KR101991178B1 - System and Methode Using Collaborative Filtering to Recommend Suitable Public Bid Information - Google Patents

Abstract

The present invention relates to a procurement announcement recommendation method and system using collaborative filtering. More specifically, the present invention relates to a procurement announcement recommendation method and system that can automatically recommend procurement announcements matching with oneself, And a method and system for recommending a procurement announcement, which can recommend the status of the procurement according to the procurement company by analyzing characteristics of the procuring company and the demanding organization. To this end, the present invention analyzes the records of the procurement announcements, obtains the user similarity between the users according to the ratio of overlapping of the procurement notices browsed by the users, and finds similar users Then, we obtain the combination matrix between the similar group users and the procurement announcements by obtaining the number of browse by the procurement announcements of the similar group users, and by using the matrix factorization technique for each of the procurement announcements, And provides the predetermined number of high procurement announcements having high relevance among the procurement announcements that the specific user has not read as a first recommendation notice to the specific user.

Description

[0001] The present invention relates to a collaborative filtering method and system,

The present invention relates to a procurement announcement recommendation method and system using collaborative filtering. More specifically, the present invention relates to a procurement announcement recommendation method and system using collaborative filtering, and more particularly, We propose a collaborative filtering method that can analyze the characteristics of procurement announcements by using collaborative filtering recommendation system and recommend users to procurement announcements appropriate to each user. And a method and system for recommending a procurement announcement.

At the present time, the R & D subjects in each field such as the country and the enterprise are developing the big data technology as the future strategy technology in order to utilize all the data generated in the connection and contact process as valuable ones. Trying hard. In particular, global companies are using Big Data to solve various problems, respond to issues, and utilize them as an important tool for future strategy and strategic decision making. According to WIKIBON data in 2015, the world's big data market will exceed $ 84.6 billion in 2026, domestic big data market will be 262.3 billion won by 2015, 30.2% . Many big data-related software are being researched and developed according to the growth potential and demand of Big Data.

Especially, it is based on the development of data accumulation analysis and utilization method in the information communication field, that is, wired / wireless communication technology and various mobile devices are rapidly developing. Big data technology, which collects and analyzes data to predict the future and create new value, is gradually expanding beyond the introduction stage. Big data technology is widely applied to enhance the competitiveness of existing industries and the efficiency of public services, and to create and innovate new service products.

As the value of Big Data increases, Public Sectors such as Public Procurement Service (PPS) have been making efforts to develop public system by reflecting big data technology in electronic procurement order system. As part of this, Big Data is used to analyze similarity And provide customized information. However, no clear results have been achieved until now.

On the other hand, various public institutions performing procurement business operate an electronic procurement system to support ordering of demand organizations and procurement companies. The e-procurement system is a service that supports all of the tasks related to the support of buyers of the public works through the information system. Most of the e-procurement systems currently in use are inconveniences such as retrieving announcements or company information one by one. It is difficult to easily find the surge in procurement announcements.

In particular, in order to find a suitable procurement notice for bidding in a company to which it belongs, it is necessary to review and review the contents one by one. This requires a lot of time and effort, so that users of the procurement notification system have been complaining. In order to solve these problems, it may be possible to recommend appropriate procurement announcements by analyzing the contents of the procurement announcement or the user's company information, but this is not easy because it requires a lot of resources such as detailed classification and analysis of the announcement contents .

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and it is an object of the present invention to provide a system and method for managing the procurement announcement, The present invention aims to provide a system and method for recommending a procurement notification suitable for a user through collaborative filtering (CF), even if it is not analyzed or classified. However, the objects of the present invention are not limited to those mentioned above, and other objects not mentioned can be clearly understood by those skilled in the art from the following description.

According to an aspect of the present invention, there is provided a method for recommending procurement announcements using collaborative filtering, the method comprising: receiving a browse record for a user viewing a procurement notification from a procurement notification system; file system in a procurement announcement recommendation system for recommending a procurement announcement to a user by analyzing the browse history to obtain a user similarity degree among users, Acquiring similarity users included in the second group; Obtaining a combination matrix between the similar group users and the procurement announcements by obtaining the number of browsing views per the procurement announcements of the similar group users; Applying a matrix factorization technique to the combination matrix to obtain a score of relevance to the similar group users for each of the procurement announcements; And providing a predetermined number of procurement announcements having a high relevance score among the procurement announcements that the specific user has not viewed as a first recommendation notice to the specific user; And a plurality of pixels.

