TWM591214U - System for content recommendation - Google Patents

Abstract

A system for content recommendation is provided. The system includes a server that includes a plurality of databases that are the proprietary databases for a plurality of clients based on data analysis of member data of each of the clients. A computation module of the system performs the method for content recommendation. In the method, the server receives the member data provided by the clients via a user interface. The proprietary database for each client is established. A plurality of recommendation algorithms are performed on the member data so as to obtain several indicators. A ranking of the indicators is referred to for the clients to select one of the recommendation algorithms. The selected recommendation algorithm is used to establish a recommendation model by a machine learning approach. A recommendation list that records a product list to be recommended is provided for each of the clients.

Description

Content recommendation system

This creation is about a product recommendation technology, especially a content recommendation system that provides customers to select a recommendation algorithm based on indicators and then calculates a recommendation list through calculation modules in the system.

According to one of the methods of recommending commodities to consumers according to the conventional knowledge, the historical statistics of the entire retail store can be obtained from the sales end to obtain popular commodities, the purpose is to adjust inventory accordingly and increase sales.

More conventional technology can use computer programs and database technology to obtain the members' favorite products based on the member’s personal consumption statistics. In addition to recommending products by specific means, they can also obtain the relevant attributes based on the attributes of the favorite products. For other products, provide more recommended products.

In addition to recommending products that consumers like, if you want to increase the sales of other products, you can also predict the recommended products that each member may accept based on the consumption habits of the entire ethnic group. However, this speculation needs to be tried and burdensome risk.

The conventional way of recommending products is mainly considered as a single indicator, that is, by recommending consumers to purchase goods to increase sales, it does not provide a variety of considerations on the sales side.

According to a content recommendation system disclosed in the disclosure, a content recommendation method is run, the method runs on a server, and a user interface provided by the server first receives customer-provided member data, which may be the customer's data Member consumption data, at least including member information, consumption content, fees and consumption date, after analysis, you can create a customer-specific database in the server.

According to the implementation of the server, there are multiple databases, which are dedicated databases created based on the analysis of member data provided by multiple customers, and then run the above content recommendation with a computing module implemented by hardware and software. method.

The content recommendation method implemented by the system can use machine learning technology to execute multiple recommendation algorithms for member data to obtain multiple indicators, such as preference accuracy, profit margin and new product ratio, to provide customers with the ability to meet customer needs. After referring to these indicators, one of the recommendation algorithms is decided, so that a recommendation model can be established with the selected recommendation algorithm to provide the customer with a recommendation list.

Preferably, the plurality of recommendation algorithms is a data analysis method using machine learning technology, through a large amount of data analysis to learn the relationship between the products sold by customers and the consumption of members, and establish a recommendation model.

Preferably, the recommendation list is a list of commodities recommended by each member for consumption according to a recommendation model.

Further, the plurality of recommendation algorithms use a machine learning technique to learn the correlation between member consumption data and customer-supplied products based on the customer-provided member data, so as to derive a recommendation algorithm to generate a product list that recommends each member After comparing the preferences of actual members, the accuracy of the preferences of each recommendation algorithm is obtained.

Further, the multiple recommendation algorithms use a machine learning technique to learn the correlation between the member consumption data and the customer's profit based on the member data provided by the customer, so as to obtain a product list that recommends each member by each recommendation algorithm, Then calculate the profit margin based on the recommended product list of each member.

Further, the plurality of recommendation algorithms use a machine learning technique to learn the correlation between member consumption data and customer-supplied products based on the customer-provided member data, so as to derive a recommendation algorithm to generate a product list that recommends each member After comparing the list of new products provided by the customer, the ratio of new products is obtained.

Then, in a plan, each recommendation algorithm can calculate the indicators based on the member data provided by the customer, such as preference accuracy, profit margin and new product ratio, and put forward a ranking for each indicator to provide customers to decide which One recommendation algorithm, and then you can run this recommendation algorithm to provide customers with a recommendation list for each member.

Further, the server provides a user interface module for implementing a user interface dedicated to communicating with each customer, which can be an interface provided by an application programming interface (API), a web browser, or a specific software program.

In order to further understand the characteristics and technical content of this creation, please refer to the following detailed description and drawings of this creation. However, the drawings provided are for reference and explanation only, and are not intended to limit this creation.

