Intelligent hotel guest needs dynamic replenishment optimization system and method

Abstract

The present invention discloses an intelligent hotel guest goods dynamic replenishment optimization system and method, the present invention relates to the field of hotel data management technology, including a central server, the central server is communicatively connected with a client data acquisition module, a global model initialization module, a local model training module, a global prediction model update module, an iterative training module, an anomaly detection module and a replenishment strategy optimization module. The intelligent hotel guest goods dynamic replenishment optimization system and method can reduce inventory backlogs and avoid cost losses caused by excess by accurately predicting demand, and can ensure that the hotel can respond to customer requirements in a timely manner and provide high-quality services by analyzing customer consumption habits. When market conditions change, such as the sudden arrival of the tourist peak season or the occurrence of special events, the hotel can quickly adjust its strategy according to the prediction model, and long-term prediction can help the hotel management to formulate a more long-term strategic plan.

Description

Technical Field

[0001] The present invention relates to the technical field of hotel data management, and in particular to an intelligent hotel guest needs dynamic replenishment optimization system and method. Background Art

[0002] In recent years, the global hotel industry has shown a stable growth trend. With the rapid development of economic globalization and the improvement of people's living standards, the tourism industry has flourished, which in turn has promoted the development of the hotel industry. The continuous expansion of the hotel market has provided huge business opportunities and development space for hotel companies. At the same time, consumers' requirements for hotel services are getting higher and higher. They are no longer limited to providing accommodation services, but pay more attention to experience and personalized needs. Therefore, with the rapid development of science and technology, the application of intelligent technology in hotel management is becoming more and more popular. As a comprehensive solution, the intelligent hotel management system is playing an increasingly important role in improving hotel management efficiency, improving customer experience, and reducing operating costs.

[0003] Existing technologies often rely on manual inventory or regular inspections to obtain inventory information. This method has the problem of information lag. When the inventory is insufficient, it often takes a while to discover and replenish, resulting in the inability to meet customer demand. In addition, traditional replenishment strategies are often based on experience or simple sales data analysis to predict demand, which lacks scientificity and accuracy and is difficult to cope with the impact of seasonal changes, special events and other factors on demand. Therefore, how to achieve real-time monitoring and dynamic adjustment of customer inventory and optimize replenishment strategies according to real-time demand changes is a problem we need to solve. To this end, an intelligent hotel customer dynamic replenishment optimization system and method are proposed. Summary of the invention

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: In the first aspect, a smart hotel guest needs dynamic replenishment optimization system includes a central server, the central server is communicatively connected with a client data acquisition module, a global model initialization module, a local model training module, a global prediction model update module, an iterative training module, an anomaly detection module and a replenishment strategy optimization module, wherein the modules are connected by electrical signals;

[0005] The client data collection module is used to collect the hotel's local data on the consumption of guest supplies and external factors in real time. The external factors include weather, special events and occupancy rate. Through real-time data collection, it is ensured that the latest inventory and sales can be obtained, providing basic data for subsequent analysis and prediction, and performing pre-processing operations such as cleaning, sorting and format conversion on the collected local data to improve data quality and availability. Through pre-processing, noise and outliers in the data are eliminated to provide more accurate data support for subsequent analysis and prediction;

[0006] The global model initialization module is used to initialize the global prediction model on the central server, including short-term and long-term prediction components and related weights and adjustment factors, providing a basic framework for the entire system to ensure the accuracy and flexibility of the prediction;

[0007] In the local model training module, each hotel uses the pre-processed local data to perform short-term and long-term prediction training, and obtains a local model through comprehensive analysis;

[0008] The global prediction model update module is used to aggregate the local models uploaded by all hotels on the central server and use the federated average algorithm to update the global prediction model, thereby improving the prediction accuracy and robustness of the global prediction model by fusing the data of multiple hotels;

[0009] In the iterative training module, the central server sends the updated global prediction model to each hotel for use in the next round of training, ensuring that each hotel can obtain the latest global prediction model in real time, and repeat the process of local training, global prediction model update and model sending until the model converges and obtains the final global prediction model. Through continuous iterative training, the prediction model is optimized to improve the accuracy and stability of the replenishment strategy;

[0010] The anomaly detection module detects and analyzes the demand for hotel guest supplies based on the iteratively trained global prediction model and the local data collected in real time, and sets an anomaly detection threshold to perform anomaly detection;

[0011] The replenishment strategy optimization module dynamically adjusts the replenishment strategy according to real-time demand changes and market conditions, optimizes the inventory structure, reduces inventory costs, and ensures that customer needs are met in a timely manner.

