CN105844302A - Depth-learning-based method for automatically calculating commodity trend indexes - Google Patents

Depth-learning-based method for automatically calculating commodity trend indexes Download PDF

Info

Publication number
CN105844302A
CN105844302A CN201610211098.4A CN201610211098A CN105844302A CN 105844302 A CN105844302 A CN 105844302A CN 201610211098 A CN201610211098 A CN 201610211098A CN 105844302 A CN105844302 A CN 105844302A
Authority
CN
China
Prior art keywords
commodity
value
class
vector
calculate
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.)
Pending
Application number
CN201610211098.4A
Other languages
Chinese (zh)
Inventor
牟川
王海强
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 Xinyili Culture Communication Co ltd
Original Assignee
Nanjing Xinyili Culture Communication 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 Xinyili Culture Communication Co ltd filed Critical Nanjing Xinyili Culture Communication Co ltd
Priority to CN201610211098.4A priority Critical patent/CN105844302A/en
Publication of CN105844302A publication Critical patent/CN105844302A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24137—Distances to cluster centroïds
    • 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]
    • G06Q30/0623—Item investigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Finance (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Accounting & Taxation (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Probability & Statistics with Applications (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a depth-learning-based method for automatically calculating commodity trend indexes. The method comprises four steps that a step 1, characteristic extraction of a commodity picture is carried out to acquire vector characteristics of the picture content; a step 2, according to the commodity vector characteristic values acquired in the step 1, clustering is carried out by employing a K-means algorithm to calculate a class center vector; a step 3, statistics of the historical data of each class of the commodity is carried out, commodity values are converted into modes of a centesimal system by utilizing a formula, the arithmetic average value of the commodity index of each class is calculated according to the center class vector acquired in the step 2, and commodity class indexes are obtained; a step 4, a distance from the commodity to each class center and the weight coefficient are calculated, an Euclidean distance from the center class vector value acquired in the step 2 and a random characteristic value of the commodity and a weight vector value are calculated through a formula; and a step 5, the weight vector W acquired in the step 4 and the trend index of each class acquired in the step 3 are final trend indexes of the commodity.

