters, with a massage bathtub, etc. The weights of the two algorithms are 0.6 and 0.4 respectively, and finally an alternative list containing "superior double room" (1350 yuan / night) and "luxury double room" (1480 yuan / night) is generated.Recommendations are pushed through customers' pre-set SMS channels, while real-time monitoring is initiated. A distributed message queue processes 1,000 customer feedback messages per second. Stream processing tools analyze the data in real time, showing that the click-through rate for "Superior Double Room" is 30% and the booking conversion rate is 5%. This data is updated to customer profiles in real time to optimize future recommendation strategies.

[0051] like Figure 2 The present invention also provides a multi-channel booking system for hotels, comprising:

[0052] The data acquisition and preprocessing module is used to acquire historical booking data from different booking channels. The historical booking data includes price levels, price fluctuations, and room type inventory data. Based on key attributes such as customer booking behavior, stay duration, and average price of historical orders, the data is segmented to obtain a room type dataset. Outliers and missing values ​​in the room type dataset are removed, and customer age group, preferred room type, price sensitivity, and historical stay frequency are extracted as key features. The key features are normalized to form a standardized feature vector describing the demand characteristics and price sensitivity of each segmented customer group.

[0053] The customer segmentation module is used to divide customers into different demand preference groups based on standardized feature vectors describing the demand characteristics and price sensitivity of each segmented customer group, using a clustering algorithm. For each demand preference group, it establishes a standardized feature vector dataset of room type demand and a price sensitivity threshold dataset under different channels and time granularities.

[0054] The demand forecasting module is used to predict the future demand for different types of housing from different channels and groups by using a time series forecasting model based on the housing demand feature vector dataset for each demand preference group. At the same time, it determines the upper limit of price adjustment range for different groups based on the price sensitivity threshold dataset.

[0055] The dynamic pricing module is used to dynamically shut down the sale of certain room types by combining real-time hotel room availability data, room demand forecasts, and price sensitivity thresholds.

[0056] The personalized recommendation module is used to match the optimal room type combination from the hotel room type library based on the dynamically adjusted price information and price sensitivity threshold, combined with the key attributes of historical check-in frequency, rating score and repurchase rate of different demand and preference groups. It uses a collaborative filtering algorithm to push personalized room type recommendations to target customers through different room availability release channels and at different time points.

[0057] It will be apparent to those skilled in the art that the application is not limited to the details of the above-exemplified embodiments and that the present application can be implemented in other particular forms without departing from the spirit or essential characteristics thereof. The embodiments should be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the above description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. No reference herein to any prior art is to be taken as an admission that the application is not entitled to antedate such prior art by virtue of prior application. Any reference to the use of a term in the singular herein shall also be taken to refer to the plural, and vice versa, unless otherwise indicated by context. Any reference to the use of a term in the present application shall be taken to refer to the use of the term in the context of the present application, and not to the use of the term in any other context. Any reference to the use of a term in the present application shall be taken to refer to the use of the term in the context of the present application, and not to the use of the term in any other context.

