Automatic matching catering menu pushing system

Abstract

The invention belongs to the technical field of catering informatization service, and provides an automatic matching catering menu pushing system, which comprises a user data acquisition module, a catering menu pushing module and a catering menu pushing module, the menu data management module stores and manages menu classification, labels and states, and obtains content feature data; the user portrait construction module generates a multi-dimensional portrait according to the multi-aspect information of the user; the menu matching module combines collaborative filtering and a content recommendation algorithm, and generates a recommendation menu based on the user portrait; the intelligent pushing module pushes dishes of nearby merchants according to the portraits, and the pushing effect is evaluated through the click rate and the conversion rate; according to the method, personalized and accurate catering menu recommendation is realized, and real-time updating and optimization are carried out in combination with geographic positions and time period factors, so that personalized pushing which better fits user preferences and dining environments is provided.

Description

Technical Field

[0001] The present invention belongs to the technical field of catering information services, and in particular to an automatic catering menu matching and pushing system. Background Art

[0002] With the rapid development of digital catering services, the catering menu push system has become an important tool to improve user experience and promote sales;

[0003] However, many existing systems make recommendations based solely on basic user information or past order history, failing to fully capture a user's diverse preferences, such as flavor preferences, dining habits, and geographic location. This results in insufficiently relevant recommendations and a failure to truly meet users' personalized needs. Furthermore, many existing systems solely utilize collaborative filtering or content-based recommendation algorithms. Collaborative filtering algorithms rely on extensive historical user behavior data, making it difficult to accurately identify similar user groups during a user's cold start phase (e.g., a newly registered user with no historical orders) or in data-sparse scenarios, leading to inaccurate recommendations.

[0004] Existing systems generally ignore the impact of a user's current location and specific time of day on their dining choices, failing to dynamically adjust recommendations, making it difficult to deliver personalized recommendations tailored to their needs. Furthermore, some systems only recommend dishes based on recommendation scores, failing to factor in the user's real-time location. This can result in recommended restaurants being too far away, reducing their willingness to consume. Furthermore, there's a lack of quantitative evaluation and feedback optimization mechanisms for recommendation effectiveness, making it impossible to dynamically adjust recommendation strategies based on metrics like click-through rate and conversion rate.

[0005] To this end, technicians in this field have proposed an automatic matching restaurant menu push system, which aims to achieve personalized and accurate restaurant menu recommendations to enhance users' dining experience and satisfaction. It combines geographic location and time factors to perform real-time updates and optimizations, thereby providing personalized push that is more in line with user preferences and dining environment. Summary of the Invention

[0006] In order to solve the above technical problems, the present invention provides an automatic matching catering menu push system to solve the problems raised in the background technology.

[0007] An automatic matching restaurant menu push system, comprising:

[0008] User data collection module, used to collect user data in multiple dimensions and dynamically update it to obtain multi-dimensional user data;

[0009] The menu data management module is used to store and manage the classification, label and real-time status of the menu and obtain the menu content feature data;

[0010] User portrait construction module, used to obtain multi-dimensional user portraits based on users' basic information, consumption behavior, taste preferences and geographical location;

[0011] A menu matching module is used to combine the collaborative filtering algorithm and the content-based recommendation algorithm according to the multi-dimensional user portrait and menu content feature data to generate a final recommended menu;

[0012] The intelligent push module is used to push the recommended menu of dishes of nearby merchants to the user based on the multi-dimensional user portrait, and evaluate the push effect based on the click rate and conversion rate to obtain the evaluation result.

[0013] Preferably, the user data collection module collects user data in multiple dimensions through user registration information, historical order records, user reviews, geographic location information, and browsing behavior data;

[0014] Integrate the geographic location acquisition function in the user-side APP to obtain the user's current location in real time;

[0015] Record the user's menu browsing time and click count in the APP, and dynamically update the user data based on the user's new orders and reviews to obtain multi-dimensional user data.

[0016] Preferably, the menu data management module is used for merchants to enter menu information including dish names, prices, pictures, descriptions, ingredient information, taste characteristics, and nutritional ingredients through the merchant management backend, and provides a visual menu entry interface;

[0017] Categorize and manage menus by cuisine and dish type, add structured tags to each dish, obtain menu content feature data, and support merchants to modify prices, replace images, and update descriptions on menus;

[0018] Monitor the status of dishes on and off the shelves in real time and synchronize them to the menu matching module.

