CN110020113B - Home product prediction method and device based on feature matching - Google Patents

Home product prediction method and device based on feature matching Download PDF

Info

Publication number
CN110020113B
CN110020113B CN201710894238.7A CN201710894238A CN110020113B CN 110020113 B CN110020113 B CN 110020113B CN 201710894238 A CN201710894238 A CN 201710894238A CN 110020113 B CN110020113 B CN 110020113B
Authority
CN
China
Prior art keywords
product data
product
convolution algorithm
subunit
project
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.)
Active
Application number
CN201710894238.7A
Other languages
Chinese (zh)
Other versions
CN110020113A (en
Inventor
包科旻
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Wujie Household Technology Co ltd
Original Assignee
Nanjing Wujie Household Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing Wujie Household Technology Co ltd filed Critical Nanjing Wujie Household Technology Co ltd
Priority to CN201710894238.7A priority Critical patent/CN110020113B/en
Publication of CN110020113A publication Critical patent/CN110020113A/en
Application granted granted Critical
Publication of CN110020113B publication Critical patent/CN110020113B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • 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
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241—Advertisements
    • G06Q30/0251—Targeted advertisements
    • 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
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241—Advertisements
    • G06Q30/0277—Online advertisement
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Marketing (AREA)
  • Economics (AREA)
  • Databases & Information Systems (AREA)
  • General Business, Economics & Management (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a method and a device for predicting household products based on feature matching, wherein the method comprises the following steps: acquiring project requirements and pushing first standby product data to a project set; selecting first product data and adding the first product data to a project set; extracting a first product feature from the first product data using a first convolution algorithm; extracting other product features from other product data in the first backup product data using a second convolution algorithm; matching a second standby product characteristic from other product characteristics according to the first product characteristic, and pushing second product data corresponding to the second standby product characteristic to the project set; and selecting second product data from the second standby product data, adding the second product data into the project set, recommending products to the project set according to the first product characteristics, and judging the product requirements of the previous user and the project requirements according to the products in the project set, so that more appropriate products can be obtained from the recommended products when products with approximately the same project requirements are continued.