Further, in addition to the above-described features, the user similarity may be calculated based on a ratio of overlapping of the procurement notices viewed by the user to the procurement announcement recommendation method using collaborative filtering.

In addition, the procurement announcement recommendation method using collaborative filtering according to the present invention is characterized in that, in addition to the above-described features, Pearson Correlation is applied to users' browsing records for each procurement announcement, To obtain a similarity degree; And obtaining a predetermined number of procurement announcements having a high publicity similarity for each of the procurement announcements included in the first recommendation announcement and providing the procurement announcement to the specific user as a second recommendation announcement; As shown in Fig.

Meanwhile, the procurement announcement recommendation system using collaborative filtering according to the present invention includes a Hadoop distributed file system (HDFS) for receiving and storing a browse record for each user who views the procurement announcement from the procurement notification system, Wherein the server analyzes the browsing history to obtain a user similarity between the users according to a ratio of overlapping of the procurement notices viewed by the user, After obtaining the users of the military users, the number of browsing of the similar user users by the procurement announcement is obtained, and the combination matrix between the users of the similar user and the procurement announcements is obtained. Then, through the matrix factorization method, The similarity group users are related to each other, One trillion months in a non-viewing notification is preferably characterized in that it provides the particular user with procurement announcement of the higher predetermined number of high relevance to the first announcement like.

Further, in addition to the above-described features, the server obtains a publicity similarity between the respective procurement announcements by applying Pearson Correlation to the user browse record for each procurement announcement, Further comprising the step of obtaining a predetermined number of procurement announcements having a high degree of similarity of announcement for each of the procurement announcements included in the announcement, and then providing the procurement announcement to the specific user as a second recommendation announcement, It is also possible to do.

As shown in the above description, the procurement announcement recommendation method and system using collaborative filtering according to the present invention allows a user who uses a procurement system to bidding his / her own company without searching for a business announcement that matches his / It is possible to automatically recommend an optimal procurement announcement.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a system overview of a method and system for recommending procurement announcements using collaborative filtering according to the present invention.
FIG. 2 is a schematic diagram showing an outline of a recommended program included in the procurement announcement recommendation system using collaborative filtering according to the present invention.
FIG. 3 is a conceptual diagram illustrating a concept of collecting browsing records in a procurement announcement recommendation method and system using collaborative filtering according to the present invention.
FIG. 4 is a conceptual diagram for obtaining user similarity between users from a browsing record in a procurement announcement recommendation method and system using collaborative filtering according to the present invention.
FIG. 5 is a conceptual diagram for obtaining a recommendation value by an ALS recommendation algorithm in a procurement announcement recommendation method and system using collaborative filtering according to the present invention.
6 is an example of a first recommendation announcement screen using ALS recommendation algorithm recommendation values in the procurement announcement recommendation method and system using collaborative filtering according to the present invention.
FIG. 7 is a conceptual diagram for obtaining publicity similarity by applying Pearson correlation in a procurement announcement recommendation method and system using collaborative filtering according to the present invention.
FIG. 8 is an example of a recommendation announcement recommendation method using collaborative filtering according to the present invention and a second recommendation announcement screen using a Pearson correlation in the system.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will be more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which: FIG. Accordingly, those skilled in the art can easily implement the technical idea of the present invention. In describing the present invention, it is to be understood that any known technology related to the present invention has already been known in the art, and if a detailed description of the known technology is considered to unnecessarily obscure the gist of the present invention, It will be omitted.

In addition, although the term used in the present invention is selected as a general term that is widely used at present, there are some terms selected arbitrarily by the applicant in a specific case. In this case, since the meaning is described in detail in the description of the relevant invention, It is to be understood that the present invention should be grasped as a meaning of a term that is not a name of the present invention.

