CN111159561A - Method for constructing recommendation engine according to user behaviors and user portrait - Google Patents
Method for constructing recommendation engine according to user behaviors and user portrait Download PDFInfo
- Publication number
- CN111159561A CN111159561A CN201911416982.1A CN201911416982A CN111159561A CN 111159561 A CN111159561 A CN 111159561A CN 201911416982 A CN201911416982 A CN 201911416982A CN 111159561 A CN111159561 A CN 111159561A
- Authority
- CN
- China
- Prior art keywords
- information
- user
- recommended
- house
- source
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 230000006399 behavior Effects 0.000 title abstract description 82
- 238000012545 processing Methods 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims abstract description 8
- 238000012163 sequencing technique Methods 0.000 claims description 29
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000012216 screening Methods 0.000 claims description 7
- 238000005516 engineering process Methods 0.000 claims description 4
- 230000003287 optical effect Effects 0.000 claims description 3
- 230000002776 aggregation Effects 0.000 claims description 2
- 238000004220 aggregation Methods 0.000 claims description 2
- 238000004891 communication Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000002860 competitive effect Effects 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000007670 refining Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000004931 aggregating effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 238000013501 data transformation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 238000011068 loading method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0645—Rental transactions; Leasing transactions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/12—Hotels or restaurants
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Physics & Mathematics (AREA)
- Finance (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- Data Mining & Analysis (AREA)
- Development Economics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses a method for constructing a recommendation engine according to user behaviors and user portraits, which relates to the technical field of data statistics, and a specific implementation mode of the method comprises the following steps: collecting user information and house source information; the user information comprises user behavior information and a preset frequency threshold, and the user behavior information is valid when the operating frequency of the user behavior is greater than the frequency thresholds corresponding to the types; the system also comprises user portrait information, wherein the user portrait information comprises user consumption information and user preference information; generating a first recall data source and a second recall data source, generating a first recommended data source according to LBS service, performing feature engineering processing through a Spark distributed framework to obtain a second recommended data source, training to obtain a ranking model to obtain a first recommended house source, generating a second recommended house source according to the requirements of a house owner, and constructing a recommendation engine. According to the implementation mode, the accuracy and the efficiency of the user behavior information are improved and the accuracy of the training model is improved by setting the frequency threshold.
Description
Technical Field
The application relates to the technical field of data statistics, in particular to a method for constructing a recommendation engine according to user behaviors and user portraits.
Background
With the development of the internet, the interests of users are more and more extensive, and with the change of the environment and living standard of users, the demands of users are also changing. At present, the LBS (Location Based Services) recommendation engine recall and sorting method has the problems in the applicability of the long-rented apartment field:
1. the current long-rent apartment collection user information is generally only according to the online record of the user, and the channel is single, so that the user information is omitted, and the sample information base is lost;
2. the recall mode industry is integrally recall by inverted indexes and user behaviors, and does not consider the properties of bulk objects such as high-quality plot prices and the like of high-quality public houses;
3. the general recommendation rough arrangement stage in the rough arrangement layer is not optimized for a high-value sparse target scene, a long-rented apartment scene needs multilayer rough arrangement, LBS and value matching strategies need to be carried out for matching of strong intentions of clients and key attributes of target objects, and matching strategies need to be carried out for secondary requirements of users and secondary house states and district characteristics of houses;
4. the ranking layer industry generally bases on the user online behavior to do the learning ToRank model ranking of pair-wise or list-wise. However, in a long-rented apartment, the conditions of on-line feedback of users and the like need to be considered, the sequencing model is more complex, multi-level sample weight depiction needs to be performed on the basis of the sequencing model, and meanwhile, the scenes such as areas, house states and the like need to be deeply subdivided on the basis of the high sparsity of the samples and the large-volume attribute of transactions, so that the recommendation quality is improved;
5. in the interference adjustment layer industry, accurate advertisement insertion is generally performed based on a bidding ranking mode, modeling is not performed based on a specific LBS mode of a long rental apartment, and the recommendation effect is not ideal.