Claims

1. A multi-channel room booking method for hotels, characterized in that, The method includes: Historical booking data from different booking channels is obtained, including price levels, price fluctuations, and room type inventory data. The historical data is preprocessed to obtain a feature matrix containing multiple time-granularity features. The feature matrix is ​​then clustered to obtain clustering results for different room types. For each clustering result, a separate room type subset is constructed, and the data is further segmented based on key attributes such as customer booking behavior, stay duration, and average price of historical orders to obtain the room type dataset. Outliers and missing values ​​were removed from the room type dataset. Customer age group, preferred room type, price sensitivity, and historical occupancy frequency were extracted as key features. The key features were normalized to form a standardized feature vector describing the demand characteristics and price sensitivity of each segment of customer groups. Based on the standardized feature vectors describing the demand characteristics and price sensitivity of each segmented customer group, a clustering algorithm is used to divide customers into different demand preference groups. For each demand preference group, a standardized feature vector dataset of room type demand and a price sensitivity threshold dataset are established under different channels and time granularities. For each demand preference group's housing type demand feature vector dataset, a time series prediction model is used to predict the future housing type demand of different channels and groups by dynamically adjusting the time granularity of the training data. At the same time, based on the price sensitivity threshold dataset, the upper limit of the price adjustment range for different groups is determined. By combining real-time hotel room availability data, room type demand forecasts, and price sensitivity thresholds, dynamic price adjustments and the suspension of sales for certain room types are implemented; including: Obtain the housing demand forecasting model and housing status data, and calculate the future demand forecast values ​​for different housing types based on the housing demand forecasting model and housing status data; The optimal pricing strategy is calculated using a genetic algorithm, and the optimal pricing strategy is based on the demand forecast and current room availability data. Receive real-time booking data after price adjustments and calculate booking volume and occupancy rate for each room type; If the number of bookings for a certain room type exceeds a preset threshold or the occupancy rate reaches the pre-booked level, a gradual shutdown mechanism will be triggered. The gradual shutdown mechanism includes restricting sales through certain channels and completely closing sales of this room type; The actual booking and occupancy data after price adjustments and room closures, along with customer feedback, are transmitted to the room demand forecasting model. The parameters of the room type demand forecasting model are adjusted in real time based on actual booking data, occupancy rate data, and customer feedback. The method of dynamically shutting down the sale of certain room types by combining real-time hotel room availability data, room type demand forecasts, and price sensitivity thresholds also includes: classifying different room types according to their current booking status based on real-time hotel room availability data; assessing the booking probability and demand for unbooked rooms over a period of time; determining whether there is overbooking or a surge in demand for booked or reserved rooms, and assessing whether to release them early or lock them in at a higher price; formulating room closure strategies based on the status of different room types and future demand forecasts; prioritizing dynamic price adjustments for room types with high future demand but currently unbooked, increasing prices to control demand; and temporarily closing or reopening room sales channels for room types with persistently low demand. Simultaneously, updating room status in real-time based on customer booking behavior patterns ensures the synchronization and accuracy of room availability information with room type demand forecasting models. Based on the dynamically adjusted pricing information and price sensitivity threshold, and combined with key attributes such as historical check-in frequency, rating scores, and repurchase rates of different demand and preference groups, a collaborative filtering algorithm is used to match the optimal room type combination from the hotel room type database and push personalized room type recommendations to target customers through different room availability release channels and at different time points.

2. The method according to claim 1, characterized in that, The process involves acquiring historical booking data from different booking channels. This historical booking data includes price levels, price fluctuations, and room type inventory data. Based on key attributes such as customer booking behavior, stay duration, and average historical order price, the data is segmented to obtain a room type dataset, including: Obtain raw data carrying reservation channel identifiers, the raw data being sent from different reservation channels; The original data is preprocessed to obtain a feature matrix, which contains features at multiple time granularities. Clustering is performed on the feature matrix to obtain clustering results for different room types; For each clustering result, a separate sub-dataset of room types is constructed; A time series decomposition instruction is executed on the room type subset dataset, the time series decomposition instruction being used to separate trend, seasonality and random fluctuation components; A price volatility indicator is calculated based on the time series decomposition results, and the price volatility indicator includes multiple time granularities. The price fluctuation index is combined with historical data to obtain a room type dataset.

3. The method according to claim 1, characterized in that, The process involves removing outliers and missing values ​​from the room type dataset, extracting customer age groups, preferred room types, price sensitivity, and historical occupancy frequency as key features, and normalizing these key features to form standardized feature vectors describing the needs and price sensitivity of each segmented customer group. These vectors include: The obtained room type dataset is preprocessed, including removing outliers and filling in missing values, to obtain a complete dataset; Customer age group, preferred room type, price sensitivity, and historical stay frequency were extracted as key features from the complete dataset; Price sensitivity is determined by dividing the standard deviation of the customer’s past booking prices over a predefined time period by the average price. The extracted key features are normalized, and the min-max scaling method is used to map the feature values ​​to a predefined range for continuous numerical features. The processed features are combined to obtain a standardized feature vector describing the demand characteristics and price sensitivity of each segmented customer group.