Embodiments set forth in the following description can be varied in various ways and have various additional embodiments, in which specific embodiments are shown in the drawings and the detailed description thereof is set forth in the accompanying drawings. It should be understood, however, that the embodiments are not intended to be limited to the particular forms, but are to be understood to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the embodiments. In connection with the description of the drawings, like reference numerals have been used for like elements.

The terminology used in the description of the various embodiments is used only to describe a specific embodiment and is not intended to limit the embodiments. The singular expressions include plural expressions unless the context clearly dictates otherwise.

It should be noted that the terms such as " comprising " or " may include " may be used in describing the embodiments to indicate the presence of the inventive function, operation, or element, Not limited. Also, in the description of the various embodiments, terms such as " comprise " or " have ", and the like, specify that the presence of stated features, integers, But do not preclude the presence or addition of one or more other features, numbers, steps, operating elements, components, or combinations thereof.

The user terminal described in various embodiments includes various electronic devices capable of wireless communication, such as a smart phone, a smart watch, a tablet PC, a mobile phone, a videophone, an electronic book reader, a desktop PC, a laptop PC, A netbook computer, a PDA, a PMP, an MP3 player, a camera and other similar devices.

BRIEF DESCRIPTION OF THE DRAWINGS Fig. The present invention relates to a method and system for using a collaborative filtering recommendation scheme, and recommends a procurement announcement most closely related to a user according to a personalization scheme. In general, the recommendation system is a system for recommending filtered content among a large amount of contents to a user. The recommendation method used by the recommendation system is a collaborative filtering recommendation method that recommends contents that users with similar tendencies are common to users, a content-based recommendation method that recommends contents that are similar to contents previously used by the user, Demographic recommendation methods that analyze demographic information, recommend rules, and recommend content. Among them, collaborative filtering recommendation method is known as the most successful recommendation method. For example, it is widely used on Internet sites such as Amazon and Netflix offering book recommendations.

The collaborative filtering recommendation method is a personalization technique for recommending a related service to a user by using other users 'preferences similar to a user' s preferred pattern, and is classified into a user - based collaboration filtering recommendation method and an item - based collaboration filtering recommendation method. Among them, the user-based collaborative filtering recommendation method measures similarity among users and recommends a product preferred by a specific user based on a product evaluated by another user having similar preferences.

1 is a schematic diagram of a system used in a procurement announcement recommendation method and system using collaborative filtering according to the present invention. As shown in FIG. 1, the procurement announcement recommendation system 10 using collaborative filtering according to the present invention preferably includes a Hadoop distributed file system (HDFS) 200 and a server 300 . The Hadoop distributed file system 200 is connected to a system in which various procurement bids announcements related to the procurement notification system 100, that is, the procurement notification system 100, are received, receives the log information from the procurement notification system 100 . Here, the log information is a record in which users of the procurement notification system 100 browse various procurement notices displayed in the procurement notification system 100, and may store information for identifying a user, Information that can identify the announcement, and the time to be viewed, that is, the click time, and the like.

FIG. 2 is a schematic diagram of a recommendation recommendation program included in the procurement announcement recommendation system 10 using collaborative filtering according to the present invention. As shown in FIG. 2, the announcement recommendation program is preferably installed in the server 300.

In addition, the announcement recommendation program is preferably implemented in Spark application for high performance, and more preferably, the Spark application is installed in all Hadoop cluster servers. The Spark application is a framework that provides cluster-based distributed processing similar to Hadoop. Unlike Hadoop, which always maintains processing results in a file system, Spark applications provide functions that can be stored and managed in memory, This is because the performance for data processing is excellent.

The announcement recommendation program reads the browsing record (click log information, user, item, simple data of the click time format, unstructured data item) stored in the Hadoop distributed file system 200 (HDFS) Filtering, CF) and store it in Nosql called Redis. Collaborative filtering (CF) is an algorithm that performs recommendation using only relationship information (here, whether or not the procurement announcement is clicked and the number of times) without using an item or an attribute of a user. It is preferable to generate the recommendation information based on the base and items (procurement announcements).