Disclosure of Invention
In view of the above, the present application discloses a method for constructing a recommendation engine according to user behavior and a user profile, comprising the steps of:
collecting user information and house source information;
the user information comprises user behavior information, the type of the user behavior information in a preset period and the operation frequency of the user behavior corresponding to each type are counted, a frequency threshold value is preset, and when the operation frequency of the user behavior is greater than the frequency threshold value corresponding to each type, the user behavior information comprises effective user behavior information; the types of the user behavior information comprise user browsing webpage information and user renting duration;
the user information further includes user representation information, the user representation information including user consumption information and user preference information; the house source information is divided into real-time user behavior information and historical user behavior information according to the effective user behavior information; the real-time user behavior information is examined and checked, and the real-time user behavior information enters an online database in real time according to a distributed stream processing open source framework; generating a log from the historical user behavior information, and inputting the log into an offline data warehouse in batches through a Spark frame calculation engine; generating a first recall data source for the online database and the offline data repository;
the house source information is recalled according to the user portrait information, and a second recall data source is obtained by screening through an optical disc arranged in an offline data warehouse and LBS (location based service);
carrying out price matching and section feature matching on the first recall data source and the second recall data source according to LBS service to obtain corresponding matching degrees, and sequentially sequencing from high to low according to the matching degrees to generate a first recommended data source;
performing feature engineering processing on the first recommended data source through a Spark distributed framework, and grading to obtain a second recommended data source;
training the second recommended data source to obtain a ranking model;
inputting the house source information into the sequencing model to obtain the grading sequencing condition of the house source information;
setting a scoring threshold value, wherein the house source with the score of the house source information larger than the scoring threshold value is a first recommended house source;
meanwhile, generating a second recommended house source according to the house owner demand;
and constructing a recommendation engine according to the first recommended house source and the second recommended house source.
Preferably, the user information is collected by means of online, offline and electrical pinning.
Preferably, the number of the first recommended house resources is greater than the number of the second recommended house resources.
Preferably, the ratio of the number of the first recommended house sources to the number of the second recommended house sources is 5: 1.
Preferably, the feature engineering process includes sequentially performing aggregation, refining, screening, and completion on the first recommended data source to obtain the second recommended data source.
Preferably, the inputting the room source information into the ranking model to obtain the ranking condition of the scores of the room source information includes: and carrying out weight depiction on the house source information, and carrying out deep division on the scenes such as regions, house states and the like according to the sparsity and the transaction attribute of the house source information to obtain the grading and sequencing conditions of the house source information.
Preferably, the recommendation engine further comprises a hotspot source.
Compared with the prior art, the method for constructing the recommendation engine according to the user behaviors and the user portrait provided by the invention has the following beneficial effects that:
1. the method for constructing the recommendation engine according to the user behaviors and the user portrait, provided by the invention, has the advantages that the frequency threshold of user behavior operation is set, and the accuracy and the efficiency of user behavior information can be improved.
2. According to the method for constructing the recommendation engine according to the user behaviors and the user portrait, the relationship between the first recommended house source and the second house source is limited, so that the income of a long-rented apartment can be improved.
3. The method for constructing the recommendation engine according to the user behaviors and the user portrait provided by the invention has rich information acquisition channels, including online APP recording, offline watching and communication and electric marketing modes of the user; a large amount of user information is collected, an integral sample model is generated, and the bulk object attributes such as high-quality parcel prices of high-quality long-rent apartments are covered.
4. The method for constructing the recommendation engine according to the user behaviors and the user portrait optimizes the personalized recommendation mode, and extracts apartments with high recommendation matching degree from the sample model library through comprehensively measuring and calculating the behaviors and the portrait of the user.
5. According to the method for constructing the recommendation engine according to the user behaviors and the user portrait, the recommendation sequencing model is optimized, a new matching strategy is formulated for the conditions of online and offline watching, signing and the like, the sequencing model is subjected to multilevel sample weight portrayal, and the complexity of the existing sequencing model is reduced; meanwhile, scenes such as regions, house states and the like are deeply subdivided according to the high sparsity of samples and the large-amount attribute of transaction, and the quality of the recommended apartment is improved.
6. The method for constructing the recommendation engine according to the user behaviors and the user portrait optimizes an intervention adjustment layer, models based on a specific LBS (location based service) mode of the long-rented apartment, and designs the characteristic recommendation engine of the long-rented apartment with bulk commodities with sparsely distributed landmass.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
FIG. 1 is a flow diagram of a method of constructing a recommendation engine based on user behavior and user profiles in accordance with the present invention;
FIG. 2 is a flow chart of another method of constructing a recommendation engine based on user behavior and user profile in accordance with the present invention.