Description

Home product prediction method and device based on feature matching
Technical Field
The invention relates to the technical field of project data management, in particular to a household product prediction method and device based on feature matching.
Background
In the field of project management, project management associations define a set of projects as: a set of associated projects that are coordinated to gain revenue and control that cannot be achieved when managing the projects individually. Coordinated management is to gain benefits and control over individual management of items. A collection of items may include related work that is outside the scope of each individual item. An item may or may not belong to any one of the sets of items, but any one of the sets of items must contain items. After entering management informatization, project data is generally managed on the internet by means of a computer. At this time, the user needs to create his/her own project set on the website, and then add a project to the project set or manage the project as a product of the project.
In the existing internet technology, a website often needs to recommend various product information related to a project set to a user for the user to adopt and select, so that the path for the user to search for a required product is shortened, and the user experience is improved. Generally, when a website recommends a product, according to historical operation data of a user on some products, such as product types in a user item set, a relevance algorithm is used to determine the relevance relationship between other products and the added product, so as to recommend product information with strong relevance with the product added by the user to the user.
However, this recommendation method only considers the historical operation data of the user, and does not comprehensively consider the association degree between other products, so the recommendation result is often very inaccurate, and particularly, when the user is a new user, because there is no historical operation data, only the initial association can be performed according to the type initially input by the user, and the product recommended for the user at this time is far from the expectation of the user.
Moreover, the existing correlation algorithm has large consumption of system resources, the calculation of the incidence relation between all products and other products is needed, the processed data volume is large, the speed is slow, especially under the condition of mass users, mass products and mass access data, the processing speed of the data is slow, the resource consumption is more serious, and therefore, the recommended products are difficult to be given in time.
Disclosure of Invention
The technical problem solved by the technical scheme of the invention is how to push product information meeting the requirements of a project set to the project set created by a user under the condition of considering the relevance between other products so as to improve the working efficiency of the user.
In order to achieve the above object, the technical solution of the present invention provides a method for predicting a home product based on feature matching, including:
acquiring project requirements and pushing first standby product data to a project set;
selecting first product data from the first standby product data, and adding the first product data to the project set;
extracting a first product feature from the first product data using a first convolution algorithm;
extracting other product features from the first backup product data other than the first product data using a second convolution algorithm;
matching a second standby product characteristic from other product characteristics according to the first product characteristic, and pushing second product data corresponding to the second standby product characteristic to the project set;
and selecting second product data from the second standby product data, and adding the second product data to the project set.
Further, the obtaining the project requirement and pushing the first standby product data to the project set includes:
selecting a setting style and a use scene corresponding to a project set when the project set is created;
and matching the first standby product data from the database according to the set style and the use scene.
Further, selecting first product data from the first backup product data, and adding the first product data to the project set includes:
marking other product data except the first product data in the first standby product data, and marking the times of pushing the other product data;
the other product data is sorted in descending order to form a marker sequence.
Further, the method also comprises the following steps:
sequentially extracting product features of the other product data within the marker sequence;
training a second convolution algorithm by using the first product data in the item set as a total training set through a first convolution algorithm;
extracting other product features for a previously set amount of the other product data in the sequence of marks using a second convolution algorithm.
Further, the method also comprises the following steps:
inserting second product data into the training content of the first convolution algorithm, wherein the weighting parameters of the elements in the second product data in the first convolution algorithm are gradually decreased according to the sequence of the elements in the marking sequence;
the weighting parameter ratio between the elements and the marking number ratio between the elements are arranged in proportion;
the first convolution algorithm inserted into the training content updates the second convolution algorithm.
In order to solve the above technical problems, the technical solution of the present invention further provides a home product prediction device based on feature matching, including: the acquisition unit is used for acquiring project requirements;
the first pushing unit is used for pushing the first standby product data to the project set;
the first selection unit is used for selecting first product data from the first standby product data;
a first adding unit, configured to add the first product data to the project set;
a first convolution unit for extracting a first product feature from the first product data using a first convolution algorithm;
a second convolution unit for extracting other product features from the other product data except the first product data in the first standby product data by using a second convolution algorithm;
the matching unit is used for matching a second standby product characteristic from other product characteristics according to the first product characteristic;
the second pushing unit is used for pushing second product data corresponding to the second standby product characteristics to the project set;
the second selection unit is used for selecting second product data from the second standby product data;
a second adding unit, configured to add the second product data to the project set.