The following are specific specific examples to illustrate the implementation of this creation, and those skilled in the art can understand the advantages and effects of this creation from the content disclosed in this specification. This creation can be implemented or applied through other different specific embodiments. The details in this specification can also be based on different views and applications, and various modifications and changes can be made without departing from the concept of this creation. In addition, the drawings in this creation are only a schematic illustration, not based on actual size, and are declared in advance. The following embodiments will further describe the relevant technical content of the creation, but the disclosed content is not intended to limit the protection scope of the creation.

It should be understood that although terms such as “first”, “second”, and “third” may be used herein to describe various elements or signals, these elements or signals should not be limited by these terms. These terms are mainly used to distinguish one component from another component, or one signal from another signal. In addition, the term "or" as used herein may include any combination of any one or more of the associated listed items, depending on the actual situation.

The disclosure provides a system for providing content recommendation services. The content recommendation method is implemented in a server that runs machine learning and other data exploration and analysis. The server can provide multiple customers to recommend products for their members through the network. The list of customers, such as companies, retailers, etc. that sell goods, can include those with various physical storefronts, and those who sell goods on online e-commerce platforms. Customers preferably have a membership system. To collect members' consumption records.

The proposed server can implement a cloud server to create different dedicated databases for different customers and provide dedicated recommendation algorithms and recommended content. The server can serve customers through a user interface. For customers, they can activate the client's computer system to connect to the server, send their membership data to the server, and perform data analysis by the server, as well as the use of machine learning technology. , Using multiple recommendation algorithms to obtain various indicators of various commercial sales, so that customers can decide one of the recommendation algorithms according to their needs, based on which a recommendation model is formed, which can effectively form a recommendation list based on the data provided by the client, It contains a list of products that customers make individual consumption recommendations for their members.

It is worth mentioning that the content recommendation system can still match a recommendation algorithm according to the product attributes, industry categories, member attributes (consumer groups), etc. sold by customers to execute the content recommendation method and provide a recommendation list. Feedback information provided by customers, as well as updated membership data, can be used to optimize the recommendation model.

FIG. 1 is a diagram showing an embodiment of the architecture of a system for implementing a content recommendation method.

The figure shows a content recommendation system implemented by the server 12, which can serve multiple customers (101, 102, 103) via the network 10. The server 12 uses a hardware computer host with various software modules to implement the functional components. For example, the calculation module 121, such as an arithmetic processor, is used to execute various algorithms, including the recommendation algorithm in the content recommendation method; there is a recommendation algorithm module 122, which includes multiple recommendation algorithms that introduce various parameters, Match the recommendation algorithm according to the needs of each customer; set up a member database 123, this member database 123 records a large number of member data provided by multiple customers, and can create a separate database for each customer, as shown in the figure for different The customer one database 111', the customer two database 112' and the customer three database 113' created by the customer.

The server 12 is provided with a user interface module 124 to implement a user interface for communicating with customers (101, 102, 103), such as an application programming interface (API) or through a web browser or a specific The user interface provided by the software program is used to manage the communication and data transmission between the server 12 and each client computer system, including data transmission, query, data reception, and message exchange.

For example, the user interface module 124 implements a web interface to provide a user interface unique to each customer (101, 102, 103). After logging in with authentication, each customer is required to provide some necessary information, such as the customer’s storefront data, Commodity information (item, description, price, historical sales data, inventory records, production records), and member data in a specific time period, such as member's consumption records, so that the server 12 can match the recommendation algorithm according to needs and characteristics, And provide indicators derived from various calculations.

The server 12 is connected to each client computer system through the network 10, including customer one 101, customer two 102 and customer three 103 in the figure. Each customer preferably has its own membership system and has established membership data Create respective databases, such as database one 111 for customer one 101, database two 112 for customer two 102, and database three 113 for customer three 103. In addition to the member's identification information and basic information, the member data is the customer's member consumption data, including at least the member's information, consumption content, fees and consumption date.

After the server 12 accepts the member data of customer one 101, customer two 102, and customer three 103 through the user interface, it creates separate databases, such as customer one database 111' and customer two database in the figure. 112' and customer three database 113'.