Detailed Description
The technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It should be noted that the described embodiments are merely some embodiments, rather than all embodiments, of the invention and are merely illustrative in nature and in no way intended to limit the invention, its application, or uses. The protection scope of the present application shall be subject to the definitions of the appended claims.
Example 1:
fig. 1 is a flowchart of a method for constructing a recommendation engine according to user behaviors and a user profile according to the present invention, and as shown in fig. 1, the present embodiment provides a method for constructing a recommendation engine according to user behaviors and a user profile, which includes the steps of:
step 101, collecting user information and house source information;
the user information comprises user behavior information, the types of the user behavior information in a preset period and the operation frequency of the user behavior corresponding to each type are counted, a frequency threshold value is preset, and when the operation frequency of the user behavior is greater than the frequency threshold value corresponding to each type, the user behavior information comprises effective user behavior information; the types of the user behavior information comprise webpage browsing information of a user and renting time of the user;
the user information also comprises user portrait information which comprises user consumption information and user preference information; it can be understood that when the user browses a house information by using the app, the operation is invalid when the user is not actually needed by mistake, and when the browsing of a house information exceeds a certain frequency or the browsing of the house of the same kind exceeds a certain frequency, the user can definitely know that the user wants to live in the house, namely the user behavior information at the moment is valid information, so that the frequency threshold value of the user behavior operation is set, and the accuracy and the efficiency of the user behavior information can be improved.
In step 101, user information may be collected by online, offline, and electrical pinning. It can be understood that, in the first online acquisition mode, according to browsing and searching records of a user on a webpage and an APP, user intention information is acquired; the second is an offline acquisition mode, which acquires the intention information of the user through the carrying, signing and communication of a specially-assigned person; the third is a mode of electric marketing, which is used for determining the intention of the user through communication with the user by telephone; the method is characterized by enriching an information acquisition channel, acquiring a large amount of user information, generating an integral sample model, and covering the attributes of a large number of targets such as the price of a high-quality land parcel owned by a high-quality long-rent apartment.
102, dividing the house source information into real-time user behavior information and historical user behavior information according to the effective user behavior information; the real-time user behavior information is examined and checked, and the information enters an online database in real time according to a distributed stream processing open source framework (Flink); generating a log from the historical user behavior information, and inputting the log into an offline data warehouse HIVE in batches through a Spark frame calculation engine; generating a first recall data source by an online database and an offline data warehouse HIVE;
and the house source information is recalled according to the user portrait information, and a second recall data source is obtained by screening through an optical disk UDF (Universal description framework) arranged in a Hive offline data warehouse and LBS (location based service).
In step 102, optionally, the real-time behavior can capture the current intention of the user through two parts of historical behavior and real-time behavior recall. Wherein, the Flink executes any stream data program in a data parallel and pipeline mode, and the pipeline runtime system of the Flink can execute batch processing and stream processing programs. Spark is a Hadoop MapReduce-like universal parallel framework sourced by UC Berkeley AMP lab (AMP laboratories, burkeley, university, ca). HIVE is a data warehouse tool based on Hadoop, is used for data extraction, transformation and loading, and is a mechanism capable of storing, inquiring and analyzing large-scale data stored in Hadoop.
In step 102, optionally, the data is recalled by way of workplace, place of life, and point of interest, with the least importance. But as a solution for cold start without insufficient collection of user behavior. UDF is an english abbreviation of Universal Disc Format (Universal Disc Format), a Universal Disc file system established by the international organization for standardization in 1996, and standard Packet Writing technology (PW) is used to simplify the use of a recorder.
103, performing price matching and section feature matching on the first recall data source and the second recall data source according to LBS service to obtain corresponding matching degrees, and sequentially sequencing from high to low according to the matching degrees to generate a first recommended data source, wherein the first recommended data source comprises an online data source and an offline data source;
in step 103, the first recall data source and the second recall data source are sorted according to the LBS service price matching strategy and the segment characteristics matching strategy to generate a first recommended data source.