[0019] Preferably, the user portrait building module is used to obtain a basic information portrait based on user registration information, wherein the basic information includes gender, age, occupation, and income level;

[0020] By analyzing the user's historical order records, a consumption behavior profile is obtained, which includes consumption frequency, consumption amount, frequently ordered dishes, and consumption time period;

[0021] Obtain a taste preference profile based on the user's ordered dishes, user reviews, and browsing behavior data. The taste preference profile includes the user's favorite cuisines and taste characteristics.

[0022] Combining the user's location information and order data, we can generate a location profile that includes the user's ordering preferences and consumption frequency at different locations.

[0023] By combining the basic information portrait, consumption behavior portrait, taste preference portrait and geographic location portrait, a multi-dimensional user portrait is obtained.

[0024] Preferably, the menu matching module includes:

[0025] According to the menu content feature data, the dish i is represented by feature vector, and the feature vector of dish i is obtained as f i ;

[0026] Use the following formula to generate the taste preference vector p based on the multi-dimensional user portrait u :

[0027]

[0028] Among them, H u is the historical order collection of user u, f j is the feature vector of dish j;

[0029] Based on the taste preference vector p u and the characteristic vector of dish i is f i , use the following formula to calculate the similarity between user preferences and menu content features, and generate a content-based recommendation score result S for each dish CB :

[0030]

[0031] Based on the multi-dimensional user profile, recommendations are generated based on the taste preferences of similar users, and user similarity is calculated using the following formula:

[0032]

[0033] Among them, I uv is the set of dishes that user u and user v have jointly evaluated. and Represents the average rating of user u and user v, sim(u,v) is the similarity between user u and user v, r u,i and r v,i Represent the ratings of user u and user v on dish i respectively;

[0034] Based on the weighted scores of the Top-K similar users, the collaborative filtering score is generated using the following formula:

[0035]

[0036] in, is the set of k users most similar to user u, S CF The recommendation score results for each dish based on collaborative filtering.

[0037] Preferably, the menu matching module further includes:

[0038] According to the geographical location and consumption behavior characteristics in the multi-dimensional user portrait, the recommendation score result S of each dish obtained based on collaborative filtering is CF And the content-based recommendation score result S for each dish CB Perform weighted fusion, expressed as:

[0039] S(i)=α·S CB (i)+β·S CF (i)+γ·F(Geo,CoBe)

[0040] Among them, α, β, γ are weight coefficients, F(Geo, CoBe) is the adjustment function of geographical location and consumption behavior characteristics, S CB (i), S CF (i) represents the recommendation score of the i-th dish obtained based on the collaborative filtering algorithm and the content-based recommendation algorithm, respectively, and S(i) is the final recommendation score of the i-th dish;

[0041] A recommendation menu is generated according to the final recommendation score result S(i).

[0042] Preferably, the intelligent push module includes:

[0043] Obtain the user's real-time location coordinates (x u ,y u ), for the merchant set B=(b1,b2,...,b n ), calculate each merchant b j The coordinates (x j ,y j ) and the distance d(b j ,u);

[0044] By presetting the distance threshold D max , filter out d j ≤D max Merchant collection B near ;

[0045] According to the user consumption behavior profile, the dishes in the recommended menu are filtered and the dishes are selected by filtering the dishes that are related to the user's frequently ordered categories. u Matching dish subset Recommend menu items and exclude dishes whose prices are significantly out of the user's spending power;

[0046] For dish i that meets the conditions, use the following formula to generate the final push priority P(i):

[0047] P(i)=δ·S(i)+λ·w d +μ·w t

[0048] Among them, δ, λ, μ are weight coefficients, w d is the merchant distance weight, w t Increase the weight of dishes matched to real-time time periods;

[0049] Arrange the dishes in descending order according to P(i) and generate a push list R={i1,i2,...,i m}, push according to the push preferences set by the user.

[0050] Preferably, the intelligent push module further includes:

[0051] Based on the click-through rate and conversion rate, use the following formula to calculate the comprehensive performance score:

[0052] A=θ·CTR+(1-θ)·CVR

[0053] Among them, θ is the weight coefficient, CTR is the click-through rate, and CVR is the conversion rate. The push effect is evaluated to obtain the evaluation result A.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] 1. The present invention deeply depicts user needs through multi-dimensional portraits, combines collaborative filtering and content recommendation algorithms to improve the matching degree between dishes and user preferences, and incorporates geographic location and time factors into the push logic to meet needs.