4. The method according to claim 1, characterized in that, The process involves using standardized feature vectors describing the demand characteristics and price sensitivity of each segmented customer group, employing a clustering algorithm to divide customers into different demand preference groups. For each demand preference group, a standardized feature vector dataset of room type demand and a price sensitivity threshold dataset are established across different channels and time granularities, including: Receive a standardized feature vector carrying customer demand characteristics and price sensitivity; Customers are grouped using a clustering algorithm based on the standardized feature vectors. The clustering algorithm detects and processes outliers using the interquartile range method. Obtain the silhouette coefficients for different K values, determine the optimal number of clusters K, and obtain the customer segmentation results; For each cluster, order information from different booking channels is obtained from the historical order database, and a feature vector reflecting the group's preferences for different channels is constructed. Historical order data is aggregated by time granularity to obtain a standardized vector reflecting the demand characteristics of the group at different time granularities; Using a random forest regression model, the input feature vector and the standardized vector are used to calculate the price sensitivity threshold for each demand preference group under different channels and time granularities.

5. The method according to claim 1, characterized in that, The aforementioned housing demand feature vector dataset for each demand preference group employs a time series prediction model. By dynamically adjusting the time granularity of the training data, it predicts future housing demand from different channels and groups. Simultaneously, based on a price sensitivity threshold dataset, it determines the upper limit of price adjustment for different groups, including: Acquire time series data and adjust the time granularity of the training data based on the historical span of the time series data and the prediction target; Using adjusted time-granularity data, we can predict the demand for different types of housing among different demand groups through different channels; Determine whether the mean absolute percentage error of the prediction results is lower than a preset threshold; Obtain price sensitivity data from the price sensitivity threshold dataset; For groups with specific demand preferences, calculate the price sensitivity distribution across different channels and time granularities. The preset quantile of the price sensitivity distribution is taken as the upper limit of the price adjustment range; Optimize the price adjustment strategy using linear programming; By taking the maximization of expected returns as the objective function and using the upper limit of the price adjustment range as a constraint, the optimal price adjustment scheme is obtained.

6. The method according to claim 1, characterized in that, Based on real-time room availability data, different room types are categorized according to their current booking status. For unbooked rooms, the probability of booking and demand for these rooms in the future are assessed. For booked and reserved rooms, it is determined whether there is overbooking or a sudden surge in demand, and it is assessed whether early release or price increase is necessary to lock in the rooms, including: Receive real-time room status data carrying room type status identifiers, which is generated in real time by the hotel management system; Based on the room status data, the room types are divided into three states: unbooked, booked, and reserved, and the percentage of rooms in each state is calculated. The random forest algorithm is used to evaluate the booking probability of unbooked room types. The input features of the random forest algorithm include historical booking data, seasonal factors, weekend / weekday bookings, and current room rates. The ARIMA model is used to predict future demand for different room types; the ARIMA model is trained based on historical booking data. Determine whether there is overbooking of room types in the booked and reserved categories. If the ratio of the current number of bookings to the total number of room types exceeds a preset threshold, it is determined that there is overbooking. Calculate the probability that a reserved room type will be converted into an actual booking. The conversion probability is derived from historical reservation data. Based on the overbooking judgment results, the demand surge judgment results, the booking probability, the demand forecast, and the reservation conversion probability, the dynamic programming algorithm is used to calculate the optimal room allocation scheme. Determine whether it is necessary to release reserved rooms in advance or lock some rooms at a higher price, and generate a list of corresponding action suggestions.