And further, an LBS and value matching strategy is carried out for matching the strong intention of the client and the key attributes of the target objects, and a matching strategy is carried out for the secondary requirements of the user and the characteristics of the secondary house state and the section of the house, so that a rough sequencing result is obtained. And optimizing an individualized recommendation mode, comprehensively measuring and calculating the behaviors and pictures of the user, and extracting a high-quality apartment with high recommendation matching degree from a sample model library. Meanwhile, a recommended sequencing model can be optimized, a new matching strategy is formulated for the conditions of online and offline watching, signing and the like, multi-level sample weight depiction is carried out on the sequencing model, and the complexity of the existing sequencing model is reduced; meanwhile, scenes such as regions, house states and the like are deeply subdivided according to the high sparsity of samples and the large-amount attribute of transaction, and the quality of the recommended apartment is improved.
In the above steps, it can be understood that the user information correspondingly generates a first recall data source, a second recall data source and a third recall data source through the user demand information, the user behavior information and the user portrait information house source information, and processes the first recall data source, the second recall data source and the third recall data source to generate a first recommended data source; some user information with relevancy can be recalled from massive user information and roughly sorted, the subsequent fine sorting can be carried out, the workload of work is reduced, and the response time of the obtained result is favorably prolonged.
104, performing feature engineering processing on the first recommended data source through a Spark distributed framework, and grading to obtain a second recommended data source;
it is to be understood that since the first recommended data source includes an online data source and an offline data source; the Spark distributed framework comprises an offline calculation module and an online module calculation module, the online data source of the first recommended data source is processed by the Spark distributed framework comprising the online calculation module, and the offline data source of the first recommended data source is processed by the Spark distributed framework comprising the offline calculation module, so that the response time of the obtained result can be shortened.
105, training a second recommended data source to obtain a ranking model;
step 106, inputting the house source information into a ranking model to obtain the scoring ranking condition of the house source information;
in step 106, the house source information is further subjected to weight portrayal, and scenes such as areas, house states and the like are deeply divided according to the sparsity and the transaction attributes of the house source information to obtain the ranking score of the house source information.
Step 107, setting a score threshold value, wherein the house source with the score of the house source information larger than the score threshold value is a first recommended house source;
in steps 104-107, it is understood that the feature engineering process includes aggregating, refining, screening, and complementing the first recommended data source to obtain a second recommended data source. And further training a corresponding sorting model after the first recommended data source is collected and processed through characteristic engineering measurement and calculation of a Spark frame, performing weight portrayal on sample data, performing deep division on scenes such as regions, house states and the like according to the high sparsity of data information and the large-amount attribute of transaction, giving accurate sorting scores of house source information, and finally recommending high-quality long-rented apartment resources to users.
Step 108, generating a second recommended house source according to the demands of the homeowners;
and step 109, constructing a recommendation engine according to the first recommended house source and the second recommended house source. Optionally, the recommendation engine further includes a hotspot source. In steps 108-109, it can be understood that after modeling based on LBS specific to the long rental apartment, the system inserts some specific house sources into specific positions in the recommendation list according to customized requirements of the company such as the out-of-house tilt policy, the operation activity policy, the advertisement promotion, etc., and recommends more selection spaces for the user; and if the recommendation result is insufficient, the system automatically recommends the completion of the related hot spot competitive house resources. An optimization intervention adjustment layer is arranged, modeling is carried out based on a specific LBS mode of the long-renting apartment, and a characteristic recommendation engine of the long-renting apartment with large commodities with sparsely distributed landmasses is designed.
And the number of the first recommended house sources is larger than that of the second recommended house sources. The ratio of the number of the first recommended house sources to the number of the second recommended house sources is preferably 5: 1. The method can ensure that the recommended requirements of the users are met, and can also improve the advertising benefits of long-rented apartments and the like.
Example 2:
FIG. 2 is a flowchart illustrating a method for building a recommendation engine according to user behavior and user profile according to the present invention, and as shown in FIG. 2, this embodiment provides a method for building a recommendation engine according to user behavior and user profile, which includes the steps of:
step 201, user information acquisition:
the information collection method mainly includes three methods: the method comprises the steps that a first online acquisition mode is adopted, and user intention information is obtained according to browsing search records of a user on a webpage and an APP; the second is an offline acquisition mode, which acquires the intention information of the user through the carrying, signing and communication of a specially-assigned person; the third is a mode of electric marketing, which is used for determining the intention of the user through communication with the user by telephone; constructing a sample model of the information acquired by all channels, storing the sample model in a database and facilitating retrieval; enriching collected information channels, including on-line APP recording, off-line communication with watching and selling modes of a user; a large amount of user information is collected, an integral sample model is generated, and the attributes of large objects such as high-quality parcel prices of high-quality long-rent apartments are covered.