Further, the obtaining unit further comprises a selecting subunit, and the selecting subunit is used for selecting the setting style and the use scene corresponding to the item set;
the first pushing unit further comprises a matching subunit, and the matching subunit is used for matching the first standby product data from the database according to the set style and the use scene.
Further, the first selecting unit further includes:
the marking subunit is used for marking other product data except the first product data in the first standby product data;
and the sequencing subunit is used for sequencing the other product data in a descending order to form a marking sequence.
Further, the method also comprises the following steps:
a first extraction subunit configured to extract product features of the other product data within the tag sequence in order;
a training subunit, configured to train a second convolution algorithm by using the first product data in the item set as a total training set through a first convolution algorithm;
a second extraction subunit, configured to extract, by using a second convolution algorithm, other product features from a previously set number of the other product data in the marker sequence.
Further, the method also comprises the following steps:
an insertion subunit, configured to insert second product data into the training content of the first convolution algorithm;
a first setting subunit, configured to set the weighting parameters of the elements in the second product data in the first convolution algorithm in a gradually decreasing manner according to the order of the elements in the mark sequence;
and the second setting subunit is used for setting the weighting parameter ratio between the elements and the marking number ratio between the elements in a direct proportion relationship.
An updating subunit, configured to update the second convolution algorithm by using the first convolution algorithm inserted with the training content.
The technical scheme of the invention at least comprises the following beneficial effects: the first product characteristics of the first product data are extracted by using a first convolution algorithm, products are recommended to the project set according to the first product characteristics, the product requirements of the user and the project requirements can be judged according to the products in the project set, and therefore more appropriate products can be obtained from the recommended products when the products with approximately the same project requirements continue; and after the marked products are adopted, the weighting parameters of the adopted and highly marked products in the second convolution algorithm are improved, so that the later recommended products can be distinguished from the previous product requirements to better meet the product requirements of the current user, the accuracy of the recommendation result is improved, the products recommended to the user more closely fit the expectation, and the use experience is improved.
According to the technical scheme, the products with the project requirements conforming to the project set but not adopted are marked, after the marked products are adopted, the training set of the first convolution algorithm training second convolution algorithm is inserted, the weighting parameters of the products are improved, the influence of the marked products on the following recommended products is improved, the recommended products conform to the product requirements of the current user, repeated recommendation is not needed for many times, the products in the marking sequence are directly subjected to feature extraction, and the range of the extracted objects is narrowed. The convolution algorithm reduces the consumption of system resources, does not need to calculate the incidence relation between all products and other products, reduces the processed data volume, and improves the speed of recommending the products, and particularly obviously improves the speed of recommending the products under the conditions of mass users, mass products and mass access data.
Drawings
FIG. 1 is a schematic flow chart of a method according to a first embodiment of the present invention;
fig. 2 is a flowchart illustrating a method S100 according to an embodiment of the present invention;
fig. 3 is a flowchart illustrating a method S200 according to an embodiment of the present invention;
fig. 4 is a flowchart illustrating a method S300 according to an embodiment of the present invention;
fig. 5 is a flowchart illustrating a method S500 according to an embodiment of the present invention;
FIG. 6 is a block diagram of an apparatus according to a second embodiment of the present invention;
FIG. 7 is a block diagram of an obtaining unit according to a second embodiment of the present invention;
FIG. 8 is a block diagram of a second push unit according to a second embodiment of the present invention;
FIG. 9 is a block diagram of a second selection unit according to a second embodiment of the present invention;
FIG. 10 is a block diagram of a second first convolution unit according to an embodiment of the present invention;
FIG. 11 is a block diagram of a second convolution unit according to a second embodiment of the present invention;
fig. 12 is a block diagram of a second push unit according to the second embodiment of the present invention.
Reference numerals: 100. an acquisition unit; 101. selecting a subunit; 200. a first pushing unit; 201. a matching subunit; 300. a first selecting unit; 301. a tagging subunit; 302. a sorting subunit; 400. a first adding unit; 500. a first convolution unit; 501. a first extraction subunit; 502. a training subunit; 600. a second convolution unit; 601. a second extraction subunit; 700. a matching unit; 800. a second pushing unit; 801. an insertion subunit; 802. a first setting subunit; 803. a second setting subunit; 804. updating the subunit; 900. a second selecting unit; 1000. a second adding unit.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the process of implementing the invention, the inventor carries out deep analysis on the prior art, finds that the product information recommended to the user by the website in the prior art does not meet the requirements of the user, and the reason for the inaccuracy is that: when the website recommends products, according to historical operation data of a user on certain products, such as product types in a user project set, the relevance algorithm is used for determining the relevance relation between other products and the added products, and therefore product information with strong relevance with the products added by the user is recommended to the user. However, this recommendation method only considers the historical operation data of the user, and does not comprehensively consider the association degree between other products, and particularly, when the user is a new user, because there is no historical operation data, only the initial association can be performed according to the type initially input by the user, and the product recommended to the user at this time is far from the expectation of the user.