The above description of the membership data provided by the clients at each end forms the big data of individual clients in the server 12, so that the server 12 can provide accurate and customized recommendation algorithms for specific industries and specific client needs. The data processing method using machine learning technology is to analyze the data through algorithms, learn the characteristics of the data with a large amount of data and algorithms, and learn the correlation with specific indicators. For example, to obtain a higher profit margin, in addition to providing a list of recommended products that satisfy members (that is, consumers), they also need to be promoted at an appropriate time and price so that they can provide appropriate recommended products according to their preferences. , Increase return rate, sales rate, and then increase profit margin.

FIG. 2 then shows another embodiment of the system for implementing the content recommendation method in another schematic diagram.

In this illustration, a server 20 implementing a content recommendation system is proposed. The server 20 uses a computing module 201 to perform a content recommendation method, which includes communicating with a client through a user interface 203, including receiving client member data 22, member data It can be continuously updated, so that the calculation module 201 can optimize the recommended model after analyzing through big data.

The server 20 creates multiple databases based on the data provided by the clients at each end, which are dedicated databases created based on the analysis of member data provided by multiple clients, such as database one 221, database two 222 and data Library three 223, the number is not limited to that shown in the figure. The server 20 is provided with a plurality of recommendation algorithms, such as recommendation algorithm I 231, recommendation algorithm II 232 and recommendation algorithm III 233, using machine learning technology with several different recommendation algorithms (recommended algorithm I 231, Recommendation algorithm 2 232 and recommendation algorithm 3 233) and big data analysis learn the characteristics of the data for each customer’s membership data, so as to establish a recommendation model for recommending members’ consumer products and generate several indicators, as shown in the figure Index one 211, index two 212, and index three 213, the number of recommended algorithms and indicators is not limited to those shown in the figure.

The machine learning is a technique for prospecting data and analyzing data, which can extract meaningful information based on member data, and can establish characteristic descriptions for individual members based on the member data provided by customers, draw members' preferences, and cooperate with customers. To meet the purpose of the recommendation list.

3 shows a schematic diagram of another embodiment of a system for implementing a content recommendation method. This embodiment shows a machine learning algorithm 301 applied in a content recommendation system. The data provided by the customer may include member consumption records 33, product information 35, and customer information. 37 and environmental factors 39. Member data can be obtained through the user interface 31, and then a recommendation model 303 is established based on the results of data exploration and analysis, and then a recommendation list 305 is provided according to customer needs.

The consumption record 33, as the above-mentioned member consumption data, includes at least member information, consumption content, cost, location, and consumption date, wherein the characteristics can form a basis for member's preference judgment and become one of the parameters in the recommendation algorithm.

The commodity information 35 is commodity information sold by the client, such as the name, category, description, place of origin and price of the commodity. In addition to the commodity already sold, it can also include the commodity that will be sold in the future. Each product has its attributes. After being marked (labeled), products with the same or similar attributes can be categorized and become the content of the recommendation list 305, which is one of the parameters in the recommendation algorithm.

The customer data 37, including storefront data (such as location, business hours, size), industrial attributes, turnover, capital, etc., of the merchandise being sold become one of the parameters in the recommendation algorithm.

In addition, the factors that affect the recommendation result may also include environmental factors 39, such as seasonality, climate, and instant weather, which may form a reference basis for member consumption and become one of the parameters of the recommendation algorithm.

FIG. 4 then shows a flowchart of an embodiment of a content recommendation method executed in the disclosed content recommendation system.

In this process, at the beginning, as in step S401, the system receives the member data provided by the customer. The member data is the member's consumption data of the customer, including at least the member's information, consumption content, fees and consumption date.

In step S403, the software program in the system performs big data analysis and analyzes the member data. The purpose is to create a customer-specific database on the system side as in step S405, so that in step S407, multiple data can be executed for each customer's member data. Recommendation algorithms. Each recommendation algorithm is a data analysis method using machine learning technology. Through a large amount of data analysis, the relationship between the products sold by customers and the consumption of members is learned, a recommendation model is established, and then the data is introduced to generate a recommendation list. Multiple indicators are calculated from this recommendation list, as shown in step S409.

