A catering service data intelligent analysis method and system

Abstract

This invention provides a method and system for intelligent analysis of catering service data, belonging to the field of data analysis and processing technology. The method includes: collecting historical ordering service records of the restaurant and performing table-type association analysis; constructing a time-series curve of the dish structure for each table type; performing global time-segment ordering distribution analysis for each table type and generating multiple global time-segment structure lists; constructing multiple crowd density distribution vectors; optimizing the crowd density distribution of the global time-segment structure lists; generating local time-segment structure lists for each table type; determining multiple dish types in the restaurant and performing local dish popularity analysis; constructing multiple local popularity lists; constructing multiple dish association graphs for each table type based on historical ordering service records; generating multiple target dish combinations based on the dish association graphs; and obtaining global dish combination recommendation data for each table type. This invention significantly improves the accuracy of dish recommendations and the dining experience for customers.

Description

Technical Field

[0001] This invention relates to the field of data analysis and processing technology, and in particular to a method and system for intelligent analysis of catering service data. Background Technology

[0002] With the widespread application of information and intelligent technologies in the catering industry, catering service data has gradually become an important resource for promoting refined restaurant operations and personalized services. Among them, ordering data, as an important component of catering service data, can reflect multi-dimensional information such as customer consumption preferences, dining habits, time characteristics, and table type characteristics, which is of great value for dish recommendations, menu optimization, and marketing strategy formulation.

[0003] Some technologies analyze food popularity based on a single indicator such as the total number of orders, which tends to identify highly popular dishes. However, this approach can easily lead to a certain category of highly popular dishes dominating the analysis results, thus masking relatively popular dishes in other categories. Furthermore, customer ordering behavior is significantly influenced by factors such as time of day, table type, and dining environment. For example, customers tend to prefer convenient dishes during lunchtime, while they may choose more complex and elaborate dishes during dinner; customers at small tables typically opt for simple combinations, while customers at large tables tend to prefer more diverse combinations including soups and cold dishes.

[0004] Therefore, there is an urgent need for an intelligent analysis method and system that can comprehensively analyze multi-dimensional features such as dish categories, time periods, table types, and pairing patterns in catering service data, so as to provide more accurate data support for merchants in various application scenarios such as formulating set menus, carrying out advertising promotions, and displaying dishes in real time. Summary of the Invention

[0005] The present invention provides a method and system for intelligent analysis of catering service data, which aims to solve at least one technical problem existing in the above-mentioned background art.

[0006] To achieve the above objectives, the first aspect of the present invention provides a method for intelligent analysis of catering service data, comprising:

[0007] Collect historical ordering service records from the restaurant, including table type and dish information for multiple orders, perform dish-table type correlation analysis on the historical ordering service records, and construct a time-series curve of the dish structure for each table type;

[0008] Based on the time series curve of the dish structure, a global time period order distribution analysis is performed on each table type to generate a global time period structure list for each table type, including the first candidate dish structure corresponding to multiple global time periods;

[0009] Based on historical ordering service records, multiple crowd density distribution vectors are constructed for each global time period. Based on the crowd density distribution vectors, the crowd density distribution of the global time period structure list is optimized to generate a local time period structure list for each table type, including the second candidate dish structure for multiple local crowd density intervals.

[0010] Identify multiple dish types in the restaurant and conduct local analysis of dish popularity based on local time period structure lists, constructing local popularity lists for multiple dish types under each table type and each local flow density interval;

[0011] Based on historical ordering service records, a dish association map is constructed for each table type in each local crowd density range. Target dish combinations for each local crowd density range are generated based on the dish association map, resulting in global dish combination recommendation data for each table type.

[0012] Preferably, the global time-slot structure list is optimized based on the crowd density distribution vector to generate a local time-slot structure list for each table type, including:

[0013] For each first candidate dish structure, the crowd density distribution vector corresponding to the global time period is used to determine the crowd density distribution interval for each global time period based on multiple crowd density distribution vectors.

[0014] Based on historical ordering service records and the distribution range of crowd density, multiple local structure frequency distribution vectors corresponding to the structure of the first candidate dish for each table type in the global time period are extracted. The local structure frequency distribution vectors include the structure frequency parameters corresponding to multiple local crowd density intervals.

