CN109121093A - A kind of user's portrait construction method and system based on passive type WiFi and depth cluster - Google Patents
A kind of user's portrait construction method and system based on passive type WiFi and depth cluster Download PDFInfo
- Publication number
- CN109121093A CN109121093A CN201810762690.2A CN201810762690A CN109121093A CN 109121093 A CN109121093 A CN 109121093A CN 201810762690 A CN201810762690 A CN 201810762690A CN 109121093 A CN109121093 A CN 109121093A
- Authority
- CN
- China
- Prior art keywords
- cluster
- portrait
- user
- wifi
- passive type
- 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
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/029—Location-based management or tracking services
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/80—Services using short range communication, e.g. near-field communication [NFC], radio-frequency identification [RFID] or low energy communication
Abstract
It draws a portrait construction method and system the present invention relates to a kind of user based on passive type WiFi and depth cluster, which includes: multiple WiFi detectors, one for the server of data processing training, the database of storing data.This method comprises: clustering is carried out in conjunction with deep learning, to depict typical portrait by passive type WiFi technology scanning cell phone.User's portrait construction method and system proposed by the present invention based on passive type WiFi and depth cluster, it is easy to spread without carrying special installation or additional downloads terminal using the universal equipment with WiFi function;Noninductive detection is carried out using passive type WiFi, wide coverage and easy to operate;The prediction of stream of people's distribution is carried out using deep learning K-means model, realizes the accurate quarter of user's portrait.
Description
Technical field
The present invention relates to passive type WiFi location technology, machine learning method, deep learning method, especially one kind to be based on
The user's portrait construction method and system of passive type WiFi and depth cluster.
Background technique
User's portrait is the labeling user gone out according to informations such as user's social property, living habit and consumer behaviors
Model.User's portrait core value is to understand user, the potential demand for guessing user, the positioning crowd characteristic of fining, digging
Dig potential user group.Since 21 century, with the continuous development and innovation of smart phone and Internet technology, mobile interchange
The product that net is combined as the two, development is swift and violent in recent years.As people use the continuous increasing of the duration and frequency of smart phone
Add, user behavior data exponentially increases, but the time of user increasingly tends to fragmentation, and the information of each dimension is also richer
Richness, if can portray " user's portrait ", so that the information about user's various aspects being collected into, may include population category
Property, hobby, shopping preferences, social attribute etc., many features of user can be described well, product personnel are unfolded
Targetedly deisgn product all plays the role of operation personnel's development precision marketing personalized recommendation vital.But
How to carry out complete precisely capture with effectively analysis to the feature of " user's portrait " is always the pain spot institute for portraying " user's portrait "
?.
The research of current phase user portrait focuses primarily upon behavioral study of the mankind under internet environment, and user's portrait is
Product under internet development, while also having pushed the continuous development and growth of Internet market.The number of user's portrait at this stage
It relies primarily on networks congestion control according to source to be captured and portrayed, the network behavior of user can no doubt draw a portrait for user and provide
A large amount of data sample, but under the overall background of life diversification, there is significant limitations for the network behavior of user, make
The portrait of user must be portrayed inaccurate.
Along with the development of indoor positioning technologies, it is based especially on the appearance and maturation of passive type WiFi technology, passes through road
By scanning cell phone, server is transmitted data to, server carries out positions calculations, to obtain location information.Meanwhile in recent years
Deep learning starts the highest attention by researcher and businessperson as a kind of new machine learning method.If can rely on
Passive type WiFi technology captures the actual life behavior of user, according to a large amount of location data, in conjunction with deep learning K-
Crowd is carried out clustering by means model, is carried out complete effectively portray to the portrait of user and is possibly realized, depicts some allusion quotations
The portrait of type, allow user portrait it is more accurate apparent.
Summary of the invention
The purpose of the present invention is to provide it is a kind of based on passive type WiFi and depth cluster user draw a portrait construction method and
System, to overcome defect existing in the prior art.
To achieve the above object, the technical scheme is that a kind of user based on passive type WiFi and depth cluster
It draws a portrait building system, comprising: multiple WiFi detectors, one for the server of data processing training, storing data
Database.