For the multiple indicators, according to the embodiment, for any combination of preference accuracy, profit margin and new product ratio calculated by the multiple recommendation algorithms for the customer's member data, each recommendation algorithm can be obtained These kinds of indicators, and the actual implementation can cover more indicators related to commercial sales, such as total revenue, return rate, inventory rate, etc.

According to an embodiment, a plurality of recommendation algorithms use a machine learning technique to learn the association between member consumption data and customer-supplied products based on customer-provided member data, so as to derive a product list that recommends each member by each recommendation algorithm, The preference accuracy of each recommendation algorithm is obtained after comparing the preferences of actual members; multiple recommendation algorithms use a machine learning technology to learn the correlation between member consumption data and the customer's profit based on the member data provided by the customer, Each recommendation algorithm is used to generate a product list that recommends each member, and then the profit rate is calculated based on the product list recommended by each member; and multiple recommendation algorithms use a machine learning technology to learn member consumption data and based on the member data provided by the customer The correlation between the products provided by the customer is used to obtain a list of products recommended by each member by each recommendation algorithm, and the ratio of new products is obtained by comparing the list of new products provided by the customer.

In this way, each recommendation algorithm can propose a ranking based on the indicators calculated by the customer data provided by the customer. For example, according to the accuracy of the customer's member preference, the recommendation algorithm is arranged in order from accurate to inaccurate. The profit rate, the recommendation algorithm is arranged in order from high to low, and the recommendation algorithm is arranged in order from high to low for the ratio of new products, which is provided to customers for reference. Therefore, the system provides customers through the menu. The recommended algorithm is as in step S411.

For example, when customers focus on the preferences of members based on actual needs, they may choose a recommendation algorithm that can provide a recommendation list with a higher accuracy of preferences; if the customer considers that the profit margin is greater than the preferences of members, it may choose to produce a high The recommendation algorithm of the profitability recommendation list; if the customer quote update rate is high and new products are often generated, the recommendation algorithm with a higher proportion of new products in the generated recommendation list may be considered.

FIG. 5 then shows a flowchart of an embodiment of the content recommendation method after the customer selects one of the recommendation algorithms.

In the content recommendation system, as in step S501, the computing module in the server executes the recommendation algorithm selected by the customer, performs machine learning on the member data, and obtains the characteristics in the data through data exploration, including the member The characteristics of consumption, the characteristics of commodities, etc., the purpose is to learn the correlation between the commodities sold by customers, member consumption and specific indicators, and establish a recommendation model, as in step S503.

Machine learning methods use regression algorithms, such as Decision Tree, Neural Network and Logistic Regression, and Neural Network Regression To perform data mining and data analysis on tagged data.

In step S505, the system provides a recommendation list for each customer, which records the list of commodities provided by the member for consumption, and then in step S507, based on the recommendation list, the customer's stock recommendation is also obtained.

The server can continuously receive updates of customer sales data, so that the machine learning method can continuously obtain updated data, and the recommendation model can be dynamically revised and optimized.

In summary, according to the embodiment of the content recommendation system proposed in the above disclosure, the content recommendation method in operation can be applied to the sales recommendation of commodities, and also allows companies selling commodities to obtain the recommendation algorithm provided by the system. Choose one of the various indicators out of the recommended indicators to implement a content recommendation method that can target specific indicators. The machine learning method used can build a recommendation model based on historical consumption record calculations, in addition to providing customers with specific recommendations for their members. In addition to the indicator’s commodity list, it also provides information on stocking, inventory and production needs.

The content disclosed above is only a preferred and feasible embodiment of this creation, and does not limit the scope of the patent application for this creation, so any equivalent technical changes made by using this creation specification and graphic content are included in this creation application Within the scope of the patent.

10‧‧‧ Internet 12‧‧‧Server 121‧‧‧computing module 122‧‧‧Recommended calculation module 123‧‧‧ Member database 111’‧‧‧Customer 1 database 112’‧‧‧Customer 2 database 113’‧‧‧ Customer Three Database 124‧‧‧User Interface Module 101‧‧‧Customer 1 111‧‧‧Database 1 102‧‧‧Customer 2 112‧‧‧Database 2 103‧‧‧Customer 3 113‧‧‧ Database 3 20‧‧‧Server 22‧‧‧Customer membership data 201‧‧‧computing module 203‧‧‧User interface 221‧‧‧Database 1 222‧‧‧Database 2 223‧‧‧Database 3 231‧‧‧Recommended Algorithm One 232‧‧‧Recommended Algorithm 2 233‧‧‧Recommended Algorithm 3 211‧‧‧I 212‧‧‧Indicator 2 213‧‧‧Indicator 3 301‧‧‧ machine learning algorithm 303‧‧‧Recommended model 305‧‧‧Recommended list 31‧‧‧User interface 33‧‧‧Member consumption records 35‧‧‧Commodity information 37‧‧‧Customer Information 39‧‧‧Environmental factors Steps S401～S411 Steps S501～S507 ‧‧‧ One of the content recommendation method flow