[0015] Multiple local structure frequency distribution vectors of each first candidate dish structure are fused to generate a global structure frequency distribution vector corresponding to the first candidate dish structure of each table type;

[0016] Based on the global structural frequency distribution vector, the target local pedestrian density interval to which each first candidate dish structure belongs is determined. Based on the target local pedestrian density intervals to which multiple first candidate dish structures belong, multiple second candidate dish structures for each local pedestrian density interval are determined.

[0017] Preferably, a food association map is constructed for each table type in each local crowd density range based on historical ordering service records, including:

[0018] Based on the structure of multiple second candidate dishes in local crowd density intervals, determine multiple dish types associated with each local crowd density interval, and determine multiple dish association graph nodes based on the multiple dish types associated with local crowd density intervals;

[0019] Based on the global time period and table type to which the local crowd density interval belongs, extract the local dish association sample data corresponding to the local crowd density interval from the historical food ordering service data;

[0020] Based on the local dish association sample data, calculate multiple association weights of multiple dish association graph nodes in the local flow density interval. Based on the multiple dish association graph nodes and multiple association weights, construct the dish association graph for each table type in each local flow density interval.

[0021] Preferably, the target dish combination for each local crowd density interval is generated based on the dish association map, including:

[0022] Based on the local popularity list of multiple dish types under each local flow density interval for each table type, multiple candidate dish combinations are generated under each local flow density interval for each table type. The structural stability of each candidate dish combination is detected by the dish association graph, and the structural stability index of each candidate dish combination is calculated. Based on the structural stability index, multiple target dish combinations are determined from the multiple candidate dish combinations under the local flow density interval.

[0023] Preferably, based on historical ordering service records and the distribution range of customer flow density, multiple local structure frequency distribution vectors of the first candidate dish structure for each table type are extracted for the corresponding global time period, including:

[0024] Based on the population density distribution interval, multiple local population density intervals are determined for each global time period. Each global time period is divided into multiple local time periods. Based on historical ordering service records and population density distribution intervals, multiple local population density intervals and multiple structural frequency parameters are determined for each local time period. Based on the multiple local population density intervals and structural frequency parameters, multiple local structural frequency distribution vectors corresponding to each global time period are constructed.

[0025] Preferably, structural stability testing is performed on each candidate dish combination based on the dish association graph, and the structural stability index of each candidate dish combination is calculated, including:

[0026] The association weights between any two dishes in the candidate dish combination are determined based on the dish association graph, and the average of multiple association weights is calculated as the structural stability index of the candidate dish combination.

[0027] A second aspect of the present invention provides a catering service data intelligent analysis system for implementing the above-mentioned catering service data intelligent analysis method, comprising:

[0028] The food table type association analysis module is used to collect historical ordering service records of the restaurant, including table type information and food information of multiple orders. It performs food table type association analysis on the historical ordering service records data and constructs a time series curve of food structure for each table type.

[0029] The global time-segment ordering analysis module is used to perform global time-segment ordering distribution analysis for each table type based on the time-series curve of the dish structure, and generate a global time-segment structure list for each table type, including the first candidate dish structure corresponding to multiple global time periods.

[0030] The order distribution optimization module is used to construct multiple crowd density distribution vectors for each global time period based on historical order service record data. Based on the crowd density distribution vectors, the module optimizes the crowd density distribution of the global time period structure list and generates a local time period structure list for each table type, including the second candidate dish structure for multiple local crowd density intervals.

[0031] The local dish popularity analysis module is used to identify multiple dish types in a restaurant and perform local dish popularity analysis based on a local time period structure list, constructing a local popularity list of multiple dish types under each table type and each local flow density interval;

[0032] The menu combination recommendation and analysis module is used to construct a menu association map for each table type in each local crowd density range based on historical ordering service records. Based on the menu association map, the module generates target menu combinations for each local crowd density range, thus obtaining global menu combination recommendation data for each table type.