Based on the defects of the prior art, the technical scheme of the invention provides a solution. The technical scheme of the invention.
It should be noted that, in the technical solution of the present invention, the input end and the user end refer to a mobile intelligent device or a non-mobile intelligent device, such as a mobile phone, a tablet computer, a notebook computer or an intelligent television; the server and the server side in the technical scheme of the invention are background servers used in a method for pushing and saving product data and a prediction method based on project content, one or more servers can be arranged, and the servers can include but are not limited to a database server, an application server and a WEB server.
Example one
In combination with the inventive concept of the technical scheme of the present invention, a method for predicting a home product based on feature matching, as shown in fig. 1, comprises the following steps:
step S100: the project requirements are acquired, and first standby product data are pushed to the project set. The project set is a new file created by the user side on the server side, and the first standby product data is offline product set data stored in the database.
As shown in fig. 2, step S100 further includes step S101 and step S102.
Step S101: when creating the project set, the user selects the setting style and the usage scenario corresponding to the project set, and the setting style and the usage scenario are the project requirements in step S100.
Step S102: and matching the first standby product data from the database according to the set style and the use scene. The server matches a plurality of products with the set styles corresponding to the use scenes from the offline product set in the database according to the set styles and the use scenes, and the plurality of products form first standby product data. The server side pushes the first standby product data to the client side.
Step S200: and selecting first product data from the first standby product data, and adding the first product data to the project set. The selection in step S200 is actively completed by the user side, that is, one or more products required are selected from the first standby product data with multiple products as the first product data to be adopted, and the user side adds the first product data to the newly created project set.
As shown in fig. 3, step S200 further includes step S201 and step S202.
Step S201: and marking other product data except the first product data in the first standby product data, and marking the times of pushing the other product data. Each product in the other product data has a tag representing the number of times it was pushed but not accepted. The mark can represent the correlation degree between the product and the project requirement of the user end, the larger the value of the mark, the better the fit between the product and the project requirement, but the other characteristics of the product can not meet the requirement of the user end and are not adopted.
Step S202: other product data is sorted in descending order to form a marker sequence.
Step S300: a first product feature is extracted for the first product data using a first convolution algorithm. The server uses the existing CNN convolutional neural network algorithm in the TensorFlow library as a first convolutional algorithm. The first convolution algorithm traverses each product in the first product data and calculates the demand characteristics of each product to form first product characteristics. The requirement characteristics can be the setting style and the use scene of the product.
As shown in fig. 4, step S300 further includes step S301, step S302, and step S303.
Step S301: product features of other product data within the marker sequence are extracted in order.
And the feature extraction is directly carried out on the products in the marker sequence, so that the range of the extracted object is narrowed. The convolution algorithm reduces the consumption of system resources and improves the speed and accuracy of recommended products.
Step S302: and training a second convolution algorithm by using the first convolution algorithm by taking the first product data in the item set as a total training set. The TensorFlow library inputs all products in the first product data as a total training set into the first convolution algorithm for training to produce a second convolution algorithm. It is worth noting that since the existing CNN convolutional neural network algorithm in the tensrflow library is a common technique in the art, the present invention focuses on inputting all elements in the first product data as a total training set into the CNN convolutional neural network algorithm, i.e., the first convolutional algorithm, to generate the first product features, and the second convolutional algorithm with only parameter changes compared with the first convolutional algorithm, i.e., the present invention focuses on the input object and the output object of the first convolutional algorithm, not the first convolutional algorithm itself, and therefore, the first convolutional algorithm is not described much.
Step S303: a second convolution algorithm is used to extract other product features for a previously set amount of other product data in the marker sequence. The second convolution algorithm traverses the products in each of the other product data and calculates the demand characteristics of each product to form other product characteristics.
And after the marked products are adopted, inserting a training set of a first convolution algorithm training second convolution algorithm, improving the weighting parameters of the products, improving the influence of the marked products on the later recommended products, and enabling the recommended products to better meet the product requirements of the current user without repeated recommendation.
Step S400: a second convolution algorithm is used to extract other product features from the first alternate product data in addition to the first product data.
Step S500: and matching a second standby product characteristic from other product characteristics according to the first product characteristic, and pushing second product data corresponding to the second standby product characteristic to the project set.
As shown in fig. 5, step S500 further includes step S501, step S502, and step S503.
Step S501: and inserting second product data into the training content of the first convolution algorithm, wherein the weighting parameters of the elements in the second product data in the first convolution algorithm are gradually decreased according to the sequence of the elements in the marking sequence.
The larger the numerical value of the mark is, the more times the product is recommended and not adopted is, and the more the product fits the project requirement but does not meet the product requirement of the previous user side, so that the product weighting parameter is increased, the recommended product does not deviate from the project requirement and can meet the product requirement of the current user side, the accuracy of the recommended product is improved, and the working efficiency of the user side is improved.
