CN109934596A - A kind of network food and drink businessman's over range operation judgment method - Google Patents
A kind of network food and drink businessman's over range operation judgment method Download PDFInfo
- Publication number
- CN109934596A CN109934596A CN201910064111.1A CN201910064111A CN109934596A CN 109934596 A CN109934596 A CN 109934596A CN 201910064111 A CN201910064111 A CN 201910064111A CN 109934596 A CN109934596 A CN 109934596A
- Authority
- CN
- China
- Prior art keywords
- food
- businessman
- classification
- over range
- network
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses a kind of network food and drink businessman's over range and manages judgment method, comprising steps of crawl merchant data from major network food and drink platform using crawler technology, in the database by the food business licence of businessman and food name storage on sale;The business scope data of food business licence are identified by optical character recognition methods, and in the database by the data obtained storage;It is random to extract the food name on sale stored in database and manually marked;It is trained on training set using the deep neural network based on continuous bag of words, constructs the food classification judgment models based on food name;Carry out type judgement using the food classification judgment models each food on sale to new businessman, by being compared with the business scope on businessman's food business licence, and then determine businessman whether over range operation.The present invention can effectively judge dining food title classification, so judge network food and drink businessman whether over range operation, greatly improve supervisory efficiency.
Description
Technical field
The invention belongs to technical field of network management, manage judgement side more particularly to a kind of network food and drink businessman's over range
Method.
Background technique
Food-safety problem is related to the existence and health of people, is related to national economy.In the present of network prosperity
It, although network food and drink is provided convenience to people, since businessman is large number of, information flow is frequent, manually checks quotient
Family whether violation heavy workload, take time and effort.
In network food and drink supervision, whether over range operation is to investigate the important indicator of food safety to businessman.According to China
Relevant regulations, network food and drink businessman must manage in the range of allowing on " the food business licence " that it is handled, Bu Nengchao
Range operation.Check whether over range operation mostly uses network food and drink businessman manually checks one by one now, not only heavy workload, and
And cause investigation timeliness slow since supervisor is limited.
Summary of the invention
To solve the above-mentioned problems, the invention proposes a kind of network food and drink businessman's over range to manage judgment method, can
Effectively judge dining food title classification, improve the accuracy of food name classification, and then judges whether network food and drink businessman surpasses
Range operation, greatly improves supervisory efficiency.
In order to achieve the above objectives, the technical solution adopted by the present invention is that: a kind of network food and drink businessman's over range manages judgement
Method, comprising steps of
S100 crawls merchant data from major network food and drink platform using crawler technology, by the food operation permission of businessman
Card and corresponding food name storage on sale are in the database;Food business licence is identified by optical character recognition methods
Business scope data, and in the database by the data obtained storage;
S200, random to extract food name on sale in database and carry out classification annotation, composing training collection;By each food
One kind that title is labeled as in eight based foods of national regulation (make and sell, cold food based food is made and sold, eats based food raw by warm-served food based food
It makes and sell, pastry food is made and sold, makes drink by oneself makes and sell, prepackaged food sale, food in bulk sale, other types of food are made and sold);
S300 is trained on training set using the deep neural network based on continuous bag of words, and building is based on food
The food classification judgment models that the name of an article claims;
S400 carries out type identification using the food classification judgment models each food on sale to new businessman, by with quotient
Family food business licence on business scope data be compared, and then judge businessman whether over range operation.
Further, the generation step of the continuous bag of words includes: to obtain a large amount of food name from database
Claim, is shown as vocabulary by training to limit the real vector of dimension K.
Further, the continuous bag of words construct term vector using Word2Vec, the Word2Vec passes through
Skip-gram model realization.
Further, the training objective of the skip-gram model is that study term vector indicates, pass through the term vector
Indicate predictable context of institute's predicate in sentence;
Given training word w1,w2,...,wT, the training objective of skip-gram model is that study term vector indicates, i.e., it is maximum
Change average log-likelihood:
Wherein, k is the size of trained window;
In skip-gram model, each word w and two vector uwAnd vwIt is associated, the vector uwIndicate word, it is described
Vector vwIndicate context;
In correctly predicted given word wjUnder conditions of, give training word wiProbability use softmax model are as follows:
Wherein, v indicates vocabulary quantity.