Further, a kind of user's portrait construction method based on passive type WiFi and depth cluster is also provided, indoors
Several WiFi detectors are placed in place appropriate location, and are divided into N number of region according to the space layout of indoor spaces;It is logical
It crosses WiFi detector and obtains stream of people's data of smart phone entrained by individual in N number of region, and pass through wired or wireless net
Network is transmitted and is stored in the relevant database of server;The server carries out stream of people's data by deep learning model
Clustering, and classify to the action trail of user, construct the typical portrait of user.
In an embodiment of the present invention, stream of people's data include: (X, Y) position data of MAC Address, individual;Wherein,
Each MAC Address respectively corresponds an individual, that is, count how many a MAC Address be equivalent to how many individual;Individual (X,
Y) position data is obtained by the RSSI value of detection cell phone apparatus.
In an embodiment of the present invention, the WiFi detector is communicated using UDP and is obtained by sending and receiving data server
(X, Y) position data of MAC Address, individual, and it is uploaded to the server.
In an embodiment of the present invention, the deep learning model uses K-means Clustering Model.
In an embodiment of the present invention, the K-means Clustering Model completes clustering as follows:
Step S1: k object of random selection, each object initially represent the average value or center of a cluster;For
Remaining each object arrives the distance at each cluster center according to it, they is given apart from the smallest cluster center, is then counted again
Calculate the average value of each cluster;This process is repeated, until clustering criteria function convergence;
Step S2: for the poly- of the data acquisition system D comprising the corresponding user behavior track of n MAC Address and initialization
Class number k randomly chooses k object as initial cluster center from data acquisition system D;
Step S3: according to the central value of cluster, n object in data acquisition system is all given and meets pre-determined distance range
Cluster;
Step S4: the central value of cluster is updated, that is, recalculates the central value of each cluster;
Step S5: calculation criterion function;
Step S6: it if criterion function meets preset threshold, exits;Otherwise, return step S2;
Step S7: by the above-mentioned steps that move in circles, the corresponding user behavior track of each MAC Address is subjected to cluster point
Analysis, obtains classification results, to obtain typical user's portrait.
In an embodiment of the present invention, the criterion function uses two subcenter of global error function or front and back
Error change;
The global error function are as follows:
Errors of centration changes twice for the front and back are as follows:
Wherein, E Representative errors, k represent k cluster centre, and i represents ith cluster, xjRepresent j-th of sample, SiIt represents
The sample set of ith cluster, uiRepresent i-th of mean vector, uibIt is a preceding center of gravity of group i, uiaIt is the latter of group i
Secondary center of gravity.
Compared to the prior art, the invention has the following advantages: it is proposed by the present invention a kind of based on passive type WiFi
The user's portrait construction method and system clustered with depth, (including but not limited to using the universal equipment with WiFi function
Mobile phone), deep learning is utilized further according to the location data of passive type WiFi in conjunction with passive type WiFi technology acquisition location data
The prediction of K-means model progress stream of people's distribution.K-means model in deep learning can be handled location data
Obtained user behavior track is grouped, to obtain more accurate user's portrait.
Detailed description of the invention
Fig. 1 is the system construction drawing of user's portrait building system based on passive type WiFi and depth cluster in the present invention.
Fig. 2 is user's portrait schematic diagram of construction method in the present invention based on passive type WiFi and depth cluster.
Fig. 3 is the clustering flow chart based on K-means model in the present invention.
Specific embodiment
With reference to the accompanying drawing, technical solution of the present invention is specifically described.
The present invention provides a kind of user's portrait building system based on passive type WiFi and depth cluster, as shown in Figure 1, packet
Multiple WiFi detectors are included, the WiFi detector includes but is not limited to router;One service for data processing training
Device, the database of a storing data, database are not limited to any relevant database.