Step S502: the weighting parameter ratio between the elements is set in proportion to the number of times of marking between the elements.
Step S503: the first convolution algorithm inserted into the training content updates the second convolution algorithm.
After the marked products are adopted, the weighting parameters of the adopted products with high marks in the second convolution algorithm are improved, so that the later recommended products can be distinguished from the previous product requirements to better meet the product requirements of the current user, the accuracy of the recommendation result is improved, the products recommended to the user are more closely matched with the expectation, and the use experience of a user side is improved.
Step S600: and selecting second product data from the second standby product data, and adding the second product data to the project set.
According to the technical scheme, the first product characteristic of the first product data is extracted by using a first convolution algorithm, the product is recommended to the project set according to the first product characteristic, and the product requirement and the project requirement of the previous user can be judged according to the product in the project set, so that a plurality of products which are in accordance with the project requirement and are approximately the same are pushed, and the user side can obtain more appropriate products from the recommended products; and after the marked products are adopted, the weighting parameters of the adopted and highly marked products in the second convolution algorithm are improved, so that the later recommended products can be distinguished from the previous product requirements to better meet the product requirements of the current user, the accuracy of the recommendation result is improved, the products recommended to the user more closely fit the expectation, and the use experience is improved.
And simultaneously, products with project requirements conforming to the project set but not adopted are marked, after the marked products are adopted, a training set of a first convolution algorithm training second convolution algorithm is inserted, the weighting parameters of the products are improved, the influence of the marked products on the later recommended products is improved, the recommended products are enabled to conform to the product requirements of the current user better, repeated recommendation is not needed for many times, the products in the marking sequence are directly subjected to feature extraction, and the range of the extracted objects is narrowed. The convolution algorithm reduces the consumption of system resources, does not need to calculate the incidence relation between all products and other products, reduces the processed data volume, and improves the speed of recommending the products, and particularly obviously improves the speed of recommending the products under the conditions of mass users, mass products and mass access data.
Example two
In combination with the above inventive concept of the technical solution of the present invention, a home product prediction device based on feature matching, as shown in fig. 6, includes:
an acquiring unit 100 is used for acquiring the project requirement. As shown in fig. 7, the obtaining unit 100 further includes a selecting subunit 101, where the selecting subunit 101 is configured to select a setting style and a usage scenario corresponding to the item set.
A first pushing unit 200 for pushing the first standby product data to the set of items. As shown in fig. 8, the first pushing unit 200 further includes a matching subunit 201, and the matching subunit 201 is configured to match the first standby product data from the database according to the set style and the usage scenario.
The first selecting unit 300 is configured to select a first product data from the first spare product data. As shown in fig. 9, the first picking unit 300 further includes a marking subunit 301 and a sorting subunit 302.
A marking subunit 301, configured to mark other product data in the first spare product data except the first product data.
A sorting subunit 302, configured to sort the other product data in descending order to form a tag sequence.
A first adding unit 400 for adding the first product data to the set of items.
A first convolution unit 500 for extracting a first product feature for the first product data using a first convolution algorithm. As shown in fig. 10, the first convolution unit 500 further includes a first extraction subunit 501 and a training subunit 502.
A first extraction subunit 501, configured to extract product features of other product data in the tag sequence in order.
A training subunit 502, configured to train a second convolution algorithm by using the first convolution algorithm with the first product data in the item set as the total training set.
A second convolution unit 600, configured to extract other product features from the other product data in the first spare product data except the first product data by using a second convolution algorithm. As shown in fig. 11, the second convolution unit 600 further includes a second extraction sub-unit 601.
A second extraction subunit 601, configured to extract other product features from a preset number of other product data in the marker sequence by using a second convolution algorithm.
The matching unit 700 is used for matching a second spare product characteristic from other product characteristics according to the first product characteristic.
The second pushing unit 800 is configured to push second product data corresponding to the second standby product characteristic to the project set. As shown in fig. 12, the second pushing unit 800 further includes an inserting sub-unit 801, a first setting sub-unit 802, a second setting sub-unit 803, and an updating sub-unit 804.
An inserting subunit 801 is configured to insert the second product data into the training content of the first convolution algorithm.
A first setting subunit 802, configured to set the weighting parameters of the elements in the second product data in the first convolution algorithm to be gradually decreased according to the order of the elements in the mark sequence.
A second setting subunit 803, configured to set the weighting parameter ratio between the elements and the marking number ratio between the elements in a direct proportional relationship.
An updating subunit 804 is configured to update the second convolution algorithm by using the first convolution algorithm inserted with the training content.
A second selecting unit 900, configured to select second product data from the second standby product data;
a second adding unit 1000, configured to add the second product data to the project set.
The above description is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may occur to those skilled in the art without departing from the principle of the invention, and are considered to be within the scope of the invention.