Further, being trained on training set using the deep neural network based on continuous bag of words, construct
Food classification judgment models based on food name, the food classification judgment models include:
Input layer: trained Word2Vec term vector is used, each food name is converted into vector, the vector
[X0 X1 X2 ... Xi ... Xn] dimension be 200;
Hidden layer: including two hidden layers Layer1 and Layer2, dimension is 300 and 64 respectively;It is learned by the way that training is automatic
Feature is practised, is introduced by sigmode activation primitive non-linear;
Original output layer: dimension 7;
Softmax output layer: the corresponding probability of each classification is exported by softmax output layer, takes maximum probability value pair
The classification answered is classification as a result, final output dimension is 7.
Further, exporting the corresponding probability of each classification, calculation formula by the softmax output layer are as follows:
Further, in step s 200, food name manually to be marked to the label of its generic, is instructed
Practice collection, verifying collection and test set, and is stored in cloud server;
In step S300, the training set input food classification judgment models of tape label are trained, adjusting and optimizing ginseng
Number drops to critical value until verifying collects the loss function in food classification judgment models, food classification is verified on test set
Judgment models simultaneously guarantee classification accuracy, final curing food classification judgment models.Food classification judgment models are carried out continuous
Update optimization, improve food classification judgment models calculate accuracy rate.
Further, in the step S400, judge new businessman whether over range operation, comprising steps of
S401 carries out classification judgements using food classification judgment models food on sale all to businessman, will differentiate result with
Business scope on businessman's food business licence is compared, and if there is the classification exceeded, then counts over range food number;
S402, according to the actual situation, by threshold decision, if the sold food of the businessman not food in its business scope
Number is more than upper limit value, then determines the businessman for over range operation.
Using the technical program the utility model has the advantages that
The present invention is to guarantee network dining food safety and realize effectively supervision, a large amount of by crawling network food and drink businessman
Information, analysis modeling, and big data and artificial intelligence technology are combined, the whether super model of food and drink businessman can be efficiently checked in the short time
Operation is enclosed, realizes quickly detection.It can effectively classify to dining food title, improve the accuracy rate of food classification, in turn
Judge network food and drink businessman whether over range operation, greatly improve supervisory efficiency.
Detailed description of the invention
Fig. 1 is the flow diagram that a kind of network food and drink businessman's over range of the invention manages judgment method;
Fig. 2 is the structural schematic diagram of food classification judgment models in the embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention is made into one with reference to the accompanying drawing
Step illustrates.
In the present embodiment, shown in Figure 1, the invention proposes a kind of network food and drink businessman's over range to manage judgement side
Method, comprising steps of
S100 crawls merchant data from major network food and drink platform using crawler technology, by the food operation permission of businessman
Card and corresponding food name storage on sale are in the database;Food business licence is identified by optical character recognition methods
Business scope data, and in the database by the data obtained storage;
S200, random to extract food name on sale in database and carry out classification annotation, composing training collection;
S300 is trained on training set using the deep neural network based on continuous bag of words, and building is based on food
The food classification judgment models that the name of an article claims;
S400 carries out type identification using the food classification judgment models each food on sale to new businessman, by with quotient
Family food business licence on business scope data be compared, and then judge businessman whether over range operation.
As the prioritization scheme of above-described embodiment, the generation step of the continuous bag of words includes: to obtain from database
A large amount of food name is taken, is shown as vocabulary by training to limit the real vector of dimension K.
The continuous bag of words construct term vector using Word2Vec, and the Word2Vec passes through skip-gram model
It realizes.
The training objective of the skip-gram model is that study term vector indicates, indicates predictable by the term vector
Context of institute's predicate in sentence;
Given training word w1,w2,...,wT, the training objective of skip-gram model is that study term vector indicates are as follows:
Wherein, k is the size of trained window;
In skip-gram model, each word w and two vector uwAnd vwIt is associated, the vector uwIndicate word, it is described
Vector vwIndicate context;
In correctly predicted given word wjUnder conditions of, give training word wiProbability use softmax model are as follows:
Wherein v indicates vocabulary quantity.