Further, a kind of user's portrait construction method based on passive type WiFi and depth cluster is also provided, such as Fig. 2 institute
Show, several WiFi detectors are placed in place appropriate location indoors, including but not limited to router;And according to indoor spaces
Space layout is divided into N number of region.Then, it is gone to obtain in this N number of region entrained by individual by WiFi detector
(X, Y) position data of stream of people's data of smart phone, mainly MAC Address, individual, wherein MAC Address is with smart phone
Unique corresponding, each MAC Address respectively corresponds an individual, that is, count how many a MAC Address be equivalent to how many
Body;(X, Y) position data of individual can be through but not limited to the RSSI value of detection cell phone apparatus, and combines and be not limited to any one
The location algorithm of kind is calculated.The data such as finally obtained MAC Address, (X, Y) position data are passed through into wired or wireless net
Network is stored in server and is not limited in any relevant database, by deep learning K-means Clustering Model to clothes
Position data in device of being engaged in carries out clustering, and classifies to the action trail of user, draws to depict the typical of user
Picture.
Further, in the present embodiment, surrounding WiFi equipment information is acquired using passive type WiFi technology, so as to
The location information of personnel is navigated to by mobile phone, and non-user equipment is filtered using particle filter algorithm.
Further, in the present embodiment, it is obtained by the sending and receiving data server in WiFi detector using DP communication
MAC Address, position data are uploaded to data processing training server.
Further, in the present embodiment, data processing training server utilizes the MAC Address and location information obtained,
Action trail is portrayed for the corresponding individual of each MAC.
Further, in the present embodiment, as shown in figure 3, K-means Clustering Model is realized in accordance with the following steps:
Step S1: before position data is sent into K-means model training, algorithm randomly chooses k object first, each
Object initially represents the average value or center of a cluster.For remaining each object, each cluster center is arrived according to it
Distance gives them apart from the smallest cluster center, then recalculates the average value of each cluster.This process is repeated, until
Clustering criteria function convergence.Criterion function generally uses two ways: first, global error function, corresponding (3-1);Second,
Errors of centration changes twice for front and back, corresponding (3-2).
Wherein E Representative errors, k represent k cluster centre, and i represents ith cluster, xjRepresent j-th of sample, SiRepresent
The sample set of i cluster, uiRepresent i-th of mean vector, uibIt is a preceding center of gravity of group i, uiaIt is the rear primary of group i
Center of gravity.
Step S2: for the poly- of the data acquisition system D comprising the corresponding user behavior track of n MAC Address and initialization
Class number k randomly chooses k object as initial cluster center from data acquisition system D;
Step S3: according to the central value of cluster, n object in data acquisition system is all given the cluster of most " similar ";At this
In embodiment, " similar " according to judging apart from length, " cluster " here is to be grouped;
Step S4: the central value of cluster is updated, that is, recalculates the central value of each cluster;
Step S5: calculation criterion function;
Step S6: it is exited if criterion function meets threshold value, otherwise return step S2;
Step S7: by the above-mentioned steps that move in circles, the corresponding user behavior track of each MAC Address is subjected to cluster point
Analysis, obtains classification results, to obtain typical user's portrait.
The above are preferred embodiments of the present invention, all any changes made according to the technical solution of the present invention, and generated function is made
When with range without departing from technical solution of the present invention, all belong to the scope of protection of the present invention.
Claims (7)
- A kind of building system 1. user based on passive type WiFi and depth cluster draws a portrait characterized by comprising multiple WiFi Detector, one for the server of data processing training, the database of storing data.
- The construction method 2. a kind of user based on passive type WiFi and depth cluster draws a portrait, which is characterized in that place is suitable indoors When several WiFi detectors are placed in position, and N number of region is divided into according to the space layout of indoor spaces;Pass through WiFi Detector obtains stream of people's data of smart phone entrained by individual in N number of region, and is transmitted by wired or wireless network And it is stored in the relevant database of server;The server carries out cluster point to stream of people's data by deep learning model Analysis, and classify to the action trail of user, construct the typical portrait of user.
- The construction method 3. a kind of user based on passive type WiFi and depth cluster according to claim 2 draws a portrait, it is special Sign is that stream of people's data include: (X, Y) position data of MAC Address, individual;Wherein, each MAC Address respectively corresponds An individual, that is, count how many a MAC Address be equivalent to how many individual;(X, Y) position data of individual passes through detection The RSSI value of cell phone apparatus obtains.