[0033] The present invention has the following beneficial effects:

[0034] This invention collects and analyzes historical ordering service records from restaurants to construct a time-series curve reflecting the changes in dish combinations for different table types over time. It divides the entire time period according to the business scenario to generate a preliminary first-candidate dish structure. Further, it introduces a pedestrian density distribution vector to further segment the pedestrian flow within the entire time period, considering the impact of pedestrian flow changes on ordering. Based on the structural frequency distribution characteristics, it generates a more refined second-candidate dish structure. On this basis, it analyzes the popularity of dishes and the correlation characteristics between dishes, constructs a dish correlation graph using historical data, and finally selects target dish combinations that highly match the ordering habits of the target group by combining a local popularity list. By comprehensively considering multiple factors such as dining time, table type, pedestrian density, and dish correlation, it achieves deep intelligent mining of historical data, providing restaurants with a highly adaptable and reasonably matched decision-making basis for dynamic menu recommendations, precise set meal design, and marketing strategy optimization, significantly improving the accuracy of dish recommendations and the dining experience of the customer group. Attached Figure Description

[0035] Figure 1This is a flowchart illustrating an intelligent data analysis method for catering services provided in an embodiment of the present invention.

[0036] Figure 2 This is a schematic diagram of the structure of a catering service data intelligent analysis system provided in an embodiment of the present invention. Detailed Implementation

[0037] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention.

[0038] Please see Figure 1 The first aspect of this invention provides a method for intelligent analysis of catering service data, comprising the following steps:

[0039] Step S1: Collect historical ordering service records of the restaurant, including table type information and dish information for multiple orders. Perform dish and table type correlation analysis on the historical ordering service records data and construct a time-series curve of the dish structure for each table type.

[0040] In this step, data related to the restaurant's ordering service can be collected over a specific historical period, such as the past three or six months. This data should include at least order identification information, table type information, and dish information. Order identification information may include order numbers, order placement times, and settlement times for different orders. Table type information may include table numbers, table type and capacity (e.g., two-person, four-person, six-person tables), and dish information may include dish numbers, dish names, dish categories, and prices (e.g., meat main courses, seafood, vegetables, soups, snacks, desserts).

[0041] It is worth noting that the specific table type and dish category classification can be determined based on the restaurant's actual menu structure and business practices. Different restaurants have different business positioning, menu design, and customer structure, so their dish classification systems may differ and are not limited to the exemplary classifications given in this embodiment. This embodiment does not impose specific limitations on them. The main purpose of this invention is to deeply analyze the differences in customer ordering tendencies corresponding to different table sizes. For example, large tables generally tend to order more types and larger portions of dishes, while small tables tend to have a more concise combination of dishes and generally do not cover a large number of dish categories. Considering this difference, the suitability of popular dishes in the restaurant as a whole to users may need to comprehensively consider information such as dining time, number of diners, and individual user preferences, rather than simply recommending dishes based on the restaurant's most popular dishes.

[0042] The collected historical data allows for correlation analysis of table type information and dish types in different orders. This analysis can statistically analyze the frequency, quantity ratio, and time distribution of various dishes at different table types, ultimately generating a time-series curve of the dish structure for each table type based on extensive historical data. This curve reflects the changes in the dish category structure of a particular table type over different time periods. For example, for a two-person table in a restaurant, the combination of dish types and their corresponding quantities appearing every hour constitutes a curve showing the change in the dish structure combination of each table type over time.

[0043] Step S2: Perform global time-slot order distribution analysis for each table type based on the time-series curve of the dish structure, and generate a global time-slot structure list for each table type, including the first candidate dish structure corresponding to multiple global time slots.

[0044] In this step, based on the aforementioned time-series curve of the menu structure, the restaurant's operating hours can be divided into several global time periods. These global time periods are mainly used to capture differences in menu combinations over a large time span. For example, based on business needs, the corresponding time periods for lunch, dinner, and late-night snacks are determined and recorded as global time periods. Examples include 11:00–14:00 and 16:00–20:00. Within each global time period, the menu category combinations and corresponding proportions for each table type are statistically analyzed to obtain the mainstream menu structure characteristics for that time period. Finally, based on the statistical results, a global time period structure list for each table type is generated. Each global time period corresponds to multiple first-candidate menu structures. For example, during lunchtime, a two-person table may frequently feature combinations such as a meat main dish + vegetables, or a meat main dish + soup.

[0045] Step S3: Construct multiple flow density distribution vectors for each global time period based on historical ordering service records. Optimize the flow density distribution of the global time period structure list based on the flow density distribution vectors to generate a local time period structure list for each table type.

[0046] It's worth noting that the actual customer flow or density at a restaurant can also influence customer ordering behavior. For example, during peak hours, some customers may prefer dishes that are served quickly and are easy to eat, in order to shorten waiting time and improve dining efficiency; while during low-peak hours, customers may prefer dishes that are more complex to prepare and take longer to eat, in order to enrich their dining experience.