[0012] Preferably, in the client data collection module, the process of acquiring the local data of the consumption of customer-needed products and external factors includes:

[0013] Deploy sensors in hotels to collect data on the consumption of hotel supplies, and use APIs to obtain real-time weather data, special event information, and external factors such as occupancy rates;

[0014] Integrate the hotel's local data on guest product consumption and external factors, and pre-process the local data, where the pre-processing operations include data cleaning, data sorting, and format conversion steps;

[0015] Based on the preprocessed local data, feature extraction is performed. For the consumption of customer necessities, the types of customer necessities, consumption time series and consumption change rate characteristics are obtained. For weather factors, the temperature, humidity and sunny or rainy conditions characteristics are obtained. For special event factors, the types of special events and special event duration characteristics are obtained. For occupancy rate factors, the occupancy rate time series and occupancy rate change rate characteristics are obtained to obtain a comprehensive feature set.

[0016] Preferably, in the local model training module, the process of acquiring the local model includes:

[0017] Extract the comprehensive feature set, traverse the relevant data of customer demand consumption and external factor characteristics in the comprehensive feature set, and divide the comprehensive feature set into a training set and a test set. The training set is used for model training, and the test set is used for model verification.

[0018] Using the characteristic data of the consumption of customer needs in a short period of time, combined with the characteristic data of external factors such as weather, special events and occupancy rate, the short-term prediction model is trained, and the short-term trend prediction is carried out using the trained short-term prediction model. The weighted moving average algorithm is used to process and analyze the short-term data to obtain the short-term prediction results.

[0019] Analyze historical feature data over a long period of time, identify long-term trends, smooth historical feature data to reduce the impact of data fluctuations on forecast results, train long-term forecast models, conduct long-term forecasts of customer needs, analyze the overall trend of customer needs consumption, and obtain long-term forecast results;

[0020] The analysis results of short-term and long-term forecasts are assigned weights and merged to obtain preliminary forecast results. According to the needs of the hotel, a local model is obtained through comprehensive analysis and uploaded to the central server for summary analysis.

[0021] Preferably, the calculation expression of the short-term prediction result is:

[0022] ;

[0023] ;

[0024] in, For short-term forecasts The predicted consumption of customer demand at a point in time, is the total number of time points, For short-term forecasts The weight of each time point, For the The consumption of customer needs at a point in time, For the Weather data at a certain point in time, For the Special event data at a point in time, For the Occupancy rate data at a certain point in time, is the adjustment factor;

[0025] The calculation expression of the long-term prediction result is:

[0026] ;

[0027] in, For long-term forecasts The predicted consumption of customer demand at a point in time, is the total number of time points, For long-term forecasts The weight of each time point, For the The amount of customer demand consumed at a point in time.

[0028] Preferably, in the global prediction model updating module, the process of updating the global prediction model includes:

[0029] After completing local training, each hotel uploads its local model to the central server through an encrypted channel, and the local models are aggregated to prepare for the update of the global model;

[0030] The central server uses the federated average algorithm to perform weighted averaging on the local models uploaded by all hotels, calculate the average value of the model parameters, and assign different weights to each local model based on the data of each hotel. The weighted average model is calculated using the assigned weights to update the parameters of the global model.

[0031] The global model is updated using the parameters calculated using the federated averaging algorithm, and the updated global model is verified using the retained dataset on the central server to evaluate the performance of the model and calculate the performance metrics of the model.

[0032] Preferably, in the iterative training module, the process of obtaining the final global prediction model includes:

[0033] After completing the update of the global prediction model, the central server will send the updated global prediction model to each hotel through a secure channel to ensure that each hotel can obtain the latest and optimized global prediction model in real time for use in the next round of training. The central server will also assign a version number to the sent global prediction model for easy tracking and management, and simultaneously record the time, version, and recipient information of the global prediction model.

[0034] Each hotel receives the global prediction model sent by the central server and uses the latest local data to further train and adjust the global prediction model.

[0035] After the training is completed, the hotel uploads the updated local model to the central server through an encrypted channel again, repeats the process of local training, global prediction model update and model distribution, forming an iterative training cycle until the global prediction model reaches a convergence state, obtains the final global prediction model, and performs the final guest product prediction to obtain the final predicted value of the guest product demand;

[0036] The calculation expression of the predicted value of the customer demand is:

[0037] ;

[0038] in, is the forecast value of the final customer demand for the product, For short-term forecasts The predicted consumption of customer demand at a point in time, For long-term forecasts The predicted consumption of customer demand at a point in time, is the mixing coefficient of short-term prediction and long-term prediction.