- The construction method 4. a kind of user based on passive type WiFi and depth cluster according to claim 3 draws a portrait, it is special Sign is that the WiFi detector communicates the position (X, Y) for obtaining MAC Address, individual using UDP by sending and receiving data server Data are set, and are uploaded to the server.
- The construction method 5. a kind of user based on passive type WiFi and depth cluster according to claim 3 draws a portrait, it is special Sign is that the deep learning model uses K-means Clustering Model.
- The construction method 6. a kind of user based on passive type WiFi and depth cluster according to claim 5 draws a portrait, it is special Sign is that the K-means Clustering Model completes clustering as follows:Step S1: k object of random selection, each object initially represent the average value or center of a cluster;For residue Each object, the distance at each cluster center is arrived according to it, they are given apart from the smallest cluster center, is then recalculated every The average value of a cluster;This process is repeated, until clustering criteria function convergence;Step S2: for the data acquisition system D comprising the corresponding user behavior track of n MAC Address and the cluster numbers of initialization Mesh k randomly chooses k object as initial cluster center from data acquisition system D;Step S3: according to the central value of cluster, n object in data acquisition system is all given the cluster for meeting pre-determined distance range;Step S4: the central value of cluster is updated, that is, recalculates the central value of each cluster;Step S5: calculation criterion function;Step S6: it if criterion function meets preset threshold, exits;Otherwise, return step S2;Step S7: by the above-mentioned steps that move in circles, carrying out clustering for the corresponding user behavior track of each MAC Address, Classification results are obtained, to obtain typical user's portrait.
- The construction method 7. a kind of user based on passive type WiFi and depth cluster according to claim 6 draws a portrait, it is special Sign is that errors of centration changes the criterion function twice using global error function or front and back;The global error function are as follows:Errors of centration changes twice for the front and back are as follows:Wherein, E Representative errors, k represent k cluster centre, and i represents ith cluster, xjRepresent j-th of sample, SiIt represents i-th The sample set of cluster, uiRepresent i-th of mean vector, uibIt is a preceding center of gravity of group i, uiaIt is the rear primary weight of group i The heart.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810762690.2A CN109121093A (en) | 2018-07-12 | 2018-07-12 | A kind of user's portrait construction method and system based on passive type WiFi and depth cluster |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810762690.2A CN109121093A (en) | 2018-07-12 | 2018-07-12 | A kind of user's portrait construction method and system based on passive type WiFi and depth cluster |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109121093A true CN109121093A (en) | 2019-01-01 |
Family
ID=64862670
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810762690.2A Pending CN109121093A (en) | 2018-07-12 | 2018-07-12 | A kind of user's portrait construction method and system based on passive type WiFi and depth cluster |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109121093A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110012355A (en) * | 2019-01-18 | 2019-07-12 | 阿里巴巴集团控股有限公司 | Method and apparatus for choosing media content |
CN110532309A (en) * | 2019-07-15 | 2019-12-03 | 浙江工业大学 | A kind of generation method of Library User's portrait system |
CN113596898A (en) * | 2020-04-30 | 2021-11-02 | 华为技术有限公司 | Method, device, equipment, storage medium and system for determining terminal portrait |
CN114722081A (en) * | 2022-06-09 | 2022-07-08 | 杭银消费金融股份有限公司 | Streaming data time sequence transmission method and system based on transfer library mode |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170032384A1 (en) * | 2015-07-29 | 2017-02-02 | Geofeedia, Inc. | System and Method for Analyzing Social Media Users Based on User Content Posted from Monitored Locations |
CN106910333A (en) * | 2015-12-23 | 2017-06-30 | 神州数码信息系统有限公司 | A kind of citizen's protecting track privacy technology based on directional statistics |
CN107801202A (en) * | 2017-10-31 | 2018-03-13 | 广东思域信息科技有限公司 | A kind of user's portrait method based on WiFi accesses |