[0047] In this step, considering that differences in actual customer traffic may affect user ordering, we further subdivide each global time period based on indicators such as historical order volume, number of customers visiting the restaurant, and table occupancy rate. This determines the restaurant's crowd density characteristics within different local time periods, constructing a crowd density distribution vector for each global time period. We also further analyze the impact of crowd density changes on the menu structure. For example, among multiple first-candidate menu structures, will there be local differences among user groups under different customer traffic levels? For instance, will low customer traffic lead to a preference for menu structures ABC, while high customer traffic might lead to a preference for menu structures EF? Ultimately, we optimize and adjust the global time period structure list to generate a more granular local time period structure list. The local time period structure list for each table type includes second-candidate menu structures corresponding to multiple crowd density intervals, which can more accurately characterize the ordering features under specific customer traffic conditions.

[0048] As an example implementation process, the global time-slot structure list is optimized based on the pedestrian density distribution vector to generate a local time-slot structure list for each table type, specifically including:

[0049] Step S31: For each first candidate dish structure, determine the crowd density distribution interval for each global time period based on the crowd density distribution vectors.

[0050] Specifically, for each first candidate dish, the restaurant's foot traffic over time is determined based on historical ordering records. For example, foot traffic density features are extracted from the data for each day's global time period to construct a foot traffic density distribution vector for that daily global time period. All foot traffic density features are then aggregated to obtain the range of foot traffic density changes within that global time period, i.e., the foot traffic density distribution interval.

[0051] Step S32: Based on historical ordering service records and the distribution range of crowd density, extract the multiple local structure frequency distribution vectors of the first candidate dish structure for each table type in the corresponding global time period.

[0052] Specifically, for the pedestrian density distribution interval, multiple local pedestrian density intervals can be further divided at equal intervals. Similarly, for each global time period, multiple local time periods can be further divided at equal intervals. For example, the global time period from 11:00 to 14:00 is divided into local time periods every half hour. For the period from 11:00 to 14:00 every day, the structural frequency parameter of a specific table type in the restaurant under a specific dish structure is determined under each local method. For example, the ratio of the number of times a meat main dish + vegetable dish structure appears on a small table in the restaurant from 11:00 to 11:30 to the total number of times all dish structures appear in the global time period is used as the structural frequency parameter under the specific dish structure. The local pedestrian density interval corresponding to the pedestrian density under each local time period is further determined. Finally, by combining the local pedestrian density interval and structural frequency parameter of each local time period, multiple first candidate dish structures of each table type can be constructed, and multiple local structural frequency distribution vectors corresponding to the global time period of each table type can be obtained. These vectors include the structural frequency parameters corresponding to the multiple local pedestrian density intervals. Specifically, a local structural frequency distribution vector can be constructed from the data of each day in historical data.

[0053] Step S33: Fuse the multiple local structure frequency distribution vectors of each first candidate dish structure to generate the global structure frequency distribution vector corresponding to the first candidate dish structure of each table type.

[0054] The process of fusing multiple local structural frequency distribution vectors of the first candidate dish structure mainly involves further determining the variation in the frequency of occurrence of the first candidate dish structure under different local pedestrian density intervals. Specifically, this can be achieved by summarizing multiple structural frequency parameters under each local pedestrian density interval from the multiple local structural frequency distribution vectors, calculating their mean as a global structural distribution feature, and finally combining the global structural distribution features of the first candidate dish structure under different local pedestrian density intervals to generate the global structural frequency distribution vector corresponding to the first candidate dish structure. This vector is used to indicate the impact of the dish structure on changes in restaurant pedestrian flow under specific demand scenarios, such as lunchtime scenarios.

[0055] Step S34: Determine the target local crowd density interval to which each first candidate dish structure belongs based on the global structural frequency distribution vector. Based on the target local crowd density intervals to which multiple first candidate dish structures belong, determine multiple second candidate dish structures for each local crowd density interval.

[0056] The target local pedestrian density interval is primarily used to characterize typical pedestrian flow scenarios where the first candidate dish structure appears. Specifically, it can be sorted based on the global structural distribution characteristics of multiple local pedestrian density intervals of the first candidate dish structure, and a preset number of local pedestrian density intervals are selected and marked as target local pedestrian density intervals. For example, if the pedestrian density distribution interval is divided into six local pedestrian density intervals, the two with the largest global structural distribution characteristics are selected and marked as target local pedestrian density intervals. Finally, based on the target local pedestrian density interval to which the first candidate dish structure belongs, the representative first candidate dish structure under each local pedestrian density interval is determined and marked as the second candidate dish structure, i.e., a more granular representative candidate dish structure under different customer flow conditions.