[0039] Preferably, in the anomaly detection module, the anomaly detection process includes:

[0040] The final global prediction model completed through iterative training is deployed in each hotel. The hotel's actual consumption of guest supplies and local data of external factors are collected in real time to detect and analyze the hotel's guest supply demand and generate a predicted value for the guest supply demand.

[0041] Based on the deviation between the predicted value of the demand for guest products and the actual consumption of guest products, the anomaly detection coefficient is calculated, and the anomaly detection threshold is set according to the historical data of each hotel. The anomaly detection coefficient is compared with the anomaly detection threshold. If the anomaly detection coefficient exceeds the set anomaly detection threshold, it is judged as an anomaly;

[0042] Once an abnormality is detected, the system automatically triggers an alarm and adjusts the replenishment strategy based on the abnormal situation. The hotel responds quickly and adjusts the replenishment strategy based on the alarm prompt to ensure sufficient supply of customer needs and meet customer needs while avoiding inventory backlogs and cost losses.

[0043] Record detailed information of abnormal events, including time, predicted value of customer demand, actual consumption of customer demand, and abnormal detection coefficient.

[0044] Preferably, the calculation expression of the abnormality detection coefficient is:

[0045] ;

[0046] ;

[0047] in, is the anomaly detection coefficient, For the The actual consumption of customer needs at a point in time, For the The forecast value of customer demand for products at a point in time, is the anomaly detection threshold.

[0048] Preferably, in the replenishment strategy optimization module, the process of dynamically adjusting the replenishment strategy includes:

[0049] Continuously monitor the real-time consumption and customer demand of customer products, and use the global forecasting model to generate forecasts of customer product demand;

[0050] Analyze market trends, including seasonal changes, holidays, special events, and monitor supply chain status, regularly check current inventory levels and inventory turnover, assess inventory health, and identify risks of excess or shortage;

[0051] Develop replenishment strategies based on demand forecasts and market conditions, and classify inventory to differentiate between fast-moving and slow-moving items, and set safety stock levels for key customer needs to prevent stockouts;

[0052] Analyze replenishment costs, including purchasing costs, transportation costs, and storage costs; analyze replenishment strategy execution data, including replenishment frequency and inventory accuracy; prepare and output replenishment strategy execution reports to provide decision support for management.

[0053] In a second aspect, a smart hotel guest supplies dynamic replenishment optimization method is implemented based on a smart hotel guest supplies dynamic replenishment optimization system, comprising the following steps:

[0054] Step 1: Deploy sensors and integrate API interfaces to collect local data on the consumption of guest supplies and external factors in the hotel in real time, and extract the correlation features of the consumption of guest supplies and external factors after preprocessing the local data to obtain a comprehensive feature set;

[0055] Step 2: Based on the relevant data of the comprehensive feature set, each hotel trains a local model, including a short-term prediction model and a long-term prediction model, and uploads the trained local models of each hotel to the central server. The federated average algorithm is used to perform weighted average on all local models, and the global prediction model is iteratively updated.

[0056] Step 3: Based on the iteratively updated global prediction model and the local data collected in real time, the customer demand for the product is detected and analyzed, and the deviation between the predicted value and the actual value is calculated, and the anomaly detection coefficient is calculated based on the deviation;

[0057] Step 4: Set an anomaly detection threshold based on the historical data of each hotel, compare the anomaly detection coefficient with the anomaly detection threshold, and if the anomaly detection coefficient exceeds the threshold, trigger the optimization adjustment of the replenishment strategy;

[0058] Step 5. The hotel executes replenishment operations according to the optimized replenishment strategy, dynamically adjusts the replenishment strategy of customer needs according to real-time demand changes, and optimizes the inventory structure.

[0059] The present invention provides an intelligent hotel guest needs dynamic replenishment optimization system and method. It has the following beneficial effects:

[0060] 1. The intelligent hotel guest needs dynamic replenishment optimization system and method can reduce inventory backlogs and avoid cost losses caused by excess inventory through accurate demand forecasting. By analyzing customer consumption habits, it can ensure that the hotel can respond to customer requirements in a timely manner and provide high-quality services. When market conditions change, such as the sudden arrival of the peak tourist season or the occurrence of special events, the hotel can quickly adjust its strategy based on the forecasting model. Long-term forecasts can help hotel management formulate longer-term strategic plans.