Step 202, recall:
after a user searches for information of a long rental apartment near 'XX prefecture', firstly, user requirements are acquired through a sample database: recalling according to the user behavior: the real-time behavior of the user is subjected to data cleaning and enters a database in real time through a Flink processing framework; historical behavior logs are generated for each behavior of the user and collected, and the collected historical behavior logs are input into a HIVE offline data warehouse in batches through a Spark calculation engine; synchronously storing the warehousing data into a Redis real-time database;
collecting and carefully selecting the acquired data information, then carrying out multilayer rough sequencing on the long-rent apartment scenes, carrying out LBS and value matching strategies aiming at the matching of the strong intentions of the clients and the key attributes of the target objects, and carrying out matching strategies on the secondary requirements of the users, the secondary house state and the district characteristics of the house to obtain a rough sequencing result; a new matching strategy is formulated for the conditions of on-line and off-line watching, signing and the like, and a multi-level sample weight depiction is carried out on the sequencing model, so that the complexity of the existing sequencing model is reduced; meanwhile, scenes such as regions, house states and the like are deeply subdivided according to the high sparsity of samples and the large-amount attribute of transaction, and the quality of the recommended apartment is improved.
Step 203, sorting:
for recalled house source data rough sorting information, after comprehensively acquiring historical behaviors of users and house-state plot business circle images and performing characteristic engineering measurement and calculation processing on a Spark distributed framework, respectively performing corresponding training on online data and offline data to construct a sorting model, performing weight portrayal on sample data, performing deep division on scenes such as regions, house states and the like according to the high sparsity of data information and the large-volume attribute of transaction, giving accurate sorting scoring of house source information, and finally recommending high-quality long-renting apartment resources to the users;
step 204, intervention and adjustment:
after modeling based on the specific LBS of the long rental apartment, the system inserts some specific house sources into specific positions in a recommendation list according to customized requirements of a company such as a house-leaving inclined policy, an operation activity policy, advertisement promotion and the like, and recommends more selection spaces for users; and if the recommendation result is insufficient, the system automatically recommends the completion of the related hot spot competitive house resources. And optimizing an intervention adjustment layer, modeling based on a specific LBS mode of the long-renting apartment, and designing a characteristic recommendation engine of the long-renting apartment with large commodities with sparsely distributed plots.
It is understood that the recall method of the present invention may be based on only one of the user behavior information and the image information, or may be based on both of the above two methods. The embodiment only shows that the recall is performed according to the user behavior, and the embodiment does not make specific requirements on the recall mode and can be set according to actual conditions.
Example 3:
with continued reference to FIG. 2, FIG. 2 is a flowchart illustrating a method for building a recommendation engine based on user behavior and a user profile according to the present invention; the embodiment provides a method for constructing a recommendation engine according to user behaviors and user portraits, which comprises the following steps:
step 301, user information acquisition:
the information collection method mainly includes three methods: the method comprises the steps that a first online acquisition mode is adopted, and user intention information is obtained according to browsing search records of a user on a webpage and an APP; the second is an offline acquisition mode, which acquires the intention information of the user through the carrying, signing and communication of a specially-assigned person; the third is a mode of electric marketing, which is used for determining the intention of the user through communication with the user by telephone; constructing a sample model of the information acquired by all channels, storing the sample model in a database and facilitating retrieval; enriching collected information channels, including on-line APP recording, off-line communication with watching and selling modes of a user; a large amount of user information is collected, an integral sample model is generated, and the attributes of large objects such as high-quality parcel prices of high-quality long-rent apartments are covered.
Step 302, recall:
after a user searches for information of a long rental apartment near 'XX prefecture', firstly, user requirements are acquired through a sample database: recall from user profile: mining relevant information of a long-rented apartment near the apartment through a Hive interest business district, screening the information through content and LBS, and storing the information into a Redis recall list index base;
collecting and carefully selecting the acquired data information, then carrying out multilayer rough sequencing on the long-rent apartment scenes, carrying out LBS and value matching strategies aiming at the matching of the strong intentions of the clients and the key attributes of the target objects, and carrying out matching strategies on the secondary requirements of the users, the secondary house state and the district characteristics of the house to obtain a rough sequencing result; a new matching strategy is formulated for the conditions of on-line and off-line watching, signing and the like, and a multi-level sample weight depiction is carried out on the sequencing model, so that the complexity of the existing sequencing model is reduced; meanwhile, scenes such as regions, house states and the like are deeply subdivided according to the high sparsity of samples and the large-amount attribute of transaction, and the quality of the recommended apartment is improved.
