KR20200080355A - Recipe Recommendation System using Big Data - Google Patents

Abstract

The present invention relates to a system for recommending a recipe using big data and, more specifically, to a system for constructing big data by collecting user data and recipe data and recommending a recipe to a user through algorithm, which recommends a recipe based on a big data analysis. The system comprises: a member authentication step (S10); a BMI confirmation step (S20); a recipe search step (S30); a recipe inquiry step (S40); and a recipe evaluation step (S50).

Description

Recipe recommendation system using big data

The present invention relates to a recipe recommendation system using big data, and more specifically, it collects user data and recipe data, builds big data, and recommends recipes to users through an algorithm that recommends recipes based on big data analysis. System.

Modern people are increasingly interested in food for a healthy lifestyle. In particular, people who diet by diet, etc., use a food exchange table first with their existing efforts. The food exchange table is divided into six food groups: food, grain, fish, vegetable, fat, milk, and fruit groups, depending on the nutrients. Made by using this, it can be exchanged and consumed freely within the same food group, which helps a person who wants to manage weight to manage balanced nutrition and total calorie intake.

However, the food exchange table is mainly defined as food ingredients, and since the cooking method using the ingredients is not described, it is difficult to consider the total calories of foods that change according to the cooking method and the compatibility of foods, and because the amount of exercise is not considered, There is a problem that diet is difficult to control.

Therefore, it is possible to recommend using big data to recommend the recipe itself to the user. Big data is data that is generated in a digital environment. It is very large in scale, short in generation cycle, and diverse in shape and number. From the above-mentioned increase in the size and type of big data, it is necessary to analyze the relationship between various opinions of users and data, and to be able to predict new data. That is, by analyzing various conditions (eg, weight, palatability, etc.) of users, a recipe that the user is satisfied should be predicted.

Korean Patent Publication No. 10-2018-0128147 (Sikshin Co., Ltd., published on Dec. 3, 2018)

The present invention has been devised to solve the above-described conventional problems, and has an object to collect user information using big data and recommend recipes suitable for users based on the information.

In addition, we want to collect various information from users and visualize and recommend the results.

In a recipe recommendation system using a database that includes user data and recipe data, a member authentication step for registering a user who wishes to use the system or modifying and querying member information of a subscriber, a key of the registered user, A BMI check step for calculating BMI through weight, gender, and searching for pre-stored BMI information, a recipe search step for searching a recipe for food to be cooked, a recipe search step for searching for a recipe searched in the recipe search step, and the user It characterized in that it comprises a recipe evaluation step of proceeding with the evaluation of the recipe used.

In the recipe search step, a name search in which a search is performed through a recipe name or a material name search, a personalized search for searching a recipe matching information of a registered user, and a pre-search recipe for a user having similar information to the registered user are recommended. Characterized in that it performs at least one search step selected from among recommended searches.

The recipe evaluation step is characterized in that at least one evaluation step selected from among evaluation inputs for evaluating the selected recipe, evaluation modification for modifying the pre-stored evaluation, and evaluation inquiry for pre-stored recipe evaluation information is performed. .

The recipe search step is characterized by searching by applying a statistical technique or knowledge discovery technique, and finding and applying a statistical rule or pattern in the database.

The personalized search and the recommended search are characterized by searching for a recipe by determining a user's gender, age, BMI, daily calorie intake, user's condition, allergy and preference.

The effect according to the present invention, there is an effect that can be used anytime, anywhere by constructing big data that combines user information and recipe information. In addition, by analyzing various conditions through a recommendation algorithm analyzed based on big data, there is an effect that can provide a user with a satisfactory recipe.

1 is a view showing the configuration of the entire system according to an embodiment of the present invention
2 is a view showing the structure of a system according to an embodiment of the present invention
3 is a diagram showing the steps of a recommendation system according to an embodiment of the present invention.
4 is a view showing a function and a user interface of a system according to an embodiment of the present invention
5 is a view showing a user interface screen according to an embodiment of the present invention

Advantages and features of the embodiments of the present invention, and methods for achieving them will be made clear by referring to embodiments described below in detail with reference to the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only the embodiments allow the disclosure of the present invention to be complete, and common knowledge in the technical field to which the present invention pertains. It is provided to completely inform the person having the scope of the invention, and the present invention is only defined by the scope of the claims. The same reference numerals refer to the same components throughout the specification.

In describing embodiments of the present invention, when it is determined that a detailed description of known functions or configurations may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted. In addition, terms to be described later are terms defined in consideration of functions in an embodiment of the present invention, which may vary according to a user's or operator's intention or practice. Therefore, the definition should be made based on the contents throughout this specification.