Description

The method automatically calculating commodity trend index based on degree of depth study
Technical field
The present invention relates to a kind of method automatically calculating commodity trend index based on degree of depth study, especially electronics business Business field, convenience goods trend intelligent identifying system quickly calculates trend index.
Background technology
The Internet has become as the main path of people's shopping now, and especially young crowd is more willing to select to purchase on the net How thing, be quickly found out up-to-date trend commodity according to personal like when of shopping on the web, becomes young more and more The demand that people is maximum.At present, the commodity pictorial style arrived for visual experience, it is converted into the concrete technology phoenix quantifying trend index Hair rare thing, can only search by the way of artificial top set, need to take more time and energy on buying experience.
Summary of the invention
In order to solve above-mentioned technical problem, the invention provides a kind of commodity trend that automatically calculates based on degree of depth study and refer to The method of number, the method comprises the steps:
Step 1: all trend commodity selling the network platform, uses CNN degree of deep learning model that commodity picture is carried out feature Extract, original image data is converted into the vector characteristics of representative picture content.
Step 2: commodity are clustered according to picture feature, the commodity picture characteristic vector drawn according to step 1, use K- Means algorithm clusters, and marks off N number of different classification (N 1), and the characteristic vector of each picture belongs to concrete In one classification, institute's directed quantity of identical category takes arithmetic mean of instantaneous value, i.e. draws the center vector of the category.
Step 3: calculate the classification index of commodity；
Step 4: calculate the commodity distance to all kinds of centers；
Step 5: statistical weight, draws weight vectors W by step 4, is multiplied by the trend index of each class that step 3 draws, Draw the trend index that commodity are final.
Further, the calculating merchandise classification index described in step 3, concrete steps have: carry out history sales volume to of all categories Statistics: add up the average click volume of all commodity and average sales volume, maximum click volume is defined as: Xmax, minimum click volume definition For: Xmin, utilize formula: commodity value is converted into hundred-mark system by (X-Xmin)/(Xmax-Xmin) * 100, draws commodity of all categories Trend index.
Further, the calculating merchandise classification index described in step 4, concrete steps have: calculate commodity to each class center Distance R, and the value of all 1/R is classified as and be 1 value, this value is weight coefficient；The class center that will draw in step 2 Vector value and the arbitrary eigenvalue of commodity, by formula r=sqrt (∑ (yi-xi) ^2) calculate between the two European away from From, calculate weight vectors value by formula wi=1/ri/ (∑ 1/ri).
By technique scheme it can be seen that compared with prior art, the advantage of the application is: inside commodity picture Concrete content is according to the operation history at the commercial family of electricity, complete quantization trend index.Consider the feature of commodity itself, and The hobby of user, and the classification of numerous commodity, trend style of changing over cannot quantify completely, it is impossible to going through of indexation History.
Accompanying drawing explanation
Specific embodiment mode below in conjunction with the accompanying drawings is described in further detail.
Fig. 1 is the method schematic diagram automatically calculating commodity trend index that the present invention learns based on the degree of depth；
Fig. 2 is the image, semantic feature of 1024 dimensions of shirt；
Fig. 3 is the characteristic vector of shirt.
Detailed description of the invention
In order to more fully understand the technology contents of the present invention, below in conjunction with specific embodiment, technical scheme is entered One step introduction and explanation, but it is not limited to this.
See accompanying drawing 1, step 1: use CNN degree of deep learning model that commodity picture carries out feature extraction, setting commodity 1: X1, x2 ... x1024}, each commodity extract the image, semantic feature of 1024 dimensions according to image content, as a example by shirt, such as figure Shown in 2；
Step 2: commodity are clustered according to the feature in Fig. 2, the commodity picture characteristic vector drawn according to step 1
Clustering with K-means algorithm, mark off N number of different classification (N 1), the characteristic vector of each picture belongs to In a concrete classification, institute's directed quantity of identical category takes arithmetic mean of instantaneous value, i.e. draws the center vector of the category.Such as figure Shown in 3:
Class1:{x1,x2,...x1024},Class2:{x1,x2,...x1024}；
Step 3: to the history sales statistics that carries out of all categories: add up average click volume and average sales volume, the maximum point of all commodity The amount of hitting is defined as: Xmax, and minimum click volume is defined as: Xmin, utilizes formula (X-Xmin)/(Xmax-Xmin) * 100 by commodity Value is converted into hundred-mark system, draws the trend index of commodity of all categories.
Step 4: calculate the commodity distance to all kinds of centers: first calculate commodity distance R to each class center,
And the value of all 1/R is classified as and be 1 value, this value is weight coefficient；The class center vector that will draw in step 2 The eigenvalue that value is arbitrary with commodity, calculates Euclidean distance between the two by formula r=sqrt (∑ (yi-xi) ^2), logical Cross formula wi=1/ri/ (∑ 1/ri) and calculate weight vectors value.
As shown in the table: for new commodity or the commodity do not sold in company, extract its feature X={x1, x2, ... x1024}, calculate the calculating weight of X and each classification
Step 5: statistical weight, draws weight vectors W by step 4, is multiplied by the trend index of each class that step 3 draws, Draw the trend index that commodity are final.

Claims (3)