User-based announcement recommendation allows users to provide announcement information that they recommend. It is desirable that the user-based recommendation program is implemented by utilizing the Spark ML library, ALS module, in such a way that the user similarity degree is obtained and the similar user highly recommends the unseen announcement in this announcement. And item-based announcement recommendation should be based on the announcements you have previously seen, so as to provide other announcement information that is relevant to the announcement.

FIG. 3 is a conceptual diagram illustrating a browsing history collected from the procurement notification system 100. It is preferable to use a log collector called Flume when collecting browsing records or logs from the procurement notification system 100. This is because Flume reliably and reliably collects logs from different servers, and the main feature is that logs can be collected from a variety of sources and stored in Hadoop. It also has the advantage that it is easy to connect the agents and draw the desired structure.

Accordingly, in the procurement announcement recommendation system using the collaboration filtering according to the present invention, the log is collected through the Flume. When the click log of the procurement announcement is generated in the WAS server through the Flume technology, the transmission / And the procurement announcement click log information is stored as the browsing record including the user (user), the item (procurement notification), the click time, and the like, which are raw data for announcement recommendation.

On the other hand, the server 300 analyzes the browsing history and obtains the user similarity between the users of the procurement notification system 100. The server 300 uses the user similarity to include each user in an upper certain ratio And the like. Here, it is preferable that the user similarity is calculated according to the ratio of the users who have browsed the respective procurement notices to each other.

FIG. 4 is a conceptual diagram of the server 300 in which the server 300 obtains user similarities from users' browsing records and obtains similar user groups for each user using the user similarities. As shown in FIG. 4, the user 1 browsed the announcement numbers 1001, 1002, and 1003, the user 2 browsed the announcement numbers 1002 and 1004, the user 3 the announcement numbers 1001 and 1003, the user 4 the announcement numbers 1003 and 1005, The publicity numbers 1001 and 1002 are recorded as viewing. As a result of analyzing the user similarity to other users based on the specific user user1, the user3 and the user5 overlap with the notices read by the specific user user1 , user2 and user4 are overlapped one by one. Therefore, user3 and user5 are users having a relatively high degree of user similarity with respect to a specific user user1. Therefore, user3 and user5 will be similar users of user1, which is a specific user.

After the server 300 obtains the user similarity and the similar user for a specific user, it is preferable to apply the ALS (Alternating Least Squares) recommendation algorithm to obtain the first recommendation announcement. FIG. 5 is a conceptual diagram for obtaining a recommendation value using an ALS (Alternating Least Squares) recommendation algorithm in the present invention.

As shown in FIG. 5, ALS (Alternating Least Squares) recommendation algorithm is applied in the procurement announcement recommendation system using collaborative filtering according to the present invention. In the input process, The user clicks the procurement announcement, and the number of times each similar group user clicks on each procurement announcement in Grouping and Count process is derived. That is, the similar group users obtain the number of browsing by the procurement announcement, and then obtain the combination matrix between the similar group users and the procurement announcements.

Next, the matrix analysis is used to analyze the relationship between each similar user and each procurement announcement. A matrix factorization technique is applied to the combination matrix, It is preferable that the sum of the scores of association with all the users in the user group in the similar group is the relevance score for each procurement announcement .

Next, the server 300 is requested to obtain a unread preview notification that has not yet been browsed by the specific business operator, and the first predetermined number of procurement notices having a high relevance score among the procured public notices It is preferable to provide the recommendation notice to the specific user. FIG. 6 shows an example of a first recommendation announcement screen to which an ALS recommendation algorithm recommendation value is applied in the present invention.

According to another embodiment of the present invention, there is also provided a method for obtaining similarity between procurement announcements by applying Pearson Correlation and recommending a proper procurement announcement to a user by using the same, The method comprising the steps of: obtaining a publicity similarity between the respective procurement announcements by applying a Pearson correlation to users' browsing records of the procurement announcements of the procurement announcements of the procurement announcements, And providing the second recommendation announcement to the specific user.