CN108260087A (en) * | 2018-01-31 | 2018-07-06 | 福州大学 | A kind of indoor stream of people's distribution forecasting method based on WIIFI long short-term memories |
-
2018
- 2018-07-12 CN CN201810762690.2A patent/CN109121093A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170032384A1 (en) * | 2015-07-29 | 2017-02-02 | Geofeedia, Inc. | System and Method for Analyzing Social Media Users Based on User Content Posted from Monitored Locations |
CN106910333A (en) * | 2015-12-23 | 2017-06-30 | 神州数码信息系统有限公司 | A kind of citizen's protecting track privacy technology based on directional statistics |
CN107801202A (en) * | 2017-10-31 | 2018-03-13 | 广东思域信息科技有限公司 | A kind of user's portrait method based on WiFi accesses |
CN108260087A (en) * | 2018-01-31 | 2018-07-06 | 福州大学 | A kind of indoor stream of people's distribution forecasting method based on WIIFI long short-term memories |
Non-Patent Citations (2)
Title |
---|
李冰,王悦 ,刘永祥: "大数据环境下基于K_means的用户画像与智能推荐的应用", 《现代计算机》 * |
蒲杰方,卢荧玲: "基于聚类算法和神经网络的客户分类模型构建", 《软件》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110012355A (en) * | 2019-01-18 | 2019-07-12 | 阿里巴巴集团控股有限公司 | Method and apparatus for choosing media content |
CN110532309A (en) * | 2019-07-15 | 2019-12-03 | 浙江工业大学 | A kind of generation method of Library User's portrait system |
CN113596898A (en) * | 2020-04-30 | 2021-11-02 | 华为技术有限公司 | Method, device, equipment, storage medium and system for determining terminal portrait |
WO2021218228A1 (en) * | 2020-04-30 | 2021-11-04 | 华为技术有限公司 | Method for determining terminal profile, apparatus, device, storage medium and system |
EP4132069A4 (en) * | 2020-04-30 | 2023-09-27 | Huawei Technologies Co., Ltd. | Method for determining terminal profile, apparatus, device, storage medium and system |
CN114722081A (en) * | 2022-06-09 | 2022-07-08 | 杭银消费金融股份有限公司 | Streaming data time sequence transmission method and system based on transfer library mode |
CN114722081B (en) * | 2022-06-09 | 2022-09-02 | 杭银消费金融股份有限公司 | Streaming data time sequence transmission method and system based on transfer library mode |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106912015B (en) | Personnel trip chain identification method based on mobile network data | |
CN109121093A (en) | A kind of user's portrait construction method and system based on passive type WiFi and depth cluster | |
CN107291888B (en) | Machine learning statistical model-based living recommendation system method near living hotel | |
CN104408043B (en) | A kind of information processing method and server | |
CN104123398B (en) | A kind of information-pushing method and device | |
CN105094305B (en) | Identify method, user equipment and the Activity recognition server of user behavior | |
CN108595461B (en) | Interest exploration method, storage medium, electronic device and system | |
CN101990003B (en) | User action monitoring system and method based on IP address attribute | |
CN104348855B (en) | Processing method, mobile terminal and the server of user information | |
EP3512221A1 (en) | Empirical expert determination and question routing system and method | |
CN106506705A (en) | Listener clustering method and device based on location-based service | |
CN105718465A (en) | Geofence generation method and device | |
CN109447150A (en) | A kind of plants ' aesthetics method, apparatus, electronic equipment and storage medium | |
JP2011108245A (en) | Program for generating inference model to determine activity type to user from contextual information | |
CN106790468A (en) | A kind of distributed implementation method for analyzing user's WiFi event trace rules | |
CN110334293A (en) | A kind of facing position social networks has Time Perception position recommended method based on fuzzy clustering | |
Huang et al. | Unsupervised interesting places discovery in location-based social sensing | |
CN109948066A (en) | A kind of point of interest recommended method based on Heterogeneous Information network | |
McKitrick et al. | Collecting, analyzing, and visualizing location-based social media data: review of methods in GIS-social media analysis | |
CN109688532A (en) | A kind of method and device dividing city function region | |
CN106407412A (en) | A friend recommendation method | |
CN110069619A (en) | Source of houses methods of exhibiting, device, equipment and computer readable storage medium | |
CN110874437A (en) | Personalized interest point recommendation method based on multiple interest point pair sorting | |
CN108062366A (en) | Public culture information recommendation system | |
CN106899644A (en) | Processing method of taking pictures and server |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190101 |
|
RJ01 | Rejection of invention patent application after publication |