[0057] It is worth noting that, in the process of determining the first candidate dish structure, those skilled in the art can think of screening out some discrete samples with low frequency of occurrence, and using multiple representative dish structures as the first candidate dish structure to ensure that the final determined second candidate dish structure is more representative and reduce the perception bias of group behavior that may be caused by small groups.

[0058] Step S4: Determine the multiple dish types of the restaurant and conduct local analysis of dish popularity based on the local time period structure list, and construct a local popularity list of multiple dish types under each table type and each local flow density interval.

[0059] In this step, based on table type, and considering the multiple sub-dishes under each dish category (e.g., the multiple dishes a restaurant can offer in the vegetable category), the popularity data for each dish type is determined within each local foot traffic density interval. For example, the popularity of each dish can be represented by the number of purchases. Specifically, the number of orders for each sub-dish in each local time period can be determined. Combined with the corresponding local foot traffic density interval for each time period, the total number of orders for each sub-dish in each local foot traffic density interval is finally calculated, normalized, and used as the popularity feature of the sub-dish. This results in a local popularity list for each dish type, reflecting the relative popularity of multiple sub-dishes in each category within a specific scenario.

[0060] Step S5: Based on historical ordering service records, construct a dish association map for each table type in each local crowd density range. Generate target dish combinations for each local crowd density range based on the dish association map, and obtain global dish combination recommendation data for each table type.

[0061] In this step, to identify the pairing relationships of specific sub-dishes under different dish types, based on historical ordering service records and the aforementioned local time-slot structure lists for different table types, the pairing characteristics between different sub-dishes are further analyzed. A dish association map is constructed for each table type in different local traffic density ranges. Based on this map, considering dish popularity and actual pairing relationships, fine-grained target dish combinations for different scenarios can be determined. The target dish combinations from different local time slots are integrated to obtain global dish combination data for each table type. This data serves as a reference for ordering different user groups during different dining times and with varying restaurant traffic. It facilitates restaurant optimization of dishes, enabling improvements in real-time dish display and dynamic menu generation, and also provides data support for various business scenarios such as set menu design and advertising promotion.

[0062] As an example implementation process, a food association map is constructed for each table type in each local crowd density range based on historical ordering service records. Specifically, this includes:

[0063] Step S51: Based on the structure of multiple second candidate dishes in the local crowd density interval, determine multiple dish types associated with each local crowd density interval, and determine multiple dish association graph nodes based on the multiple dish types associated with the local crowd density interval.

[0064] Specifically, for any local crowd density range of a table type, the multiple dish types associated with it are all the dish types that appear in multiple second candidate dish structures, and the multiple sub-dishes involved in these dish types are used as nodes in the dish association graph.

[0065] Step S52: Based on the global time period and table type to which the local crowd density interval belongs, extract the local dish association sample data corresponding to the local crowd density interval from the historical ordering service record data.

[0066] Among them, the sample data of local dish associations reflects the actual association between different dishes when customers order food in specific local time periods, specific table types and specific restaurant traffic scenarios. It can characterize which dishes are often ordered together and which dishes have a strong ordering correlation.

[0067] Step S53: Calculate multiple association weights of multiple dish association graph nodes in the local crowd density interval based on the local dish association sample data, and construct the dish association graph of each table type in each local crowd density interval based on the multiple dish association graph nodes and multiple association weights.

[0068] The calculation of the association weight between any two dish association graph nodes involves using local dish association sample data to calculate the ratio between the number of times the two dish association graph nodes appear in the same order and the total number of orders associated with those two nodes. The higher the association weight, the more likely customers are to choose both dishes in the same order, indicating a strong association between them. Conversely, a lower association weight indicates a weaker association, meaning customers are more likely to not choose or only choose one dish in the same order. Ultimately, based on the identified multiple dish association graph nodes and their associated weights, dish association graphs corresponding to different table types and different local crowd density ranges can be constructed, facilitating a better identification of customer dish selection preferences in specific dining scenarios.