Hereinafter, a recipe recommendation system using big data according to an embodiment of the present invention will be described in more detail with reference to FIGS. 1 to 5.

1 is a view showing the configuration of an entire system according to an embodiment of the present invention. A recipe recommendation system using big data according to an embodiment of the present invention, as shown in FIG. 1, builds big data including user data and recipe data, and recommends a recipe to a user through a recommendation algorithm . At this time, the recipe recommendation system using big data according to an embodiment of the present invention is characterized in that it is implemented through an application.

At this time, the big data 100 may use a NoSQL database (hereinafter, NoSQL). The NoSQL is a non-relational database, without a schema (structure of data, a method of expressing data, and a structure in which relations between data are defined in a formal language) or a flexible database. Unlike the existing relational database (RDBMS), NoSQL does not define a relationship between data. That is, a data table is just one data table and cannot define a relationship with another table.

Since NoSQL does not need to define a relationship, it is less complex than an existing relational database (RDBMS) and can store large amounts of data. In addition, there is no scalability through a distributed structure and quick response in case of server failure through mutual replication of data and service interruption.

Since the schema is flexible unlike the relational database (RDBMS), there is no type restriction on the value of the field part except the key structure used as an ID. In NoSQL, the consistency of data is not always guaranteed, and various and complex data queries are impossible, so when used for a big data server, the cost for data consistency is very high compared to a relational database (RDBMS). It is efficient for the utilized system.

In addition, the recommendation system according to an embodiment of the present invention applies statistical techniques and knowledge discovery techniques to product and service recommendation problems, and is referred to as data mining.

The data mining is to systematically and automatically find statistical rules or patterns within big data, and uses various weighing techniques ranging from statistics to pattern recognition. The fields of application of the data mining are largely classified as classification, clustering, association, sequencing, and prediction. The recommendation system is largely composed of a user model (User Model), user information (User Info), item information (Item Info), and preferences (Preference). In addition, it is divided into explicit preferences such as ratings and reviews, and implicit preferences such as number of views and clicks, purchases, and follow.

In addition, in recommending a recipe to a user through a recommendation algorithm, the recommendation algorithm may use Collaborative Filtering (CF) or Contents Based Filtering (CBF). Hereinafter, the recommended algorithm will be described in more detail.

Collaborative filtering (CF) is a method of calculating a similarity between a user and a similar user in consideration of the preferences or preferences of a similar user, and then predicting and recommending an item that the user has not evaluated or used. In the above method, a problem in which the prediction accuracy of the system decreases due to a lack of initial data causes a problem of cold start (a phenomenon in which the prediction accuracy of the system decreases due to a lack of initial data), but there is a lot of data. Shows very high accuracy.

The collaborative filtering does not require content information or user information of the item, and has the advantage of being relatively easy to implement, but has a disadvantage of low accuracy in a system having a similarity calculation cost and a small amount of data.

In addition, collaborative filtering is composed of user-based collaborative filtering and item-based collaborative filtering, and user-based collaborative filtering defines that a person with similar tendencies will like items of similar tendency afterwards. Hereinafter, user-based collaborative filtering will be described in more detail.

The user-based collaborative filtering uses matrix factorization, as shown in the following image. That is, using the user with the X-axis and the items as the Y-axis, fill in the matrix of each user's item preferences (such as ratings) and predict the blank space.

It can be seen through the left and right graphs that the pattern similarity of the graph is similar to Alice and User1. In the above case, the absolute values of the scores of Alice and User1 are not important and are not considered. The formula for calculating similarity for the prediction is as follows.

= 2.4 /

= 2.4 )(

= 2.4 /

= 2.4)

Through the calculation of the above formula, the following results can be obtained.

At this time, Sim represents similarity. Through the above result, it can be confirmed that users similar to Alice are User1 and User2.

Also, it is possible to infer the evaluation scores of Alice's Item 5 based on the values of User 1 and User 2.

The above equation is an equation for inferring Alice's evaluation score. By substituting a value for the above equation, the calculation may be performed to infer the value of 4.87.

Item-based collaborative filtering is a collaborative filtering that infers that if the past user liked the item'ga', they would also like the'gy' item similar to the'ga' item. When using the item-based collaborative filtering, the calculation cost can be reduced compared to the user-based collaborative filtering, and the frequency of calculation is relatively low because the item category is not changed significantly. At this time, the similarity between items uses a cosine similarity method. Below is a formula for finding the cosine similarity.