Claims (6)

1. A household product prediction method based on feature matching is characterized by comprising the following steps:
acquiring project requirements and pushing first standby product data to a project set;
selecting first product data from the first standby product data, and adding the first product data to the project set;
extracting a first product feature from the first product data using a first convolution algorithm;
extracting other product features from the first backup product data other than the first product data using a second convolution algorithm;
matching a second standby product characteristic from other product characteristics according to the first product characteristic, and pushing second product data corresponding to the second standby product characteristic to the project set;
selecting second product data from the second standby product data, and adding the second product data to the project set;
sorting the other product data in descending order to form a tagged sequence;
sequentially extracting product features of the other product data within the marker sequence;
training a second convolution algorithm by using the first product data in the item set as a total training set through a first convolution algorithm;
extracting other product features from a previously set amount of the other product data in the marker sequence using a second convolution algorithm;
inserting second product data into the training content of the first convolution algorithm, wherein the weighting parameters of the elements in the second product data in the first convolution algorithm are gradually decreased according to the sequence of the elements in the marking sequence;
the weighting parameter ratio between the elements and the marking number ratio between the elements are arranged in proportion;
the first convolution algorithm inserted into the training content updates the second convolution algorithm.
2. The method of claim 1, wherein the obtaining the project requirement pushing first backup product data to the set of projects comprises:
selecting a setting style and a use scene corresponding to a project set when the project set is created;
and matching the first standby product data from the database according to the set style and the use scene.
3. The method of claim 1, wherein selecting first product data from the first backup product data, and adding the first product data to the set of items comprises:
marking other product data except the first product data in the first standby product data, and marking the number of times that the other product data is pushed.
4. A household product prediction device based on feature matching is characterized by comprising:
an acquisition unit (100) for acquiring a project requirement;
a first pushing unit (200) for pushing first standby product data to the set of items;
a first selecting unit (300) for selecting a first product data from the first spare product data;
a first adding unit (400) for adding the first product data to the set of items;
a first convolution unit (500) for extracting a first product feature from the first product data using a first convolution algorithm;
a second convolution unit (600) for extracting other product features from the first backup product data other than the first product data using a second convolution algorithm;
a matching unit (700) for matching a second spare product characteristic from the other product characteristics according to the first product characteristic;
a second pushing unit (800) for pushing second product data corresponding to the second standby product characteristic to the project set;
a second selecting unit (900) for selecting second product data from the second standby product data;
a second adding unit (1000) for adding the second product data to the set of items;
a sorting subunit (302) for sorting the other product data in descending order to form a marker sequence;
a first extraction subunit (501) for sequentially extracting product features of the other product data within the marker sequence;
a training subunit (502) for training a second convolution algorithm using a first convolution algorithm with the first product data in the item set as a total training set;
a second extraction subunit (601) for extracting other product features for a previously set number of said other product data in said marker sequence using a second convolution algorithm;
an insertion subunit (801) for inserting second product data into the training content of the first convolution algorithm;
a first setting subunit (802) for setting the weighting parameters of the elements in the second product data in the first convolution algorithm in a gradually decreasing order in the sequence of markers;
a second setting subunit (803) for setting a weighting parameter ratio between the elements and a marking number ratio between the elements in a direct proportional relationship;
an updating subunit (804) for updating the second convolution algorithm by using the first convolution algorithm inserted with training content.
5. The device according to claim 4, wherein the obtaining unit (100) further comprises a selecting subunit (101), and the selecting subunit (101) is used for selecting a setting style and a usage scenario corresponding to the item set;
the first pushing unit (200) further comprises a matching subunit (201), and the matching subunit (201) is used for matching the first standby product data from the database according to the set style and the use scene.
6. The apparatus according to claim 5, wherein the first selection unit (300) further comprises:
a marking subunit (301) for marking other product data than the first product data in the first spare product data.
CN201710894238.7A 2017-09-28 2017-09-28 Home product prediction method and device based on feature matching Active CN110020113B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710894238.7A CN110020113B (en) 2017-09-28 2017-09-28 Home product prediction method and device based on feature matching