1. the method automatically calculating commodity trend index based on degree of depth study, it is characterised in that: the method
Comprise the steps:
Step 1: use CNN degree of deep learning model that commodity picture is carried out feature extraction, original image data is converted into generation The vector characteristics of table image content；
Step 2: use K-means algorithm to cluster the picture feature vector of step 1, mark off N number of different classification (N 1), the characteristic vector of each picture belongs in a concrete classification, and institute's directed quantity of identical category takes arithmetic average Value, draws the center vector of the category；
Step 3: calculate the classification index of commodity；
Step 4: calculate the commodity distance to all kinds of centers；
Step 5: statistical weight, draws weight vectors W by step 4, is multiplied by the trend index of each class that step 3 draws, draws The trend index that commodity are final.
The method automatically calculating commodity trend index based on degree of depth study the most according to claim 1, it is characterised in that: step Calculating merchandise classification index described in 3, concrete steps have: carry out history sales statistics to of all categories, add up the flat of all commodity All click volume and average sales volumes, maximum click volume is defined as: Xmax, and minimum click volume is defined as: Xmin, utilizes formula (X- Commodity value is converted into hundred-mark system by Xmin)/(Xmax-Xmin) * 100, draws the classification index of commodity.
The method automatically calculating commodity trend index based on degree of depth study the most according to claim 1, it is characterised in that: step Calculating commodity described in 4 are to the distance at all kinds of centers, and concrete steps have:
Calculate commodity to distance R at each class center, and the value of all 1/R is classified as and be 1 value, this value is weight coefficient； By the class center vector value drawn in step 2 and the arbitrary eigenvalue of commodity, by formula r=sqrt (∑ (yi-xi) ^2 ) calculate Euclidean distance between the two, calculate weight vectors value by formula wi=1/ri/ (∑ 1/ri).
CN201610211098.4A 2016-04-07 2016-04-07 Depth-learning-based method for automatically calculating commodity trend indexes Pending CN105844302A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610211098.4A CN105844302A (en) 2016-04-07 2016-04-07 Depth-learning-based method for automatically calculating commodity trend indexes

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610211098.4A CN105844302A (en) 2016-04-07 2016-04-07 Depth-learning-based method for automatically calculating commodity trend indexes

Publications (1)

Publication Number Publication Date
CN105844302A true CN105844302A (en) 2016-08-10

Family

ID=56598014

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610211098.4A Pending CN105844302A (en) 2016-04-07 2016-04-07 Depth-learning-based method for automatically calculating commodity trend indexes

Country Status (1)

Country Link
CN (1) CN105844302A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108022116A (en) * 2016-11-01 2018-05-11 北京京东尚科信息技术有限公司 To the method, system and terminal device of user modeling
CN108268898A (en) * 2018-01-19 2018-07-10 大象慧云信息技术有限公司 A kind of electronic invoice user clustering method based on K-Means
CN108470285A (en) * 2017-02-23 2018-08-31 北京京东尚科信息技术有限公司 Method, apparatus, electronic equipment and storage medium for obtaining user data information
CN108665300A (en) * 2017-04-01 2018-10-16 北京京东尚科信息技术有限公司 A kind of method, apparatus, equipment and storage medium that data are provided
CN110020695A (en) * 2019-04-19 2019-07-16 杭州电子科技大学 K-means non-uniform quantizing algorithm for filter bank multi-carrier modulation optical communication system
CN114004605A (en) * 2021-12-31 2022-02-01 北京中科闻歌科技股份有限公司 Invoice over-limit application approval method, device, equipment and medium

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060093185A1 (en) * 2004-11-04 2006-05-04 Fuji Xerox Co., Ltd. Moving object recognition apparatus
CN102841946A (en) * 2012-08-24 2012-12-26 北京国政通科技有限公司 Commodity data retrieval sequencing and commodity recommendation method and system
CN103092861A (en) * 2011-11-02 2013-05-08 阿里巴巴集团控股有限公司 Method and system for selecting commodity representative picture
CN103345645A (en) * 2013-06-27 2013-10-09 复旦大学 Commodity image category forecasting method based on online shopping platform
CN103605765A (en) * 2013-11-26 2014-02-26 电子科技大学 Mass image retrieval system based on cluster compactness
CN104142946A (en) * 2013-05-08 2014-11-12 阿里巴巴集团控股有限公司 Method and system for aggregating and searching service objects of same type
CN104732249A (en) * 2015-03-25 2015-06-24 武汉大学 Deep learning image classification method based on popular learning and chaotic particle swarms