[0061] 2. The intelligent hotel guest supplies dynamic replenishment optimization system and method, by deploying sensors and integrating API interfaces, does not need to expose its customer data to other hotels or third parties, which helps to protect customers' personal information and the hotel's own business secrets. During the federated learning process, model updates are transmitted through encrypted channels, which increases the security of data transmission.

[0062] 3. The intelligent hotel guest supplies dynamic replenishment optimization system and method, combined with short-term and long-term forecasting models, can more accurately predict room reservations, catering consumption, towel consumption and other demands, helping hotels to better plan resources. In addition, by integrating multiple factors such as weather, holidays, historical data, etc., it can more comprehensively analyze future trends and improve the reliability of forecasts.

[0063] 4. This intelligent hotel guest supplies dynamic replenishment optimization system and method, through federated learning, can share the results of model training between hotels, promote the progress and development of the entire industry, and multiple hotels jointly participate in model training, which can produce synergistic effects and improve the overall prediction level. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a flow chart of an intelligent hotel guest supplies dynamic replenishment optimization system and method of the present invention;

[0065] Figure 2 It is a module structure diagram of the present invention;

[0066] Figure 3 This is a flowchart of abnormality detection of the present invention;

[0067] Figure 4 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0068] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

[0069] The first embodiment, as Figure 1 , Figure 2 As shown, the present invention provides a technical solution: an intelligent hotel guest needs dynamic replenishment optimization system, including a central server, the central server is communicatively connected with a client data acquisition module, a global model initialization module, a local model training module, a global prediction model update module, an iterative training module, an anomaly detection module and a replenishment strategy optimization module, wherein the modules are connected by electrical signals;

[0070] The client data collection module is used to collect the local data of the hotel's guest supplies consumption and external factors in real time. External factors include weather, special events and occupancy rate. Through real-time data collection, it ensures that the latest inventory and sales can be obtained, providing basic data for subsequent analysis and prediction, and pre-processing operations such as cleaning, sorting and format conversion are performed on the collected local data to improve data quality and availability. Through pre-processing, noise and outliers in the data are eliminated to provide more accurate data support for subsequent analysis and prediction. Sensors are deployed in the hotel to collect the consumption data of guest supplies in the hotel, and the API interface is used to obtain real-time weather data, special event information and external factor data of occupancy rate, integrating the hotel's guest supplies consumption and external factors. The local data of internal factors is collected and preprocessed. The preprocessing operation includes data cleaning, data sorting and format conversion steps. The collected local data is cleaned, including removing duplicate data, processing missing values, correcting erroneous data, etc., to improve data quality and provide accurate data support for subsequent analysis and prediction. The cleaned data is sorted through data sorting and stored in the specified format. The data is classified according to different types of customer needs and sorted in chronological order. The data format is converted, the text data is converted into numerical data, and the data of different units are unified. Based on the preprocessed local data, feature extraction is performed. For the consumption of customer needs, the types and consumption of customer needs are obtained. Time series and consumption change rate characteristics. The characteristics of customer necessities, such as shampoo, shower gel, toothbrush, towels, etc., have different consumption patterns. The consumption time series characteristics record the consumption of each customer necessities within a period of time (daily, weekly, monthly) to form time series data for analyzing consumption trends. The consumption change rate characteristics calculate the rate of change of consumption in adjacent time periods to reflect the change in consumption speed. For weather factors, obtain temperature, humidity, and sunny or rainy conditions characteristics. In the temperature characteristics, high or low temperatures affect the guests' water consumption and bathing frequency, which in turn affects the consumption of customer necessities. In the humidity characteristics, humidity affects the replacement frequency of customer necessities such as towels. In the sunny or rainy conditions characteristics, sunny or rainy days affect the guests' outdoor activity time. The consumption of guest supplies is indirectly affected. For special event factors, the special event type and special event duration characteristics are obtained. Different special events (such as conferences, exhibitions, festivals, etc.) have different celebration methods and activity arrangements in the special event type characteristics, which affect the guests' accommodation time and consumption behavior. The duration of special events in the special event duration characteristics affects the guests' accommodation plans and the demand for guest supplies. For occupancy rate factors, the occupancy rate time series and occupancy rate change rate characteristics are obtained. The occupancy rate time series characteristics record the hotel's daily or weekly occupancy rate, which is used to analyze the relationship between occupancy rate and consumption of guest supplies. The occupancy rate change rate characteristics calculate the change rate of occupancy rate in adjacent time periods to reflect the changing trend of hotel occupancy and obtain a comprehensive feature set.