FIG. 7 is a conceptual diagram for obtaining the announcement similarity by applying the Pearson correlation, and FIG. 8 is an example of the second recommendation announcement screen to which the Pearson correlation is applied. As shown in FIG. 7, the server 300 receives the information that all the users who use the procurement notification system 100 clicked the procurement announcement, and the number of times that the users clicked the respective procurement announcements in the Grouping and Count process . Next, it is preferable to analyze the relationship between each procurement announcement by matrix analysis and apply the Pearson correlation algorithm to calculate the degree of similarity between the procurement announcements.

After the announcement similarity is calculated, the server 300 obtains the procurement announcements of each of the procurement announcements included in the first recommendation announcement and the predetermined fixed number of announcement similarities, It is preferable to provide it. For example, if the specific user user1 has received the recommendation of announcement 1100, announcement 1101 and announcement 1103 with the first recommendation announcement, and assumes that the higher-order schedule number with high notification similarity is 2, the user user1, The announcement 1101 and the announcement 1023, which are the highest two announcements, are provided by the second recommendation announcement, and the announcement 1201, the announcement 1023, and the announcement 1103, A second recommendation notice can be provided.

Meanwhile, it is preferable that the announcement recommendation program executed by the server 300 stores a recommendation result, that is, a first recommendation announcement and a second recommendation announcement for each user, in Radis (ver 3.2.8). The Redis is a memory-based Nosql & Cash solution, because it is memory-based and therefore optimized for fast processing speed and key-value storage.

Although the present invention has been described with reference to the above examples, the present invention is not necessarily limited to these examples, and various modifications may be made without departing from the scope of the present invention. Therefore, the examples disclosed in the present invention are not intended to limit the scope of the present invention and are not intended to limit the scope of the present invention. The scope of protection of the present invention should be construed by the following claims, and all technical ideas within the scope of equivalents thereof should be construed as being included in the scope of the present invention.

100 procurement notification system
200 HDFS
300 server

Claims (5)

A procurement announcement recommendation system that receives a browse record of a user who views a procurement notice from a procurement notification system and stores the read record in a Hadoop distributed file system (HDFS) ,
Analyzing the browse history to obtain user similarities among users, and then obtaining similar user users whose user similarities are included in a predetermined ratio for a specific user;
Obtaining a combination matrix between each of the similar group users and each of the procurement notices by obtaining the number of browsing per procurement announcement for each of the similar group users;
Applying a matrix factorization technique to the combination matrix to obtain a sum of scores of association scores with the similar group users for each of the procurement announcements;
Providing a predetermined number of procurement announcements in the order of the highest relevance score among the procurement announcements that the specific user has browsed as the first recommendation notice to the specific user;
Analyzing a relationship between procurement announcements of each procurement announcement by a matrix analysis of users' browsing records, and applying a Pearson correlation to obtain a publicity similarity between the procurement announcements; And
Wherein the plurality of procurement announcements for each of the procurement announcements included in the first recommendation announcement are obtained in ascending order of the degree of similarity among the procured announcements not yet read by the specific user, Providing to a specific user; , ≪ / RTI &
Wherein the user similarity is calculated according to a ratio of overlapping of the procurement notices viewed by the user. delete delete A Hadoop distributed file system (HDFS) for receiving and storing a browse record for each user who views the procurement notice from the procurement notification system, and a server,
The server comprises:
- analyzing the browsing history to obtain user similarities between users according to a ratio of overlapping of the procurement notices viewed by the users, finding similar group users whose user similarity is included in a certain percentage for a specific user, A similarity matrix is obtained for each of the similar group users and each of the procurement notices by obtaining the number of browse by each procurement announcement for each of the similar group users, and for each of the procurement announcements through the matrix factorization method, A second recommendation announcement of a certain number of procurement announcements in descending order of the sum of the relevance scores among the procurement announcements not yet visited by the specific user,
- Analyzing the relationship between the procurement announcements by matrix analysis of the user browse records for each procurement announcement and then applying the Pearson Correlation to obtain the announcement similarity between the respective procurement announcements, 1 procurement announcements are obtained from the procurement announcements that the specific user has browsed among the procurement announcements in descending order of the notification similarity, A procurement announcement recommendation system using collaborative filtering delete

Images