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060093185A1 (en) * 2004-11-04 2006-05-04 Fuji Xerox Co., Ltd. Moving object recognition apparatus
CN103092861A (en) * 2011-11-02 2013-05-08 阿里巴巴集团控股有限公司 Method and system for selecting commodity representative picture
CN102841946A (en) * 2012-08-24 2012-12-26 北京国政通科技有限公司 Commodity data retrieval sequencing and commodity recommendation method and system
CN104142946A (en) * 2013-05-08 2014-11-12 阿里巴巴集团控股有限公司 Method and system for aggregating and searching service objects of same type
CN103345645A (en) * 2013-06-27 2013-10-09 复旦大学 Commodity image category forecasting method based on online shopping platform
CN103605765A (en) * 2013-11-26 2014-02-26 电子科技大学 Mass image retrieval system based on cluster compactness
CN104732249A (en) * 2015-03-25 2015-06-24 武汉大学 Deep learning image classification method based on popular learning and chaotic particle swarms

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
图书情报工作杂志社: "《电子商务研究与实践进展》", 31 October 2013, 海军出版社 *
贾世杰: "基于内容的商品图像分类方法研究", 《中国博士学位论文全文数据库》 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108022116A (en) * 2016-11-01 2018-05-11 北京京东尚科信息技术有限公司 To the method, system and terminal device of user modeling
CN108022116B (en) * 2016-11-01 2021-06-29 北京京东尚科信息技术有限公司 Method, system and terminal equipment for modeling user
CN108470285A (en) * 2017-02-23 2018-08-31 北京京东尚科信息技术有限公司 Method, apparatus, electronic equipment and storage medium for obtaining user data information
CN108470285B (en) * 2017-02-23 2021-11-12 北京京东尚科信息技术有限公司 Method, device, electronic equipment and storage medium for acquiring user data information
CN108665300A (en) * 2017-04-01 2018-10-16 北京京东尚科信息技术有限公司 A kind of method, apparatus, equipment and storage medium that data are provided
CN108268898A (en) * 2018-01-19 2018-07-10 大象慧云信息技术有限公司 A kind of electronic invoice user clustering method based on K-Means
CN110020695A (en) * 2019-04-19 2019-07-16 杭州电子科技大学 K-means non-uniform quantizing algorithm for filter bank multi-carrier modulation optical communication system
CN114004605A (en) * 2021-12-31 2022-02-01 北京中科闻歌科技股份有限公司 Invoice over-limit application approval method, device, equipment and medium

Similar Documents

Publication Publication Date Title
CN105844302A (en) Depth-learning-based method for automatically calculating commodity trend indexes
CN106651546B (en) Electronic commerce information recommendation method oriented to smart community
CN110674407B (en) Hybrid recommendation method based on graph convolution neural network
Strehl et al. Relationship-based clustering and visualization for high-dimensional data mining
CN103093376B (en) Clustering collaborative filtering recommendation system based on singular value decomposition algorithm
CN108665323B (en) Integration method for financial product recommendation system
CN108334592B (en) Personalized recommendation method based on combination of content and collaborative filtering
Ba et al. Clustering collaborative filtering recommendation system based on SVD algorithm
CN109947987B (en) Cross collaborative filtering recommendation method
CN106203313B (en) A kind of the clothes classification and recommended method of image content-based
CN105426528A (en) Retrieving and ordering method and system for commodity data
CN102968506A (en) Personalized collaborative filtering recommendation method based on extension characteristic vectors
CN106897914A (en) A kind of Method of Commodity Recommendation and system based on topic model
CN104298787A (en) Individual recommendation method and device based on fusion strategy
CN109903138A (en) A kind of individual commodity recommendation method
CN102521420B (en) Socialized filtering method on basis of preference model
CN104915440B (en) A kind of commodity rearrangement and system
CN112102029B (en) Knowledge graph-based long-tail recommendation calculation method
CN108287916A (en) A kind of resource recommendation method
CN108389113B (en) Collaborative filtering recommendation method and system
CN106886559A (en) The collaborative filtering method of good friend's feature and similar users feature is incorporated simultaneously
US20160171365A1 (en) Consumer preferences forecasting and trends finding
CN102722578A (en) Unsupervised cluster characteristic selection method based on Laplace regularization
Zhao et al. Men also do laundry: Multi-attribute bias amplification
CN111597428A (en) Recommendation method for splicing user and article with q-separation k sparsity

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20160810