[0056] 2. The present invention ensures that the recommendation results are adjusted as user habits change by updating user data and menu status in real time. The merchant side manages the menu through structured tags to reduce operating costs, and the user side obtains personalized recommendations to improve the consumption experience. It also assists merchants in optimizing the dish structure and increasing the order volume through the effect evaluation mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a block diagram of the automatic matching restaurant menu push system of the present invention. DETAILED DESCRIPTION

[0058] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0059] As attached Figure 1 As shown:

[0060] Embodiment 1: The present invention provides a system for automatically matching restaurant menus and pushing them, comprising:

[0061] The user data collection module is used to collect user data in multiple dimensions and dynamically update it to obtain multi-dimensional user data. The user data collection module collects user data in multiple dimensions through user registration information, historical order records, user reviews, geographic location information, and browsing behavior data.

[0062] Integrate the geographic location acquisition function in the user-side APP to obtain the user's current location in real time;

[0063] Record the user's menu browsing time and click count in the APP, and dynamically update the user data based on the user's new orders and reviews to obtain multi-dimensional user data.

[0064] By collecting user information data from multiple dimensions, we can build a dynamic and comprehensive user data pool, and update the data in real time to ensure the timeliness of user portraits and adapt to changes in consumption habits.

[0065] The menu data management module is used to store and manage the classification, labeling and real-time status of menus, and obtain menu content feature data; the menu data management module is used for merchants to enter menu information including dish names, prices, pictures, descriptions, ingredient information, taste characteristics, and nutritional ingredients through the merchant management background, and provides a visual menu entry interface;

[0066] Categorize and manage menus by cuisine and dish type, add structured tags to each dish, obtain menu content feature data, and support merchants to modify prices, replace images, and update descriptions on menus;

[0067] Monitor the status of dishes on and off the shelves in real time and synchronize them to the menu matching module.

[0068] By classifying menus (cuisine, type) and labeling them, and monitoring their on-shelf and off-shelf status in real time, we can build a structured menu content feature library to facilitate subsequent recommendation algorithm matching. At the same time, we can update the menu status in real time to effectively ensure the accuracy and availability of the recommendation results.

[0069] The user portrait construction module is used to obtain a multi-dimensional user portrait based on the user's basic information, consumption behavior, taste preferences and geographical location. The user portrait construction module is used to obtain a basic information portrait based on the user's registration information. The basic information includes gender, age, occupation and income level.

[0070] By analyzing the user's historical order records, we can obtain a consumption behavior profile, which includes consumption frequency, consumption amount, frequently ordered dishes, and consumption time period.

[0071] Based on the user's ordered dishes, user reviews, and browsing behavior data, a taste preference profile is obtained. The taste preference profile includes the user's favorite cuisine and taste characteristics;

[0072] Combining the user's location information and order data, we can generate a location profile that includes the user's ordering preferences and consumption frequency at different locations.

[0073] By combining basic information portraits, consumption behavior portraits, taste preference portraits and geographic location portraits, a multi-dimensional user portrait is obtained.

[0074] A multi-dimensional portrait is constructed from four dimensions: basic information, consumption behavior, taste preferences, and geographic location. Deep preferences are explored through historical orders, reviews, and browsing behaviors to form a three-dimensional user tag system. The combination of geographic location and consumption behavior can identify needs in different scenarios.

[0075] The menu matching module is used to combine the collaborative filtering algorithm and the content-based recommendation algorithm based on the multi-dimensional user portrait and menu content feature data to generate the final recommended menu;

[0076] According to the menu content feature data, the dish i is represented by feature vector, and the feature vector of dish i is obtained as f i ;

[0077] Use the following formula to generate the taste preference vector p based on the multidimensional user portrait u :

[0078]

[0079] Among them, H u is the historical order collection of user u, f j is the feature vector of dish j;

[0080] Based on the taste preference vector p u and the characteristic vector of dish i is f i , use the following formula to calculate the similarity between user preferences and menu content features, and generate a content-based recommendation score result S for each dish CB :

[0081]

[0082] Based on multi-dimensional user portraits, recommendations are generated based on the taste preferences of similar users. The following formula is used to calculate user similarity:

[0083]

[0084] Among them, I uv is the set of dishes that user u and user v have jointly evaluated. and Represents the average rating of user u and user v, sim(u,v) is the similarity between user u and user v, r u,i and r v,i Represent the ratings of user u and user v on dish i respectively;

[0085] Based on the weighted scores of the Top-K similar users, the collaborative filtering score is generated using the following formula:

[0086]

[0087] in, is the set of k users most similar to user u, S CF The recommendation score results for each dish based on collaborative filtering.