[0069] As an example implementation process, a target dish combination is generated for each local crowd density interval based on the dish association map, including:

[0070] Based on the local popularity list of multiple dish types under each local crowd density range for each table type, multiple candidate dish combinations are generated under each local crowd density range for each table type.

[0071] In generating candidate dish combinations, we can first determine multiple representative high-popularity dishes for each dish type based on the local popularity list of different dish types. Then, based on the local time period structure list of the table type, we can determine multiple second candidate dish structures under different local crowd density intervals. For example, for a small table with a meat main dish + vegetable dish structure, we can arrange and combine multiple representative high-popularity dishes of the dish type according to this combination to obtain multiple candidate dish combinations under this dish structure.

[0072] The structural stability of each candidate dish combination is tested based on the dish association graph, and the structural stability index of each candidate dish combination is calculated. Based on the structural stability index, multiple target dish combinations are determined from multiple candidate dish combinations under local population density range.

[0073] The calculation of the structural stability index can be achieved by determining the association weight between any two dishes in the candidate dish combination based on the dish association graph, and then calculating the average of multiple association weights as the structural stability index of the candidate dish combination, which is used to characterize the overall association strength between different dishes in the combination. Finally, multiple candidate dish combinations with structural stability indices greater than the preset structural stability threshold are selected and marked as the target dish combination.

[0074] It's worth noting that compared to selecting the most popular dish from each dish type as the recommended combination, the combination of dishes with higher structural stability provided by this invention better considers the pairing relationships between dishes. For example, for a combination of meat, seafood, and vegetables, although the most popular dishes in each dish type are selected, resulting in the highest overall popularity, their combination may not be common in historical ordering data. This is because, due to differences in taste, cooking methods, or serving time, their combination may not suit the actual needs of customers. By analyzing the correlation between dishes, this invention can better identify combinations that frequently occur together in actual orders, which can better meet the actual needs of customers, thereby improving the customer's dining experience and providing more accurate data support for businesses in various application scenarios such as developing set menus, conducting advertising promotions, and providing real-time dish displays or recommendations.

[0075] Please see Figure 2 The second aspect of this invention provides a catering service data intelligent analysis system, specifically used to implement the above-mentioned catering service data intelligent analysis method, including:

[0076] The food table type association analysis module is used to collect historical ordering service records of the restaurant, including table type information and food information of multiple orders. It performs food table type association analysis on the historical ordering service records data and constructs a time series curve of food structure for each table type.

[0077] The global time-segment ordering analysis module is used to perform global time-segment ordering distribution analysis for each table type based on the time-series curve of the dish structure, and generate a global time-segment structure list for each table type, including the first candidate dish structure corresponding to multiple global time periods.

[0078] The order distribution optimization module is used to construct multiple crowd density distribution vectors for each global time period based on historical order service record data. Based on the crowd density distribution vectors, the module optimizes the crowd density distribution of the global time period structure list and generates a local time period structure list for each table type, including the second candidate dish structure for multiple local crowd density intervals.

[0079] The local dish popularity analysis module is used to identify multiple dish types in a restaurant and perform local dish popularity analysis based on a local time period structure list, constructing a local popularity list of multiple dish types under each table type and each local flow density interval;

[0080] The menu combination recommendation and analysis module is used to construct a menu association map for each table type in each local crowd density range based on historical ordering service records. Based on the menu association map, the module generates target menu combinations for each local crowd density range, thus obtaining global menu combination recommendation data for each table type.