Step 303, sorting:
for recalled house source data rough sorting information, after comprehensively acquiring historical behaviors of users and house-state plot business circle images and performing characteristic engineering measurement and calculation processing on a Spark distributed framework, respectively performing corresponding training on online data and offline data to construct a sorting model, performing weight portrayal on sample data, performing deep division on scenes such as regions, house states and the like according to the high sparsity of data information and the large-volume attribute of transaction, giving accurate sorting scoring of house source information, and finally recommending high-quality long-renting apartment resources to the users;
step 304, intervention and adjustment:
after modeling based on the specific LBS of the long rental apartment, the system inserts some specific house sources into specific positions in a recommendation list according to customized requirements of a company such as a house-leaving inclined policy, an operation activity policy, advertisement promotion and the like, and recommends more selection spaces for users; and if the recommendation result is insufficient, the system automatically recommends the completion of the related hot spot competitive house resources. And optimizing an intervention adjustment layer, modeling based on a specific LBS mode of the long-renting apartment, and designing a characteristic recommendation engine of the long-renting apartment with large commodities with sparsely distributed plots.
It is understood that the recall method of the present invention may be based on only one of the user behavior information and the image information, or may be based on both of the above two methods. The embodiment only shows that the recall is performed according to the user portrait, and the recall mode is not specifically required and can be set according to the actual situation.
According to the embodiments, the application has the following beneficial effects:
1. the method for constructing the recommendation engine according to the user behaviors and the user portrait, provided by the invention, has the advantages that the frequency threshold of user behavior operation is set, and the accuracy and the efficiency of user behavior information can be improved.
2. According to the method for constructing the recommendation engine according to the user behaviors and the user portrait, the relationship between the first recommended house source and the second house source is limited, so that the income of a long-rented apartment can be improved.
3. The method for constructing the recommendation engine according to the user behaviors and the user portrait provided by the invention has rich information acquisition channels, including online APP recording, offline watching and communication and electric marketing modes of the user; a large amount of user information is collected, an integral sample model is generated, and the bulk object attributes such as high-quality parcel prices of high-quality long-rent apartments are covered.
4. The method for constructing the recommendation engine according to the user behaviors and the user portrait optimizes the personalized recommendation mode, and extracts apartments with high recommendation matching degree from the sample model library through comprehensively measuring and calculating the behaviors and the portrait of the user.
5. According to the method for constructing the recommendation engine according to the user behaviors and the user portrait, the recommendation sequencing model is optimized, a new matching strategy is formulated for the conditions of online and offline watching, signing and the like, the sequencing model is subjected to multilevel sample weight portrayal, and the complexity of the existing sequencing model is reduced; meanwhile, scenes such as regions, house states and the like are deeply subdivided according to the high sparsity of samples and the large-amount attribute of transaction, and the quality of the recommended apartment is improved.
6. The method for constructing the recommendation engine according to the user behaviors and the user portrait optimizes an intervention adjustment layer, models based on a specific LBS (location based service) mode of the long-rented apartment, and designs the characteristic recommendation engine of the long-rented apartment with bulk commodities with sparsely distributed landmass.
While the present invention has been described in detail with reference to the drawings and examples, it is to be understood that the foregoing examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention. The scope of the invention is defined by the appended claims.