[0071] The global model initialization module is used to initialize the global prediction model on the central server, including short-term and long-term prediction components and related weights and adjustment factors, providing a basic framework for the entire system to ensure the accuracy and flexibility of the prediction;

[0072] In the local model training module, each hotel uses the pre-processed local data to perform short-term and long-term forecasting training. The short-term forecast is to combine the data in the recent period with external factors to make trend forecasts. The long-term forecast is to analyze the historical data with a long span, identify the long-term trend, improve the accuracy and flexibility of the forecast, and adapt to the influence of seasonal changes and special events. The local model is obtained through comprehensive analysis, and the comprehensive feature set is extracted. The relevant data of the consumption of customer needs and the characteristics of external factors in the comprehensive feature set are traversed, and the comprehensive feature set is divided into a training set and a test set. The training set is used for model training, and the test set is used for model verification. The characteristic data of the consumption of customer needs in a short period of time (such as the last few weeks or months) is combined with weather, special events and entry. The short-term prediction model is trained with the characteristic data of external factors of occupancy rate, and the short-term trend prediction is carried out by using the trained short-term prediction model. The weighted moving average algorithm is used to process and analyze the data in the short term to obtain the short-term prediction results. The historical characteristic data of a long time span (such as one or more years) is analyzed to identify the long-term trend. The historical characteristic data is smoothed to reduce the impact of data fluctuations on the prediction results. The long-term prediction model is trained to make long-term predictions of guest needs. The overall trend of the consumption of guest needs is analyzed to obtain long-term prediction results. The analysis results of the short-term prediction and the long-term prediction are assigned weights and merged to obtain preliminary prediction results. According to the needs of the hotel, a local model is obtained through comprehensive analysis, and the local model is uploaded to the central server for summary analysis;

[0073] Furthermore, the calculation expression of the short-term prediction result is:

[0074] ;

[0075] ;

[0076] in, For short-term forecasts The predicted consumption of customer demand at a point in time, is the total number of time points, For short-term forecasts The weight of each time point, For the The consumption of customer needs at a point in time, For the Weather data at a certain point in time, For the Special event data at a point in time, For the Occupancy rate data at a certain point in time, is the adjustment factor;

[0077] The calculation expression of long-term prediction results is:

[0078] ;

[0079] in, For long-term forecasts The predicted consumption of customer demand at a point in time, is the total number of time points, For long-term forecasts The weight of each time point, For the The consumption of customer needs at a point in time;

[0080] The global prediction model update module is used to aggregate the local models uploaded by all hotels on the central server and use the federated average algorithm to update the global prediction model. By integrating the data of multiple hotels, the prediction accuracy and robustness of the global prediction model are improved. After completing local training, each hotel uploads its local model to the central server through an encrypted channel, aggregates the local models and prepares to update the global model. The central server uses the federated average algorithm to perform weighted averaging on the local models uploaded by all hotels, calculates the average value of the model parameters, and assigns different weights to each local model based on the data of each hotel. The weighted average model is calculated using the assigned weights, and the parameters of the global model are updated. The global model is updated using the parameters calculated using the federated average algorithm. The updated global model is verified using the retained data set on the central server, the performance of the model is evaluated, and the performance indicators of the model are calculated.

[0081] In the iterative training module, the central server sends the updated global prediction model to each hotel for use in the next round of training, ensuring that each hotel can obtain the latest global prediction model in real time, and repeat the process of local training, global prediction model update and model distribution until the model converges and obtains the final global prediction model. Through continuous iterative training, the prediction model is optimized to improve the accuracy and stability of the replenishment strategy. After completing the update of the global prediction model, the central server sends the updated global prediction model to each hotel through a secure channel, ensuring that each hotel can obtain the latest and optimized global prediction model in real time for use in the next round of training. A version number is assigned to the global prediction model for easy tracking and management. The time, version and recipient information of the global prediction model are recorded synchronously. Each hotel receives the global prediction model sent by the central server and uses the latest local data to further train and adjust the global prediction model. After the training is completed, the hotel uploads the updated local model to the central server through an encrypted channel again, and repeats the process of local training, global prediction model update and model sending to form an iterative training cycle until the global prediction model reaches a convergence state, and the final global prediction model is obtained. The final guest product forecast is carried out to obtain the predicted value of the final guest product demand;

[0082] The calculation expression for the predicted value of customer demand is:

[0083] ;

[0084] in, is the forecast value of the final customer demand for the product, For short-term forecasts The predicted consumption of customer demand at a point in time, For long-term forecasts The predicted consumption of customer demand at a point in time, is the mixing coefficient of short-term and long-term forecasts;

[0085] The anomaly detection module detects and analyzes the demand for hotel guest supplies based on the iteratively trained global prediction model and the real-time collected local data, and sets anomaly detection thresholds to perform anomaly detection;

[0086] The replenishment strategy optimization module dynamically adjusts the replenishment strategy according to real-time demand changes and market conditions, optimizes the inventory structure, reduces inventory costs, and ensures that customer needs are met in a timely manner.