The actual calculation is to divide the dot product of two vectors by the product of the magnitude of each vector,

The similarity between item 1 (Item 1) and item 5 (Item 5) can be obtained through the above formula, and Alice 5 (Item 5) from Alice through the similarity between Item 1 (Item 1) and Item 5 (Item 5). The evaluation score can be predicted as 5. At this time, the prediction is performed under the assumption that the similarity of item 1 is the highest among the similarities between item 5 and other items.

In performing the collaborative filtering (CF), the collaborative filtering uses user-based filtering.

As described above, in order to solve the cold start problem, which is a disadvantage of collaborative filtering, prior to guessing preferences through evaluation information of other users, a group of users with similar user information is configured, and The recipe is recommended using the user's preference information. For example, preference information is collected by dividing users by gender and age first, and through this, if there are few users initially and there is no user with similar information, preference among recipes used by users of the same gender or the same age High recipes are recommended. Even in the case of a large amount of data, the preferences can be predicted through the above-described preference calculation method (matrix decomposition) for users with similar preferences in the same age group of users of the same gender.

Content-based filtering (CBF) is a personalized recommendation system that recommends personal preferences and product content. At this time, the personalized recommendation system is a system that recommends only the preferences of a user based on a user's profile and content usage record, and is characterized by not considering evaluation information of other users. The content-based filtering has an advantage of compensating for a cold start (cold start, a phenomenon in which the prediction accuracy of the system decreases due to lack of initial data), which is a disadvantage of collaborative filtering.

The content-based filtering proceeds in three stages in the order of content information generation of items, preference creation based on correlation between users and items, and recommendation list generation. At this time, due to the recommendation result focused only on the user preference, there is a disadvantage of not recommending the item with low association.

For example, the content-based filtering is recommended based on personal user profile information. If the user likes wine, he/she recommends content with characteristics and characteristics similar to those of the wine. In this case, there is a disadvantage in that various contents cannot be accessed because the user is only recommended for contents similar to contents that the user directly used or to which a good evaluation is given.

Further, the content-based filtering includes user tendency information and user state information. The user propensity information is a system that recommends a user by including or excluding a material that a user likes or dislikes, and the user status information is a system that provides a user with a personalized recipe recommendation function as described above. Since the user propensity information and the user state information are data that is changed according to the state of the user to use, they are not stored in the big data, but are input and processed by the user while using the recipe recommendation function.

2 is a view showing the structure of a system according to an embodiment of the present invention, FIG. 3 is a view showing steps of a recommendation system according to an embodiment of the present invention, and FIG. 4 is according to an embodiment of the present invention It is a diagram showing the function and user interface of the system. A structure of a recipe recommendation system using big data according to an embodiment of the present invention will be described with reference to FIGS. 2 to 4.

2 to 4, a recipe recommendation system using big data according to an embodiment of the present invention undergoes a series of processes for recommending and providing an appropriate recipe to a user.

First, a member authentication step of a user who wants to use the system is made or a member authentication step of modifying and querying member information of a subscriber, a BMI through the user's height, weight, and gender is calculated, and the previously stored BMI information is calculated. The BMI check step to search, a recipe search step to search the recipe of the food to be cooked, a search step to search the recipe searched in the recipe search step, and a recipe evaluation step to evaluate the recipe used by the user .

Hereinafter, the series of steps will be described in more detail.

The member authentication step includes member registration, member information modification, member information inquiry, etc., membership registration as a user for using the interface, member information modification to modify the subscribed information, and inquiry of the subscribed information Includes member information inquiry.

The BMI check step is a step for calculating BMI through the height, weight, and gender of the subscribed user, and inquiring pre-stored BMI information. As a function included in the user-specific recommendation search, enter the user's key and weight information. If it does, the BMI value is calculated by referring to the user's gender. The BMI is a well-known calculation method, and a description thereof will be omitted.

In the recipe search step, a name search in which a search is performed through a recipe name or a material name search, a personalized search for searching a recipe matching information of a registered user, and a pre-search recipe for a user having similar information to the registered user are recommended. At least one search step selected among recommended searches is performed.

At this time, the recipe search step is characterized in that, as described above, a statistical technique or a knowledge discovery technique is applied to search, and a statistical rule or pattern is found and applied in the database.

In the recipe search step, in the personalized search and recommendation search, a recipe is searched by determining gender, age, BMI, calories per day, user status, allergy and preference. The above determination is made through the above-mentioned recommendation algorithm.