As the prioritization scheme of above-described embodiment, as shown in Fig. 2, utilizing the deep neural network based on continuous bag of words
It is trained on training set, constructs the food classification judgment models based on food name, the food classification judgment models packet
It includes:
Input layer: trained Word2Vec term vector is used, each food name is converted into vector, the vector
[X0 X1 X2 ... Xi ... Xn] dimension be 200;
Hidden layer: including two hidden layers Layer1 and Layer2, dimension is 300 and 64 respectively;It is learned by the way that training is automatic
Feature is practised, is introduced by sigmode activation primitive non-linear;
Original output layer: dimension 7;
Softmax output layer: the corresponding probability of each classification is exported by softmax output layer, takes maximum probability value pair
The classification answered is classification as a result, final output dimension is 7.
Further, exporting the corresponding probability of each classification, calculation formula by the softmax output layer are as follows:
As the prioritization scheme of above-described embodiment, in step s 200, food name is manually marked to its affiliated class
Other label obtains training set, verifying collection and test set, and is stored in cloud server;
In step S300, the training set input food classification judgment models of tape label are trained, adjusting and optimizing ginseng
Number drops to critical value until verifying collects the loss function in food classification judgment models, food classification is verified on test set
Judgment models simultaneously guarantee classification accuracy, final curing food classification judgment models.Food classification judgment models are carried out continuous
Update optimization, improve food classification judgment models calculate accuracy-.
As the prioritization scheme of above-described embodiment, in the step S400, judge new businessman whether over range operation, step
Suddenly include:
S401 carries out classification judgements using food classification judgment models food on sale all to businessman, will differentiate result with
Business scope on businessman's food business licence is compared, and if there is the classification exceeded, then counts over range food number;
S402, according to the actual situation, by threshold decision, if the sold food of the businessman not food in its business scope
Number is more than upper limit value, then determines the businessman for over range operation.
It is verified, 1000 network food and drink businessmans are detected, algorithm identification there are 342 over range operation, through artificial
Wherein 331 certain over range operation, accuracy rate reach 96.78%, greatly improve supervisory efficiency for verifying.
The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the above embodiments and description only describe this
The principle of invention, without departing from the spirit and scope of the present invention, various changes and improvements may be made to the invention, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appended claims and its
Equivalent thereof.
Claims (8)
1. a kind of network food and drink businessman's over range manages judgment method, which is characterized in that comprising steps of
S100 crawls merchant data from major network food and drink platform using crawler technology, by the food business licence of businessman and
Corresponding food name storage on sale is in the database;Model is managed by optical character recognition methods identification food business licence
Data are enclosed, and in the database by the data obtained storage;
S200, random to extract food name on sale in database and carry out classification annotation, composing training collection;
S300 is trained on training set using the deep neural network based on continuous bag of words, and building is based on food name
The food classification judgment models of title;
S400 carries out type identification using the food classification judgment models each food on sale to new businessman, by eating with businessman
Business scope data on product business licence are compared, so determine businessman whether over range operation.
2. a kind of network food and drink businessman's over range according to claim 1 manages judgment method, which is characterized in that the company
The generation step of continuous bag of words includes: that a large amount of food name is obtained from database, is shown as limiting by vocabulary by training
The real vector of dimension K.
3. a kind of network food and drink businessman's over range according to claim 2 manages judgment method, which is characterized in that the company
Continuous bag of words construct term vector using Word2Vec, and the Word2Vec passes through skip-gram model realization.
4. a kind of network food and drink businessman's over range according to claim 3 manages judgment method, which is characterized in that described
The training objective of skip-gram model is that study term vector indicates, indicates prediction institute's predicate in sentence by the term vector
Context;
Given training word w1,w2,...,wT, the training objective of skip-gram model is that study term vector indicates, that is, is maximized flat
Equal log-likelihood:
Wherein, k is the size of trained window;
In skip-gram model, each word w and two vector uwAnd vwIt is associated, the vector uwIndicate word, the vector vw
Indicate context;
In correctly predicted given word wjUnder conditions of, give training word wiProbability use softmax model are as follows:
Wherein, v indicates vocabulary quantity.