[0087] The second embodiment is based on the first embodiment. Figure 3 As shown in the figure, in the anomaly detection module, the anomaly detection process includes:

[0088] The final global prediction model completed by iterative training is deployed in each hotel. The hotel's actual guest product consumption and local data of external factors are collected in real time to detect and analyze the hotel's guest product demand, generate a predicted value of the guest product demand, calculate the anomaly detection coefficient based on the deviation between the predicted value of the guest product demand and the actual guest product consumption, and set the anomaly detection threshold according to the historical data of each hotel. The anomaly detection coefficient is compared with the anomaly detection threshold. If the anomaly detection coefficient exceeds the set anomaly detection threshold, it is judged as an anomaly. Once an anomaly is detected, the system automatically triggers an alarm and triggers the adjustment of the replenishment strategy according to the abnormal situation. The hotel responds quickly and adjusts the replenishment strategy according to the alarm prompt to ensure sufficient supply of guest products to meet customer needs while avoiding inventory backlogs and cost losses. The detailed information of the abnormal event is recorded, including time, predicted value of guest product demand, actual guest product consumption, and anomaly detection coefficient;

[0089] Furthermore, the calculation expression of the anomaly detection coefficient is:

[0090] ;

[0091] ;

[0092] in, is the anomaly detection coefficient, For the The actual consumption of customer needs at a point in time, For the The forecast value of customer demand for products at a point in time, is the anomaly detection threshold;

[0093] In the replenishment strategy optimization module, the process of dynamically adjusting the replenishment strategy includes:

[0094] Continuously monitor the real-time consumption and customer demand of customer-needed items, and use the global forecasting model to generate forecasts of customer-needed item demand; analyze market trends, including seasonal changes, holidays, and special events; monitor the status of the supply chain, regularly check current inventory levels and inventory turnover rates, assess the health of inventory, identify risks of excess or shortages, formulate replenishment strategies based on demand forecast results and market conditions, classify inventory, distinguish between fast-moving and slow-moving items, set safety inventory levels for key customer-needed items to prevent out-of-stocks, analyze replenishment costs, including procurement costs, transportation costs, and storage costs, analyze replenishment strategy execution data, including replenishment frequency and inventory accuracy, prepare and output replenishment strategy execution reports, and provide decision support for management.

[0095] The third embodiment is based on the first and second embodiments. Figure 4As shown, the present invention also provides a technical solution: a smart hotel guest needs dynamic replenishment optimization method, which is implemented based on a smart hotel guest needs dynamic replenishment optimization system, comprising the following steps:

[0096] Step 1: Deploy sensors and integrate API interfaces to collect local data on the consumption of guest supplies and external factors in the hotel in real time, and extract the correlation features of the consumption of guest supplies and external factors after preprocessing the local data to obtain a comprehensive feature set;

[0097] Step 2: Based on the relevant data of the comprehensive feature set, each hotel trains a local model, including a short-term prediction model and a long-term prediction model, and uploads the trained local models of each hotel to the central server. The federated average algorithm is used to perform weighted average on all local models, and the global prediction model is iteratively updated.

[0098] Step 3: Based on the iteratively updated global prediction model and the local data collected in real time, the customer demand for the product is detected and analyzed, and the deviation between the predicted value and the actual value is calculated, and the anomaly detection coefficient is calculated based on the deviation;

[0099] Step 4: Set an anomaly detection threshold based on the historical data of each hotel, compare the anomaly detection coefficient with the anomaly detection threshold, and if the anomaly detection coefficient exceeds the threshold, trigger the optimization adjustment of the replenishment strategy;

[0100] Step 5. The hotel executes replenishment operations according to the optimized replenishment strategy, dynamically adjusts the replenishment strategy of customer needs according to real-time demand changes, and optimizes the inventory structure.

[0101] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.