[0088] According to the geographical location and consumption behavior characteristics in the multi-dimensional user portrait, the recommendation score result S of each dish obtained based on collaborative filtering is CF And the content-based recommendation score result S for each dish CB Perform weighted fusion, expressed as:

[0089] S(i)=α·S CB (i)+β·S CF (i)+γ·F(Geo,CoBe)

[0090] Among them, α, β, γ are weight coefficients, F(Geo, CoBe) is the adjustment function of geographical location and consumption behavior characteristics, S CB (i), S CF (i) represents the recommendation score of the i-th dish obtained based on the collaborative filtering algorithm and the content-based recommendation algorithm, respectively, and S(i) is the final recommendation score of the i-th dish;

[0091] Based on the final recommendation score S(i), a recommendation menu is generated. By integrating collaborative filtering and content recommendation algorithms, this recommendation menu effectively balances "user individual preferences" (content recommendation) with "group common preferences" (collaborative filtering), improving recommendation diversity and accuracy. The adjustment function of geographic location and consumer behavior makes recommendations more context-sensitive.

[0092] The intelligent push module is used to push recommended menus of nearby merchants to users based on multi-dimensional user portraits, and evaluate the push effect based on click-through rate and conversion rate to obtain evaluation results.

[0093] Obtain the user's real-time location coordinates (x u ,y u ), for the merchant set B=(b1,b2,...,b n), calculate each merchant b j The coordinates (x j ,y j ) and the distance d(b j ,u);

[0094] By presetting the distance threshold D max , filter out d j ≤D max Merchant collection B near ;

[0095] According to the user's consumption behavior profile, the dishes in the recommended menu are filtered and the dishes are selected by filtering the dishes that are related to the user's frequently ordered categories. u Matching dish subset Recommend menu items and exclude dishes whose prices are significantly out of the user's spending power;

[0096] For dish i that meets the conditions, use the following formula to generate the final push priority P(i):

[0097] P(i)=δ·S(i)+λ·w d +μ·w t

[0098] Among them, δ, λ, μ are weight coefficients, w d is the merchant distance weight, w t Increase the weight of dishes matched to real-time time periods;

[0099] Arrange the dishes in descending order according to P(i) and generate a push list R={i1,i2,...,i m}, push according to the push preferences set by the user.

[0100] Based on the click-through rate and conversion rate, use the following formula to calculate the comprehensive performance score:

[0101] A=θ·CTR+(1-θ)·CVR

[0102] Where θ is the weight coefficient, CTR is the click-through rate, and CVR is the conversion rate. The push effect is evaluated to obtain evaluation result A. Nearby merchants are filtered based on the user's real-time location, and dishes are filtered based on consumer behavior. Push priority is calculated based on factors such as distance and time of day. A list is generated in descending order, and the effect is evaluated based on click-through rate and conversion rate. This allows for precise location of nearby merchants, shortening the decision-making process, and dynamically adjusting the push order to improve user click-through rate. The results of the comprehensive evaluation indicators are quantified to facilitate push strategy optimization.

[0103] From the above, we can see that by adjusting the weights through geographic location and consumption behavior characteristics, dynamic algorithm switching is achieved, and scenario factors such as real-time location, consumption time, and activity type are included in the push priority calculation. A comprehensive scoring formula is constructed in combination with click-through rate and conversion rate, and real-time feedback on push effects is provided to form a complete closed loop; merchants can directly improve the recommendation matching efficiency through structured tag management menus; users can obtain more accurate services through real-time location and scenario-based recommendations.

[0104] It will be understood that in the development of any actual embodiment, as in any engineering or design project, numerous implementation-specific decisions may be made. Such a development effort may be complex and time-consuming, but for those of ordinary skill having the benefit of this disclosure, the development effort will be a routine task of design, fabrication, and production without undue experimentation.

[0105] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An automatic matching restaurant menu push system, characterized by: include: User data collection module, used to collect user data in multiple dimensions and dynamically update it to obtain multi-dimensional user data; The menu data management module is used to store and manage the classification, label and real-time status of the menu and obtain the menu content feature data; User portrait construction module, used to obtain multi-dimensional user portraits based on users' basic information, consumption behavior, taste preferences and geographical location; A menu matching module is used to combine the collaborative filtering algorithm and the content-based recommendation algorithm according to the multi-dimensional user portrait and menu content feature data to generate a final recommended menu; The intelligent push module is used to push the recommended menu of dishes of nearby merchants to the user based on the multi-dimensional user portrait, and evaluate the push effect based on the click rate and conversion rate to obtain the evaluation result.

2. The automatic matching restaurant menu push system according to claim 1, characterized in that: The user data collection module collects user data in multiple dimensions through user registration information, historical order records, user reviews, geographic location information, and browsing behavior data; Integrate the geographic location acquisition function in the user-side APP to obtain the user's current location in real time; Record the user's menu browsing time and click count in the APP, and dynamically update the user data based on the user's new orders and reviews to obtain multi-dimensional user data.