Claims (7)
1. A method for building a recommendation engine based on user behavior and user profiles, comprising the steps of:
collecting user information and house source information;
the user information comprises user behavior information, the type of the user behavior information in a preset period and the operation frequency of the user behavior corresponding to each type are counted, a frequency threshold value is preset, and when the operation frequency of the user behavior is greater than the frequency threshold value corresponding to each type, the user behavior information comprises effective user behavior information; the types of the user behavior information comprise user browsing webpage information and user renting duration;
the user information further includes user representation information, the user representation information including user consumption information and user preference information; the house source information is divided into real-time user behavior information and historical user behavior information according to the effective user behavior information; the real-time user behavior information is examined and checked, and the real-time user behavior information enters an online database in real time according to a distributed stream processing open source framework; generating a log from the historical user behavior information, and inputting the log into an offline data warehouse in batches through a Spark frame calculation engine; generating a first recall data source for the online database and the offline data repository;
the house source information is recalled according to the user portrait information, and a second recall data source is obtained by screening through an optical disc arranged in an offline data warehouse and LBS (location based service);
carrying out price matching and section feature matching on the first recall data source and the second recall data source according to LBS service to obtain corresponding matching degrees, and sequentially sequencing from high to low according to the matching degrees to generate a first recommended data source;
performing feature engineering processing on the first recommended data source through a Spark distributed framework, and grading to obtain a second recommended data source;
training the second recommended data source to obtain a ranking model;
inputting the house source information into the sequencing model to obtain the grading sequencing condition of the house source information;
setting a scoring threshold value, wherein the house source with the score of the house source information larger than the scoring threshold value is a first recommended house source;
meanwhile, generating a second recommended house source according to the house owner demand;
and constructing a recommendation engine according to the first recommended house source and the second recommended house source.
2. The method of claim 1, wherein the user information is collected online, offline, and telemarketing.
3. The method of claim 1, wherein the number of first recommended living quarters is greater than the number of second recommended living quarters.
4. The method of claim 3, wherein the ratio of the number of first recommended living quarters to the number of second recommended living quarters is 5: 1.
5. The method of claim 1, wherein the feature engineering process comprises subjecting the first recommended data source to aggregation, refinement, screening and completion in sequence to obtain the second recommended data source.
6. The method of claim 1, wherein said inputting said source information into said ranking model and obtaining a ranking score of said source information comprises: and carrying out weight depiction on the house source information, and carrying out deep division on the scenes such as regions, house states and the like according to the sparsity and the transaction attribute of the house source information to obtain the grading and sequencing conditions of the house source information.
7. The method of claim 1, wherein the recommendation engine further comprises a hotspot source.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911416982.1A CN111159561A (en) | 2019-12-31 | 2019-12-31 | Method for constructing recommendation engine according to user behaviors and user portrait |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911416982.1A CN111159561A (en) | 2019-12-31 | 2019-12-31 | Method for constructing recommendation engine according to user behaviors and user portrait |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111159561A true CN111159561A (en) | 2020-05-15 |
Family
ID=70560235
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911416982.1A Withdrawn CN111159561A (en) | 2019-12-31 | 2019-12-31 | Method for constructing recommendation engine according to user behaviors and user portrait |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111159561A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113034241A (en) * | 2021-04-23 | 2021-06-25 | 中国平安人寿保险股份有限公司 | Product information recommendation method and computer equipment |
| CN113190702A (en) * | 2021-05-08 | 2021-07-30 | 北京百度网讯科技有限公司 | Method and apparatus for generating information |
| CN113918740A (en) * | 2021-11-24 | 2022-01-11 | 北京达佳互联信息技术有限公司 | Event information processing method and device, electronic equipment and storage medium |
| CN114466250A (en) * | 2020-11-09 | 2022-05-10 | 江苏华软智能信息科技有限公司 | Video recommendation method |