Claims

1. An intelligent hotel guest needs dynamic replenishment optimization system, including a central server, characterized in that: The central server is communicatively connected with a client data acquisition module, a global model initialization module, a local model training module, a global prediction model update module, an iterative training module, an anomaly detection module, and a replenishment strategy optimization module, wherein electrical signals are connected between the modules; The client data collection module is used to collect the local data of the hotel's guest product consumption and external factors in real time, the external factors include weather, special events and occupancy rate, and pre-process the collected local data. In the client data collection module, the process of obtaining the local data of guest product consumption and external factors includes: Deploy sensors in hotels to collect consumption data of hotel supplies, and use API interfaces to obtain real-time weather data, special event information, and external factors such as occupancy rate; Integrate the hotel's local data on guest product consumption and external factors, and pre-process the local data, where the pre-processing operations include data cleaning, data sorting, and format conversion steps; Based on the pre-processed local data, feature extraction is performed. For the consumption of customer needs, the types of customer needs, consumption time series and consumption change rate characteristics are obtained. For weather factors, the temperature, humidity and sunny and rainy conditions are obtained. For special event factors, the types of special events and special event duration characteristics are obtained. For occupancy factors, the occupancy time series and occupancy change rate characteristics are obtained to obtain a comprehensive feature set. The global model initialization module is used to initialize the global prediction model on the central server, including short-term and long-term prediction components and related weights and adjustment factors; In the local model training module, each hotel uses the pre-processed local data to perform short-term and long-term prediction training, and obtains a local model through comprehensive analysis. In the local model training module, the process of obtaining the local model includes: Extract the comprehensive feature set, traverse the relevant data of the consumption of customer demand products and external factor characteristics in the comprehensive feature set, and divide the comprehensive feature set into a training set and a test set; Using the characteristic data of the consumption of customer needs in a short period of time, combined with the characteristic data of external factors such as weather, special events and occupancy rate, the short-term prediction model is trained, and the short-term trend prediction is carried out using the trained short-term prediction model. The weighted moving average algorithm is used to process and analyze the short-term data to obtain the short-term prediction results. Analyze historical feature data over a long period of time, identify long-term trends, smooth historical feature data, train long-term prediction models, conduct long-term predictions of customer needs, analyze the overall trend of customer needs consumption, and obtain long-term prediction results; The analysis results of short-term prediction and long-term prediction are assigned weights and merged to obtain preliminary prediction results. According to the needs of the hotel, a local model is obtained through comprehensive analysis, and the local model is uploaded to the central server for summary analysis. The calculation expression of the short-term prediction result is: ； ； in, For short-term forecasts The predicted consumption of customer demand at a point in time, is the total number of time points, For short-term forecasts The weight of each time point, For the The consumption of customer needs at a point in time, For the Weather data at a certain point in time, For the Special event data at a point in time, For the Occupancy rate data at a certain point in time, is the adjustment factor; The calculation expression of the long-term prediction result is: ； in, For long-term forecasts The predicted consumption of customer demand at a point in time, is the total number of time points, For long-term forecasts The weight of each time point, For the The consumption of customer needs at a point in time; The global prediction model update module is used to aggregate the local models uploaded by all hotels on the central server and use the federated average algorithm to update the global prediction model; In the iterative training module, the central server sends the updated global prediction model to each hotel, and repeats the process of local training, global prediction model update and model sending until the model converges to obtain the final global prediction model; The anomaly detection module detects and analyzes the demand for hotel guest supplies based on the iteratively trained global prediction model and the local data collected in real time, and sets an anomaly detection threshold to perform anomaly detection; The replenishment strategy optimization module dynamically adjusts the replenishment strategy according to real-time demand changes and market conditions.

2. According to claim 1, a smart hotel guest needs dynamic replenishment optimization system is characterized by: In the global prediction model updating module, the process of updating the global prediction model includes: After completing local training, each hotel uploads its local model to the central server through an encrypted channel, and the local models are aggregated to prepare for the update of the global model; The central server uses the federated average algorithm to perform weighted averaging on the local models uploaded by all hotels, calculate the average value of the model parameters, and assign different weights to each local model based on the data of each hotel. The weighted average model is calculated using the assigned weights to update the parameters of the global model. The global model is updated using the parameters calculated using the federated averaging algorithm, and the updated global model is verified using the retained dataset on the central server to evaluate the performance of the model and calculate the performance metrics of the model.