[0081] The above are merely specific embodiments of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for intelligent analysis of catering service data, characterized in that, include: Collect historical ordering service data from the restaurant, including table type and dish information for multiple orders. Perform dish-table type correlation analysis on the historical ordering service data to construct a time-series curve of the dish structure for each table type. The time-series curve of the dish structure includes the quantity of each dish type combination in multiple time periods. Based on the time-series curve of the menu structure, the restaurant's operating hours are divided into several global time periods. According to the time-series curve of the menu structure, the global time period order distribution analysis is performed for each table type. This includes statistically analyzing the proportion of each menu type combination for any table type in each global time period, determining multiple first candidate menu structures for each global time period based on the proportion of menu type combinations, and generating a global time period structure list for each table type. Based on historical ordering service records, construct multiple flow density distribution vectors for each global time period. For each first candidate dish structure, construct the flow density distribution vector corresponding to its global time period. Based on the multiple flow density distribution vectors, determine the flow density distribution interval for each global time period. Divide each global time period into multiple local time periods. Based on the flow density distribution intervals, determine multiple local flow density intervals for each global time period. Based on historical ordering service records and the distribution range of people flow density, the number of times the first candidate dish structure appears in each local people flow density range during the global time period is counted. The ratio of the number of times the first candidate dish structure appears to the total number of times all dish structures appear during the global time period is calculated as the structure frequency parameter of the first candidate dish structure in each local people flow density range. Based on the structure frequency parameters of multiple local people flow density ranges, the local structure frequency distribution vector of the first candidate dish structure in the global time period is constructed. Multiple local structure frequency distribution vectors of each first candidate dish structure are fused to construct the global structure frequency distribution vector corresponding to the first candidate dish structure; The target local pedestrian density interval to which each first candidate dish structure belongs is determined based on the global structural frequency distribution vector. Based on the target local pedestrian density intervals to which multiple first candidate dish structures belong, multiple second candidate dish structures are determined for each local pedestrian density interval. Identify multiple dish types in the restaurant and conduct local analysis of dish popularity based on local time period structure lists, constructing local popularity lists for multiple dish types under each table type and each local flow density interval; Based on historical ordering service records, a dish association map is constructed for each table type in each local crowd density interval. This includes multiple dish types associated with multiple second candidate dish structures in the local crowd density interval, as well as the association weights of multiple dish types, to construct the dish association map for each table type in each local crowd density interval. Based on the dish association graph, target dish combinations are generated for each local crowd density interval. This includes a local popularity list of multiple dish types for each table type and each local crowd density interval, generating multiple candidate dish combinations for each table type and each local crowd density interval. The structural stability of each candidate dish combination is tested based on the dish association graph, and the structural stability index of each candidate dish combination is calculated. Based on the structural stability index, multiple target dish combinations are determined from the multiple candidate dish combinations for each local crowd density interval, resulting in global dish combination recommendation data for each table type.

2. The intelligent analysis method for catering service data according to claim 1, characterized in that, Based on the association of multiple second candidate dishes with multiple dish types in local crowd density intervals, and the association weights of multiple dish types, a dish association map for each table type in each local crowd density interval is constructed, including: Based on the structure of multiple second candidate dishes in local crowd density intervals, determine multiple dish types associated with each local crowd density interval, and determine multiple dish association graph nodes based on the multiple dish types associated with local crowd density intervals; Based on the global time period and table type to which the local crowd density interval belongs, extract the local dish association sample data corresponding to the local crowd density interval from the historical food ordering service data; Based on the local dish association sample data, calculate multiple association weights of multiple dish association graph nodes in the local flow density interval. Based on the multiple dish association graph nodes and multiple association weights, construct the dish association graph for each table type in each local flow density interval.

3. The intelligent analysis method for catering service data according to claim 2, characterized in that, Based on the dish association graph, structural stability testing was performed on each candidate dish combination, and the structural stability index of each candidate dish combination was calculated, including: The association weights between any two dishes in the candidate dish combination are determined based on the dish association graph, and the average of multiple association weights is calculated as the structural stability index of the candidate dish combination.

4. A smart data analysis system for catering services, characterized in that, The system is used to implement the intelligent data analysis method for catering services as described in any one of claims 1-3, including: The food table type association analysis module is used to collect historical ordering service records of the restaurant, including table type information and food information of multiple orders. It performs food table type association analysis on the historical ordering service records data and constructs a time series curve of food structure for each table type. The global time-segment ordering analysis module is used to perform global time-segment ordering distribution analysis for each table type based on the time-series curve of the dish structure, and generate a global time-segment structure list for each table type, including the first candidate dish structure corresponding to multiple global time periods. The order distribution optimization module is used to construct multiple crowd density distribution vectors for each global time period based on historical order service record data. Based on the crowd density distribution vectors, the module optimizes the crowd density distribution of the global time period structure list and generates a local time period structure list for each table type, including the second candidate dish structure for multiple local crowd density intervals. The local dish popularity analysis module is used to identify multiple dish types in a restaurant and perform local dish popularity analysis based on a local time period structure list, constructing a local popularity list of multiple dish types under each table type and each local flow density interval; The menu combination recommendation and analysis module is used to construct a menu association map for each table type in each local crowd density range based on historical ordering service records. Based on the menu association map, the module generates target menu combinations for each local crowd density range, thus obtaining global menu combination recommendation data for each table type.

Images