FIG. 1 shows a diagram of an embodiment of a system for implementing a content recommendation method;

FIG. 2 shows one of the schematic diagrams of the system embodiments for implementing the content recommendation method;

Figure 3 shows a second schematic diagram of an embodiment of a system for implementing a content recommendation method;

4 shows one of the flowcharts of the embodiments of the content recommendation method performed by the system; and

FIG. 5 shows a second flowchart of an embodiment of a content recommendation method executed by the system.

20‧‧‧Server

22‧‧‧Customer membership data

201‧‧‧computing module

203‧‧‧User interface

221‧‧‧Database 1

222‧‧‧Database 2

223‧‧‧Database 3

231‧‧‧Recommended Algorithm One

232‧‧‧Recommended Algorithm 2

233‧‧‧Recommended Algorithm 3

211‧‧‧I

212‧‧‧Indicator 2

213‧‧‧Indicator 3

Claims (10)

A content recommendation system, including: A server, which includes multiple databases, which are dedicated databases created based on analysis of member data provided by multiple customers; and A computing module, the content recommendation system executes a content recommendation method through the computing module, wherein a user interface provided through the server receives member data provided by a customer, and the member data is member consumption data of the customer , At least including member information, consumption content, cost and consumption date; then, the calculation module analyzes the member data to establish a customer-specific database, and executes multiple recommendation algorithms for the member data to obtain A plurality of indicators, determining one of the recommendation algorithms according to the plurality of indicators, and then executing the selected recommendation algorithm to establish a recommendation model, and providing the customer with a recommendation list through the recommendation model. The content recommendation system as described in claim 1, wherein the plurality of recommendation algorithms executed by the computing module are data analysis methods using machine learning technology, and the products sold by the customer and the products consumed by the members are learned through a large amount of data analysis Relevance, establish the recommendation model. The content recommendation system according to claim 2, wherein the recommendation list is a product list recommended by each member for consumption based on the recommendation model. The content recommendation system according to claim 3, wherein the plurality of indicators are any combination of a preference accuracy, a profit margin, and a new product ratio calculated by the plurality of recommendation algorithms for the customer's member data . The content recommendation system according to claim 4, wherein the plurality of recommendation algorithms executed by the computing module uses a machine learning technique to learn between the member consumption data and the goods provided by the customer based on the member data provided by the customer The relevance of each recommendation algorithm to generate a list of products that recommend each member, and compare the preferences of actual members to obtain the accuracy of the recommendation algorithm. The content recommendation system according to claim 4, wherein the plurality of recommendation algorithms executed by the computing module uses a machine learning technique to learn the relationship between the member consumption data and the customer's profit based on the member data provided by the customer Relevance, to obtain a recommendation list of products recommended by each member by each recommendation algorithm, and then calculate the profit rate according to the list of products recommended by each member. The content recommendation system according to claim 4, wherein the plurality of recommendation algorithms executed by the computing module uses a machine learning technique to learn between the member consumption data and the goods provided by the customer based on the member data provided by the customer Related to each recommendation algorithm to generate a list of products that recommend each member, and compare the list of new products provided to the customer to obtain the proportion of new products. The content recommendation system as described in any one of the request items 5 to 7, wherein each recommendation algorithm calculates the preference accuracy, the profit rate and the ratio of the new product based on the member data provided by the customer, and puts forward various indicators A ranking to determine one of the recommended algorithms. The content recommendation system described in claim 1 further includes a user interface module for implementing the user interface dedicated to communicating with each customer. The content recommendation system according to claim 9, wherein the user interface is an interface provided by an application program interface, a web browser, or a specific software program.

Images