When a recipe recommendation is made using the recommendation algorithm, a recommendation criterion according to a propensity is recommended based on a user's gender and age. That is, among the recipes used by users of the same gender and similar age, the best rating is recommended. For example, in the case of a man in his 20s, when the evaluation of the green juice for beginners is good, the current user is recommended to the green juice when he is in the 20s.

The recipe recommendation criteria for each user considers the user's BMI, daily calorie intake, user status, and ingredients to be excluded. For example, if the current user's BMI is in the normal range, the daily calorie intake exceeds the daily reference value, and the user is currently tired and has the information that he dislikes the banana, find the recipe in the case where the BMI is normal among all the recipes, Among the recipes with low calorie and high vitamin C content, bananas are not recommended.

In addition, when a recommendation is made through a recipe category, a recipe suitable for a user's situation is classified by category, and a search through a basic recipe and a material name is also possible.

The recipe evaluation step performs at least one evaluation step selected from an evaluation input for evaluating a selected recipe, an evaluation modification to modify a pre-stored evaluation, and an evaluation inquiry to query pre-stored recipe evaluation information.

5 is a view showing a user interface screen according to an embodiment of the present invention. As shown in FIG. 5, the interface may provide various screens according to a user's selection. On the first screen when the interface is executed, a button for logging in or registering as a member is generated at the bottom button (Application Login). 5(a) shows a main page in the interface, and may include a search through a recipe name, a search through a category classification, a search through a recommendation, a user-customized search, and a user wish list. In other words, it is a slide showing the image of a random recipe to the user, a page (①) created to post advertisements or announcements, a button (②) to go to a page for recipe inquiry, and various information applied to the application This is a page that provides them, a button (③) that contains information on nutrient components, a button (④) that moves to a page for viewing and modifying user information, a page for the first user of the application, and a simple tip for the application. It may be composed of a button (⑤) to move to the provided page. 5(b) is an interface related to a search screen, and may be composed of a search through a recipe name, a search through a category classification, a search through a recommendation, a user customized search, and a user wish list. At this time, you will search by the name of the recipe, and the button (①) that lists all the recipes containing the string, the search button (②) through category classification, the search (③) through recommendation, and the user-defined search. User's status (nutrient and improvement function) selection, BMI measurement, overseas material selection, current calorie intake (high, medium, low) will proceed in this order, and the rest of the items except the user's first choice of status Is composed of a button (④) that allows selective input, and a recipe that a user likes among the recipes that can be stored in the wish list, and a button (⑤) that searches stored recipes.

In the above description, various embodiments of the present invention have been presented and described, but the present invention is not necessarily limited thereto, and a person having ordinary knowledge in the technical field to which the present invention pertains may depart from the technical spirit of the present invention. It will be readily appreciated that branch substitution, modification and modification are possible.

100: big data
200: user data
300: recipe data
400: recommended algorithm
S10~S50: Stage of recommendation system

Claims (5)

In the recipe recommendation system using a database containing user data and recipe data,
A member authentication step (S10) in which a user who intends to use the system is registered or who modifies and inquires the member information of the existing subscriber;
A BMI check step (S20) of calculating the BMI through the height, weight, and gender of the subscribed user and querying pre-stored BMI information;
A recipe search step (S30) of searching for a recipe of food to be cooked;
A recipe inquiry step (S40) for inquiring the recipe found in the recipe search step; And
A recipe evaluation step of evaluating the recipe used by the user (S50);
Recipe recommendation system using big data, characterized in that it comprises a.
The method of claim 1, wherein the recipe search step (S30)
Name search, where the search is done by searching for recipe names or material names;
A personalized search for searching recipes matching information of the subscribed user; And
A recommendation search recommending a pre-search recipe of a user with similar information to the subscribed user; A recipe recommendation system using big data, characterized in that at least one search step selected from among them is performed.

The method of claim 1, wherein the recipe evaluation step (S50)
An evaluation input that proceeds with evaluation for the selected recipe;
An assessment correction that modifies a previously stored assessment; And
An evaluation query for querying previously stored recipe evaluation information;
A recipe recommendation system using big data, characterized in that at least one evaluation step selected from among them is performed.
According to claim 1, The recipe search step (S30)
Search by applying statistical or knowledge discovery techniques,
A recipe recommendation system using big data characterized by finding and applying statistical rules or patterns in the database.
The method of claim 2, wherein the personalized search and the recommended search are
A recipe recommendation system using big data characterized by searching for a recipe by determining a user's gender, age, BMI, calories burned daily, user's condition, allergy and preference.

Images