| CN114915844A (en) * | 2021-11-08 | 2022-08-16 | 海看网络科技(山东)股份有限公司 | Method for realizing real-time intelligent recommendation on IPTV |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107766580A (en) * | 2017-11-20 | 2018-03-06 | 北京奇虎科技有限公司 | The method for pushing and device of message |
| CN108288179A (en) * | 2018-01-25 | 2018-07-17 | 链家网（北京）科技有限公司 | A kind of user preference source of houses computational methods and system |
| CN109255126A (en) * | 2018-09-10 | 2019-01-22 | 百度在线网络技术（北京）有限公司 | Article recommended method and device |
| CN109377329A (en) * | 2018-12-25 | 2019-02-22 | 北京时光荏苒科技有限公司 | A kind of source of houses recommended method, device, storage medium and electronic equipment |
| CN109658192A (en) * | 2018-12-20 | 2019-04-19 | 重庆锐云科技有限公司 | A kind of source of houses recommended method and server |
| CN109783726A (en) * | 2018-12-21 | 2019-05-21 | 深圳市戴德梁行土地房地产评估有限公司 | A kind of real estate big data computation processing method and its system |
-
2019
- 2019-12-31 CN CN201911416982.1A patent/CN111159561A/en not_active Withdrawn
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107766580A (en) * | 2017-11-20 | 2018-03-06 | 北京奇虎科技有限公司 | The method for pushing and device of message |
| CN108288179A (en) * | 2018-01-25 | 2018-07-17 | 链家网（北京）科技有限公司 | A kind of user preference source of houses computational methods and system |
| CN109255126A (en) * | 2018-09-10 | 2019-01-22 | 百度在线网络技术（北京）有限公司 | Article recommended method and device |
| CN109658192A (en) * | 2018-12-20 | 2019-04-19 | 重庆锐云科技有限公司 | A kind of source of houses recommended method and server |
| CN109783726A (en) * | 2018-12-21 | 2019-05-21 | 深圳市戴德梁行土地房地产评估有限公司 | A kind of real estate big data computation processing method and its system |
| CN109377329A (en) * | 2018-12-25 | 2019-02-22 | 北京时光荏苒科技有限公司 | A kind of source of houses recommended method, device, storage medium and electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 张海华: "基于大数据和机器学习的大学生选课推荐模型研究" * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114466250A (en) * | 2020-11-09 | 2022-05-10 | 江苏华软智能信息科技有限公司 | Video recommendation method |
| CN113034241A (en) * | 2021-04-23 | 2021-06-25 | 中国平安人寿保险股份有限公司 | Product information recommendation method and computer equipment |
| CN113034241B (en) * | 2021-04-23 | 2024-06-11 | 中国平安人寿保险股份有限公司 | Product information recommendation method and computer equipment |
| CN113190702A (en) * | 2021-05-08 | 2021-07-30 | 北京百度网讯科技有限公司 | Method and apparatus for generating information |
| CN113190702B (en) * | 2021-05-08 | 2023-07-25 | 北京百度网讯科技有限公司 | Method and device for generating information |
| CN114915844A (en) * | 2021-11-08 | 2022-08-16 | 海看网络科技(山东)股份有限公司 | Method for realizing real-time intelligent recommendation on IPTV |
| CN113918740A (en) * | 2021-11-24 | 2022-01-11 | 北京达佳互联信息技术有限公司 | Event information processing method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111159561A (en) | Method for constructing recommendation engine according to user behaviors and user portrait | |
| CN110442790B (en) | Method, device, server and storage medium for recommending multimedia data | |
| US10572565B2 (en) | User behavior models based on source domain | |
| CN109033408B (en) | Information pushing method and device, computer readable storage medium and electronic equipment | |
| CN110019616B (en) | POI (Point of interest) situation acquisition method and equipment, storage medium and server thereof | |
| CN111143689A (en) | Method for constructing recommendation engine according to user requirements and user portrait | |
| KR20170093713A (en) | Method and device for mobile searching based on artificial intelligence | |
| CN111159341B (en) | Information recommendation method and device based on user investment and financial management preference | |
| CN111159559A (en) | Method for constructing recommendation engine according to user requirements and user behaviors | |
| CN112632405A (en) | Recommendation method, device, equipment and storage medium | |
| CN111429161B (en) | Feature extraction method, feature extraction device, storage medium and electronic equipment | |
| CN112995690B (en) | Live content category identification method, device, electronic equipment and readable storage medium | |
| CN109033286B (en) | Data statistical method and device | |
| CN114186130A (en) | Big data-based sports information recommendation method | |
| US10346856B1 (en) | Personality aggregation and web browsing | |
| CN108304570B (en) | Processing method and display method of search results, server and client | |
| CN113343149A (en) | Agent-based mobile terminal social media propagation effect evaluation method, system and application | |
| US20160196626A1 (en) | Jury research system | |
| CN115115407A (en) | Intelligent recommendation method and system for urban integrated intelligent service platform | |
| KR20020020584A (en) | Internet survey system and method and media for storing program source thereof | |
| CN111143688B (en) | Evaluation method and system based on mobile news client | |
| CN111161034A (en) | Method for constructing recommendation engine based on LBS (location based service) renting scene | |
| CN112905892A (en) | Big data processing method and big data server applied to user portrait mining | |
| CN118093678B (en) | Method and device for measuring information exposure rate in internet media | |
| CN113343046A (en) | Intelligent search sequencing system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200515 |