3. According to claim 2, a smart hotel guest needs dynamic replenishment optimization system is characterized by: In the iterative training module, the process of obtaining the final global prediction model includes: After completing the update of the global prediction model, the central server will send the updated global prediction model to each hotel through a secure channel, assign a version number to the sent global prediction model, and simultaneously record the time, version, and recipient information of the global prediction model; Each hotel receives the global prediction model sent by the central server and uses the latest local data to further train and adjust the global prediction model. After the training is completed, the hotel uploads the updated local model to the central server through an encrypted channel again, repeats the process of local training, global prediction model update and model distribution, forming an iterative training cycle until the global prediction model reaches a convergence state, obtains the final global prediction model, and performs the final guest product prediction to obtain the final predicted value of the guest product demand; The calculation expression of the predicted value of the customer demand is: ； in, is the forecast value of the final customer demand for the product, For short-term forecasts The predicted consumption of customer demand at a point in time, For long-term forecasts The predicted consumption of customer demand at a point in time, is the mixing coefficient of short-term prediction and long-term prediction.

4. According to claim 3, a smart hotel guest needs dynamic replenishment optimization system is characterized by: In the anomaly detection module, the anomaly detection process includes: The final global prediction model completed through iterative training is deployed in each hotel. The hotel's actual consumption of guest supplies and local data of external factors are collected in real time to detect and analyze the hotel's guest supply demand and generate a predicted value for the guest supply demand. Based on the deviation between the predicted value of the demand for guest products and the actual consumption of guest products, the anomaly detection coefficient is calculated, and the anomaly detection threshold is set according to the historical data of each hotel. The anomaly detection coefficient is compared with the anomaly detection threshold. If the anomaly detection coefficient exceeds the set anomaly detection threshold, it is judged as an anomaly; Once an abnormality is detected, the system automatically triggers an alarm and adjusts the replenishment strategy based on the abnormal situation. The hotel responds quickly and adjusts the replenishment strategy based on the alarm prompt; Record detailed information of abnormal events, including time, predicted value of customer demand, actual consumption of customer demand, and abnormal detection coefficient.

5. According to claim 4, a smart hotel guest needs dynamic replenishment optimization system is characterized by: The calculation expression of the abnormality detection coefficient is: ； ； in, is the anomaly detection coefficient, For the The actual consumption of customer needs at a point in time, For the The forecast value of customer demand for products at a point in time, is the anomaly detection threshold.

6. According to claim 5, a smart hotel guest needs dynamic replenishment optimization system is characterized by: In the replenishment strategy optimization module, the process of dynamically adjusting the replenishment strategy includes: Continuously monitor the real-time consumption and customer demand of customer products, and use the global forecasting model to generate forecasts of customer product demand; Analyze market trends, including seasonal changes, holidays, special events, and monitor supply chain status, regularly check current inventory levels and inventory turnover, assess inventory health, and identify risks of excess or shortage; Develop replenishment strategies based on demand forecasts and market conditions, and classify inventory to differentiate between fast-moving and slow-moving items, and set safety stock levels for key customer needs; Analyze replenishment costs, including purchasing costs, transportation costs, and storage costs; analyze replenishment strategy execution data, including replenishment frequency and inventory accuracy; and prepare and output replenishment strategy execution reports.

7. An intelligent hotel guest supplies dynamic replenishment optimization method, implemented based on the intelligent hotel guest supplies dynamic replenishment optimization system according to any one of claims 1 to 6, characterized in that: The following steps are involved: Step 1: Deploy sensors and integrate API interfaces to collect local data on the consumption of guest supplies and external factors in the hotel in real time, and extract the correlation features of the consumption of guest supplies and external factors after preprocessing the local data to obtain a comprehensive feature set; Step 2: Based on the relevant data of the comprehensive feature set, each hotel trains a local model, including a short-term prediction model and a long-term prediction model, and uploads the trained local models of each hotel to the central server. The federated average algorithm is used to perform weighted average on all local models, and the global prediction model is iteratively updated. Step 3: Based on the iteratively updated global prediction model and the local data collected in real time, the customer demand for the product is detected and analyzed, and the deviation between the predicted value and the actual value is calculated, and the anomaly detection coefficient is calculated based on the deviation; Step 4: Set an anomaly detection threshold based on the historical data of each hotel, compare the anomaly detection coefficient with the anomaly detection threshold, and if the anomaly detection coefficient exceeds the threshold, trigger the optimization adjustment of the replenishment strategy; Step 5. The hotel executes replenishment operations according to the optimized replenishment strategy, dynamically adjusts the replenishment strategy of customer needs according to real-time demand changes, and optimizes the inventory structure.

Images