3. The automatic matching restaurant menu push system according to claim 1, characterized in that: The menu data management module is used for merchants to enter menu information including dish names, prices, pictures, descriptions, ingredient information, taste characteristics, and nutritional ingredients through the merchant management background, and provides a visual menu entry interface; Categorize and manage menus by cuisine and dish type, add structured tags to each dish, obtain menu content feature data, and support merchants to modify prices, replace images, and update descriptions on menus; Monitor the status of dishes on and off the shelves in real time and synchronize them to the menu matching module.

4. The automatic matching restaurant menu push system according to claim 1, characterized in that: The user portrait construction module is used to obtain a basic information portrait based on the user registration information, and the basic information includes gender, age, occupation, and income level; By analyzing the user's historical order records, a consumption behavior profile is obtained, which includes consumption frequency, consumption amount, frequently ordered dishes, and consumption time period; Obtain a taste preference profile based on the user's ordered dishes, user reviews, and browsing behavior data. The taste preference profile includes the user's favorite cuisines and taste characteristics. Combining the user's location information and order data, we can generate a location profile that includes the user's ordering preferences and consumption frequency at different locations. By combining the basic information portrait, consumption behavior portrait, taste preference portrait and geographic location portrait, a multi-dimensional user portrait is obtained.

5. The automatic matching restaurant menu push system according to claim 1, characterized in that: The menu matching module includes: According to the menu content feature data, the dish i is represented by feature vector, and the feature vector of dish i is obtained as f i ; Use the following formula to generate the taste preference vector p based on the multi-dimensional user portrait u : Among them, H u is the historical order collection of user u, f j is the feature vector of dish j; Based on the taste preference vector p u and the characteristic vector of dish i is f i , use the following formula to calculate the similarity between user preferences and menu content features, and generate a content-based recommendation score result S for each dish CB : Based on the multi-dimensional user profile, recommendations are generated based on the taste preferences of similar users, and user similarity is calculated using the following formula: Among them, I uv is the set of dishes that user u and user v have jointly evaluated. and Represents the average rating of user u and user v, sim(u,v) is the similarity between user u and user v, r u,i and r v,i Represent the ratings of user u and user v on dish i respectively; Based on the weighted scores of the Top-K similar users, the collaborative filtering score is generated using the following formula: in, is the set of k users most similar to user u, S CF The recommendation score results for each dish based on collaborative filtering.

6. The automatic matching restaurant menu push system according to claim 5, characterized in that: The menu matching module also includes: According to the geographical location and consumption behavior characteristics in the multi-dimensional user portrait, the recommendation score result S of each dish obtained based on collaborative filtering is CF And the content-based recommendation score result S for each dish CB Perform weighted fusion, expressed as: S(i)=α·S CB (i)+β·S CF (i)+γ·F(Geo,CoBe) Among them, α, β, γ are weight coefficients, F(Geo, CoBe) is the adjustment function of geographical location and consumption behavior characteristics, S CB (i), S CF (i) represents the recommendation score of the i-th dish obtained based on the collaborative filtering algorithm and the content-based recommendation algorithm, respectively, and S(i) is the final recommendation score of the i-th dish; A recommendation menu is generated according to the final recommendation score result S(i).

7. The automatic matching restaurant menu push system according to claim 1, characterized in that: The intelligent push module includes: Obtain the user's real-time location coordinates (x u ,y u ), for the merchant set B=(b1,b2,...,b n ), calculate each merchant b j The coordinates (x j ,y j ) and the distance d(b j ,u); By presetting the distance threshold D max , filter out d j ≤D max Merchant collection B near ; According to the user consumption behavior profile, the dishes in the recommended menu are filtered and the dishes are selected by filtering the dishes that are related to the user's frequently ordered categories. u Matching dish subset And exclude dishes whose prices significantly deviate from the user's spending power; For dish i that meets the conditions, use the following formula to generate the final push priority P(i): P(i)=δ·S(i)+λ·w d +μ·w t Among them, δ, λ, μ are weight coefficients, w d is the merchant distance weight, w t Increase the weight of dishes matched to real-time time periods; Arrange the dishes in descending order according to P(i) and generate a push list R={i1,i2,...,i m }, push according to the push preferences set by the user.

8. The automatic matching restaurant menu push system according to claim 7, characterized in that: The intelligent push module also includes: Based on the click-through rate and conversion rate, use the following formula to calculate the comprehensive performance score: A=θ·CTR+(1-θ)·CVR Among them, θ is the weight coefficient, CTR is the click-through rate, and CVR is the conversion rate. The push effect is evaluated to obtain the evaluation result A.

Images