5. a kind of network food and drink businessman's over range according to claim 4 manages judgment method, which is characterized in that utilize base
It is trained on training set in the deep neural network of continuous bag of words, constructs the food classification judgement based on food name
Model, the food classification judgment models include:
Input layer: trained Word2Vec term vector is used, each food name is converted into vector, the vector [X0
X1 X2 ... Xi ... Xn] dimension be 200;
Hidden layer: including two hidden layers Layer1 and Layer2, dimension is 300 and 64 respectively;It is special by the automatic study of training
Sign is introduced non-linear by sigmode activation primitive;
Original output layer: dimension 7;
Softmax output layer: the corresponding probability of each classification is exported by softmax output layer, takes maximum probability value corresponding
Classification is classification as a result, final output dimension is 7.
6. a kind of network food and drink businessman's over range according to claim 5 manages judgment method, which is characterized in that pass through institute
It states softmax output layer and exports the corresponding probability of each classification, calculation formula are as follows:
7. a kind of any network food and drink businessman's over range manages judgment method in -6 according to claim 1, feature exists
In, in step s 200, food name is manually marked to the label of its generic, obtain training set, verifying collection and survey
Examination collection, and be stored in cloud server;
In step S300, the training set input food classification judgment models of tape label are trained, adjusting and optimizing parameter, directly
Critical value is dropped to loss function of the verifying collection in food classification judgment models, food classification is verified on test set and judges mould
Type simultaneously guarantees classification accuracy, final curing food classification judgment models.
8. a kind of network food and drink businessman's over range according to claim 7 manages judgment method, which is characterized in that described
In step S400, judge new businessman whether over range operation, step includes:
S401 carries out classification judgement using food classification judgment models food on sale all to businessman, will differentiate result and businessman
Business scope on food business licence is compared, and if there is the classification exceeded, then counts over range food number;
S402, according to the actual situation, by threshold decision, if food number of the sold food of businessman not in its business scope
More than upper limit value, then determine the businessman for over range operation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910064111.1A CN109934596A (en) | 2019-01-23 | 2019-01-23 | A kind of network food and drink businessman's over range operation judgment method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910064111.1A CN109934596A (en) | 2019-01-23 | 2019-01-23 | A kind of network food and drink businessman's over range operation judgment method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109934596A true CN109934596A (en) | 2019-06-25 |
Family
ID=66985148
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910064111.1A Pending CN109934596A (en) | 2019-01-23 | 2019-01-23 | A kind of network food and drink businessman's over range operation judgment method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109934596A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111563806A (en) * | 2020-07-20 | 2020-08-21 | 平安国际智慧城市科技股份有限公司 | Method, device, medium and electronic equipment for identifying merchant compliance in network platform |
CN114638299A (en) * | 2022-03-15 | 2022-06-17 | 平安国际智慧城市科技股份有限公司 | Intelligent zombie license identification method, device, equipment and storage medium |
CN116579786A (en) * | 2023-05-06 | 2023-08-11 | 全拓科技(杭州)股份有限公司 | Data cleaning method and system applied to big data analysis |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102663025A (en) * | 2012-03-22 | 2012-09-12 | 浙江盘石信息技术有限公司 | Illegal online commodity detection method |
US20160170982A1 (en) * | 2014-12-16 | 2016-06-16 | Yahoo! Inc. | Method and System for Joint Representations of Related Concepts |
CN107609807A (en) * | 2017-11-08 | 2018-01-19 | 厦门美亚商鼎信息科技有限公司 | A kind of network food and drink Risk Identification Method and system |
CN107832306A (en) * | 2017-11-28 | 2018-03-23 | 武汉大学 | A kind of similar entities method for digging based on Doc2vec |