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710894238.7A CN110020113B (en) 2017-09-28 2017-09-28 Home product prediction method and device based on feature matching

Publications (2)

Publication Number Publication Date
CN110020113A CN110020113A (en) 2019-07-16
CN110020113B true CN110020113B (en) 2021-04-20

Family

ID=67186431

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710894238.7A Active CN110020113B (en) 2017-09-28 2017-09-28 Home product prediction method and device based on feature matching

Country Status (1)

Country Link
CN (1) CN110020113B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104754413A (en) * 2013-12-30 2015-07-01 北京三星通信技术研究有限公司 Image search based television signal identification and information recommendation method and device
CN104992179A (en) * 2015-06-23 2015-10-21 浙江大学 Fine-grained convolutional neural network-based clothes recommendation method
CN106570192A (en) * 2016-11-18 2017-04-19 广东技术师范学院 Deep learning-based multi-view image retrieval method
CN106933996A (en) * 2017-02-28 2017-07-07 广州大学 A kind of recommendation method of use depth characteristic matching
CN107105031A (en) * 2017-04-20 2017-08-29 北京京东尚科信息技术有限公司 Information-pushing method and device

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104834641B (en) * 2014-02-11 2019-03-15 腾讯科技（北京）有限公司 The processing method and related system of network media information
US20160379229A1 (en) * 2015-06-25 2016-12-29 International Business Machines Corporation Predicting project outcome based on comments
CN105608604A (en) * 2015-12-30 2016-05-25 合一网络技术(北京)有限公司 Continuous calculation method of brand advertisement effectiveness optimization
CN106127525A (en) * 2016-06-27 2016-11-16 浙江大学 A kind of TV shopping Method of Commodity Recommendation based on sorting algorithm
CN106230849B (en) * 2016-08-22 2019-04-19 中国科学院信息工程研究所 A kind of smart machine machine learning safety monitoring system based on user behavior
CN106504064A (en) * 2016-10-25 2017-03-15 清华大学 Clothes classification based on depth convolutional neural networks recommends method and system with collocation
CN106599022B (en) * 2016-11-01 2019-12-10 中山大学 User portrait forming method based on user access data

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104754413A (en) * 2013-12-30 2015-07-01 北京三星通信技术研究有限公司 Image search based television signal identification and information recommendation method and device
CN104992179A (en) * 2015-06-23 2015-10-21 浙江大学 Fine-grained convolutional neural network-based clothes recommendation method
CN106570192A (en) * 2016-11-18 2017-04-19 广东技术师范学院 Deep learning-based multi-view image retrieval method
CN106933996A (en) * 2017-02-28 2017-07-07 广州大学 A kind of recommendation method of use depth characteristic matching
CN107105031A (en) * 2017-04-20 2017-08-29 北京京东尚科信息技术有限公司 Information-pushing method and device

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Deep Structured Scene Parsing by Learning with Image Descriptions;Shengqing Zhai et al.;《2016 IEEE Conference on Computer Vision and Pattern Recognition》;20161212;2276-2283 *

Also Published As

Publication number Publication date
CN110020113A (en) 2019-07-16

Similar Documents

Publication Publication Date Title
CN106651542B (en) Article recommendation method and device
CN108521439B (en) Message pushing method and device
CN108108821B (en) Model training method and device
CN110222880B (en) Service risk determining method, model training method and data processing method
CN106484777B (en) Multimedia data processing method and device
CN105678587B (en) Recommendation feature determination method, information recommendation method and device
CN105335491B (en) Behavior is clicked come to the method and system of user's Recommended Books based on user
CN105160545B (en) Method and device for determining release information style
CN105224554A (en) Search word is recommended to carry out method, system, server and the intelligent terminal searched for
CN105824911A (en) Video recommending method based on LDA user theme model
CN105103176A (en) Information processing device, information processing method, and program
CN110598105B (en) Recommendation method, device, equipment and medium based on probability sampling
CN104239327A (en) Location-based mobile internet user behavior analysis method and device
CN114936301B (en) Intelligent household building material data management method, device, equipment and storage medium
CN111209477A (en) Information recommendation method and device, electronic equipment and storage medium
CN107644047B (en) Label prediction generation method and device
CN106982250B (en) Information pushing method and device
CN112559777A (en) Content item delivery method and device, computer equipment and storage medium
CN105069034A (en) Recommendation information generation method and apparatus
CN110489131B (en) Gray level user selection method and device
CN114238764A (en) Course recommendation method, device and equipment based on recurrent neural network
CN108563648B (en) Data display method and device, storage medium and electronic device
CN112565902B (en) Video recommendation method and device and electronic equipment
CN110020113B (en) Home product prediction method and device based on feature matching
CN111782935A (en) Information recommendation method and device, electronic equipment and storage medium

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
GR01 Patent grant
GR01 Patent grant