7. The method according to claim 1, characterized in that, Based on the status of different room types and future demand forecasts, a room closure strategy is formulated; for room types with high future demand but currently unbooked, a dynamic pricing strategy is prioritized to increase prices and control demand; for room types with persistently low demand, room closures are temporarily suspended or sales channels are reopened; simultaneously, room status is updated in real time based on customer booking behavior patterns to ensure the synchronization and accuracy of room status information with the room demand forecasting model, including: The system obtains the current occupancy rate, the projected future demand, and the historical average occupancy rate. It then uses the random forest algorithm to construct a room closure strategy model and obtain room closure suggestions. Based on the room closure recommendations, determine whether dynamic pricing or temporary room closure is necessary; If dynamic price adjustment is required, the price elasticity value is calculated using the demand elasticity formula. Based on the price elasticity value, the price adjustment range is determined by a linear programming method, which includes setting an objective function to maximize revenue and constraints on the number of rooms and the highest and lowest prices. Analyze sales channels to identify inefficient channels; A distributed message queue is used to analyze booking data in real time, and an online gradient descent algorithm is used to optimize the demand forecasting model and update room availability information.

8. The method according to claim 1, characterized in that, Based on dynamically adjusted pricing information and price sensitivity thresholds, combined with key attributes such as historical check-in frequency, review scores, and repurchase rates for different demand and preference groups, a collaborative filtering algorithm is used to match the optimal room type combination from the hotel room type database. Personalized room type recommendations are then pushed to target customers through different room availability release channels and at different time points, including: Obtain a customer and room type matrix, which includes historical check-in frequency, rating score, and repurchase rate; The customer and room type matrix is ​​standardized to obtain standardized customer and room type data. A user-based collaborative filtering algorithm is used to calculate the similarity between customers; Based on the historical preferences of similar customer groups and the current room type inventory, combined with the current price of the room type, the customer's price sensitivity, and historical reviews, a recommendation score for each room type to the target customer is calculated. The recommended scores are sorted, and the K room types with the highest scores are selected as candidate recommendation results; The candidate recommendations are filtered based on a price sensitivity threshold to obtain the final recommended room type; Receive customer historical behavior data; Predict the optimal push time using the customer's historical behavior data; Dynamically optimize push strategies for different channels to send room type recommendations to target customers at the optimal time through the best channels.

9. A multi-channel booking system for hotels, used to implement the multi-channel booking method for hotels as described in any one of claims 1-8, characterized in that, The system includes: The data acquisition and preprocessing module is used to acquire historical booking data from different booking channels. The historical booking data includes price levels, price fluctuations, and room type inventory data. Based on key attributes such as customer booking behavior, stay duration, and average price of historical orders, the data is segmented to obtain a room type dataset. Outliers and missing values ​​in the room type dataset are removed, and customer age group, preferred room type, price sensitivity, and historical stay frequency are extracted as key features. The key features are normalized to form a standardized feature vector describing the demand characteristics and price sensitivity of each segmented customer group. The customer segmentation module is used to divide customers into different demand preference groups based on standardized feature vectors describing the demand characteristics and price sensitivity of each segmented customer group, using a clustering algorithm. For each demand preference group, it establishes a standardized feature vector dataset of room type demand and a price sensitivity threshold dataset under different channels and time granularities. The demand forecasting module is used to predict the future demand for different types of housing from different channels and groups by using a time series forecasting model based on the housing demand feature vector dataset for each demand preference group. At the same time, it determines the upper limit of price adjustment range for different groups based on the price sensitivity threshold dataset. The dynamic pricing module is used to dynamically adjust prices and shut down the sale of certain room types by combining real-time hotel room availability data, room demand forecasts, and price sensitivity thresholds. The personalized recommendation module is used to match the optimal room type combination from the hotel room type library based on the price information after dynamic price adjustment and the price sensitivity threshold, combined with the key attributes of historical check-in frequency, evaluation score and repurchase rate of different demand and preference groups. It uses a collaborative filtering algorithm to push personalized room type recommendations to target customers through different room availability release channels and at different time points.

Images