CN108614855A (en) * | 2018-03-19 | 2018-10-02 | 众安信息技术服务有限公司 | A kind of rumour recognition methods |
CN108717636A (en) * | 2018-03-19 | 2018-10-30 | 杭州祐全科技发展有限公司 | A kind of network ordering intelligent supervision method |
-
2019
- 2019-01-23 CN CN201910064111.1A patent/CN109934596A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102663025A (en) * | 2012-03-22 | 2012-09-12 | 浙江盘石信息技术有限公司 | Illegal online commodity detection method |
US20160170982A1 (en) * | 2014-12-16 | 2016-06-16 | Yahoo! Inc. | Method and System for Joint Representations of Related Concepts |
CN107609807A (en) * | 2017-11-08 | 2018-01-19 | 厦门美亚商鼎信息科技有限公司 | A kind of network food and drink Risk Identification Method and system |
CN107832306A (en) * | 2017-11-28 | 2018-03-23 | 武汉大学 | A kind of similar entities method for digging based on Doc2vec |
CN108614855A (en) * | 2018-03-19 | 2018-10-02 | 众安信息技术服务有限公司 | A kind of rumour recognition methods |
CN108717636A (en) * | 2018-03-19 | 2018-10-30 | 杭州祐全科技发展有限公司 | A kind of network ordering intelligent supervision method |
Non-Patent Citations (2)
Title |
---|
曹祺著: "《情报学视域下的数据研究 理论、原理与方法》", 31 October 2018, 武汉大学出版社 * |
顾振华编: "《食品药品安全监管工作指南》", 30 January 2017, 上海科学技术出版社 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111563806A (en) * | 2020-07-20 | 2020-08-21 | 平安国际智慧城市科技股份有限公司 | Method, device, medium and electronic equipment for identifying merchant compliance in network platform |
CN114638299A (en) * | 2022-03-15 | 2022-06-17 | 平安国际智慧城市科技股份有限公司 | Intelligent zombie license identification method, device, equipment and storage medium |
CN116579786A (en) * | 2023-05-06 | 2023-08-11 | 全拓科技(杭州)股份有限公司 | Data cleaning method and system applied to big data analysis |
CN116579786B (en) * | 2023-05-06 | 2023-11-14 | 全拓科技(杭州)股份有限公司 | Data cleaning method and system applied to big data analysis |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106611169B (en) | A kind of dangerous driving behavior real-time detection method based on deep learning | |
CN109934596A (en) | A kind of network food and drink businessman's over range operation judgment method | |
CN107918921A (en) | Criminal case court verdict measure and system | |
CN109300111A (en) | A kind of chromosome recognition methods based on deep learning | |
CN106570525A (en) | Method for evaluating online commodity assessment quality based on Bayesian network | |
CN106372072A (en) | Location-based recognition method for user relations in mobile social network | |
CN107851097A (en) | Data analysis system, data analysing method, data analysis program and storage medium | |
CN107846392A (en) | A kind of intrusion detection algorithm based on improvement coorinated training ADBN | |
CN107515873A (en) | A kind of junk information recognition methods and equipment | |
CN107180070A (en) | A kind of risk information is classified, recognized and method for early warning and system automatically | |
CN107784312A (en) | Machine learning model training method and device | |
CN103092975A (en) | Detection and filter method of network community garbage information based on topic consensus coverage rate | |
CN105654196A (en) | Adaptive load prediction selection method based on electric power big data | |
CN106952190A (en) | False source of houses typing Activity recognition and early warning system | |
CN107122432A (en) | CSR analysis method, device and system | |
CN110569904A (en) | method for constructing machine learning model and computer-readable storage medium | |
CN108304479A (en) | A kind of fast density cluster double-layer network recommendation method based on graph structure filtering | |
CN109214444B (en) | Game anti-addiction determination system and method based on twin neural network and GMM | |
Zhang | Food safety risk intelligence early warning based on support vector machine | |
CN109214464A (en) | A kind of doubtful stealing customer identification device and recognition methods based on big data | |
CN109409541A (en) | The method for realizing abandoned car battery reverse logistic feasibility assessment | |
CN114764682B (en) | Rice safety risk assessment method based on multi-machine learning algorithm fusion | |
CN104376024B (en) | A kind of document similarity detection method based on seed words | |
CN107992613A (en) | A kind of Text Mining Technology protection of consumers' rights index analysis method based on machine learning | |
He et al. | [Retracted] Visualization and Analysis of Mapping Knowledge Domain of Heterogeneous Traffic Flow |
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: 20190625 |
|
RJ01 | Rejection of invention patent application after publication |