CN108830492B - Method for determining spot-check merchants based on big data - Google Patents

Method for determining spot-check merchants based on big data Download PDF

Info

Publication number
CN108830492B
CN108830492B CN201810650948.XA CN201810650948A CN108830492B CN 108830492 B CN108830492 B CN 108830492B CN 201810650948 A CN201810650948 A CN 201810650948A CN 108830492 B CN108830492 B CN 108830492B
Authority
CN
China
Prior art keywords
merchant
merchants
market
inspection
check
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.)
Expired - Fee Related
Application number
CN201810650948.XA
Other languages
Chinese (zh)
Other versions
CN108830492A (en
Inventor
高嵘
丁熠
赵良吉
秦臻
张黎
邓伏虎
赵洋
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.)
Chengdu Boyoi Technology Co ltd
University of Electronic Science and Technology of China
Original Assignee
Chengdu Boyoi Technology Co ltd
University of Electronic Science and Technology of China
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 Chengdu Boyoi Technology Co ltd, University of Electronic Science and Technology of China filed Critical Chengdu Boyoi Technology Co ltd
Priority to CN201810650948.XA priority Critical patent/CN108830492B/en
Publication of CN108830492A publication Critical patent/CN108830492A/en
Application granted granted Critical
Publication of CN108830492B publication Critical patent/CN108830492B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Educational Administration (AREA)
  • Economics (AREA)
  • Tourism & Hospitality (AREA)
  • General Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Physics & Mathematics (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a method for determining a spot check merchant based on big data, which comprises the following steps: s1, classifying all merchants in the market according to the inspection items according to the data in the market supervision platform, and calculating the predicted failure rate F (i) of each merchant; s2, calculating the final inspection priority F of each merchant according to the predicted failure rate of each merchant*(i) (ii) a S3, according to F, a plurality of merchants in the market*(i) In descending order according to F*(i) And automatically generating a merchant list, updating the merchant checking condition after the field check is finished, and storing corresponding data into the market supervision platform. The spot check merchant generation method automatically generates the merchant list required to be subjected to field check on the same day every day according to the event information data which occurs in the recent market in the supervision platform in real time, and effectively solves the problems of inaccuracy and low efficiency in the spot check process of the merchants in the traditional market.

Description

Method for determining spot-check merchants based on big data
Technical Field
The invention belongs to the technical field of Internet big data application, and particularly relates to a method for determining a spot-check merchant based on big data.
Background
Strengthening the safety supervision and management of the marketing quality of the edible agricultural products is an important way for ensuring the quality safety of the edible agricultural products. Therefore, centralized trading markets in various regions begin to establish own practical agricultural product market supervision platforms so as to realize the information acquisition and recording of quality information and sales information of edible agricultural products. Such platforms typically include three types of users: market regulators, sellers (merchants), and buyers (buyers). The market supervisor comprises a market management department and a basic supervision department of government, and can carry out online and offline (on-site) inspection on basic information of sellers and daily purchase and sale information through the platform, and give out inspection results and related correction requirements on the platform, such as a booth name, a mobile phone number, an identity card number, a management certificate, a management field of the market, a booth number and the like, and provide daily purchase and sale information, such as the variety, total weight, unit price, production place, purchase certificate, quality inspection condition and the like of input agricultural products, the variety, unit price, weight, amount of money sold and the like, so as to be inspected by the market supervisor; the buyer refers to a buyer of the product, and the buyer can evaluate and grade the seller corresponding to the purchased product on the platform, and can also complain the seller on the platform. For example, as shown in fig. 1, the platform interface is used when a supervisor performs field inspection, an inspector can fill in according to actual conditions to finally draw a conclusion whether the platform is qualified, and if the platform is not qualified, the platform can automatically generate requirements for what aspects need to be modified according to the inspection conditions, or the inspector can manually input the requirements for modification.
For a market supervisor, the workload of checking all the booths on site every day is too large and the efficiency is low, which may cause a lot of repeated labor, and the traditional spot inspection method is that the supervisor draws part of the booths for inspection in sequence or randomly every day, which has the advantages of simplicity and easy implementation, but this way may cause the supervisor not to know the food safety problem occurring in the recent market in time. Big data is a data set which cannot be captured, managed and processed by a conventional software tool within a certain time range, and is a massive, highly-growing and diversified information asset which can have stronger decision-making power, insight discovery power and process optimization power only by a new processing mode. The real-time updated various data information on the market supervision platform is big data information, so that the data information of the platform is integrated for more efficiently and accurately performing quality spot check on the merchants on the market, and the list of the merchants needing spot check automatically by the supervision party becomes a direction needing further thinking according to the content of the merchants.
Disclosure of Invention
Aiming at the defects in the prior art, the method for determining the spot check merchant based on the big data solves the problems that the spot check process of the traditional market merchant is not accurate enough, the efficiency is low, and the real-time food quality safety cannot be known in time.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that: a method for determining a spot check merchant based on big data comprises the following steps,
s1, classifying all merchants in the market according to the inspection item data in the market supervision platform, and calculating the prediction reject ratio F (i) of each merchant;
wherein i ═ 1,2.. m, m denotes the total number of merchants in the market;
s2, calculating the final inspection priority F of each merchant according to the predicted failure rate of each merchant*(i)；
S3, according to F, multiple merchants in the market*(i) In descending order, arrange F*(i) Automatically generating a merchant list by the K merchants with the maximum value, updating the checking condition of the merchants after the on-site checking is finished, and storing corresponding data in a market supervision platform;
and K is the number of merchants needing on-site inspection on the day and manually set by the supervisor.
Further, the step S1 is specifically:
s11, classifying all merchants in the market according to whether the inspection items of the sold products are the same or not;
s12, establishing a data set D ═ D { D } based on the historical check records of each type of merchants according to the data in the market platform1,D2,…DL}；
Wherein the subscript L represents the number of sales merchants belonging to the same class of products;
Di＝{(x1,y1),(x2,y2),…(xn,yn) Denotes the check record, x, for merchant inThe condition of each attribute of the merchant at the nth check is represented and recorded as
Figure BDA0001704773970000031
Figure BDA0001704773970000032
Value, y, representing attribute v at the time of the nth checknIndicates the result of the nth inspection, if the inspection is qualified yn0, otherwise yn＝1；
The attributes are corresponding inspection items when a supervisor carries out field inspection;
s13, according to the data set D of each type of merchant as N: 1 is randomly divided into a training set S and a test set T;
wherein the number of samples in the training set S is N times of the number of samples in the T;
s14, learning for each type of merchants by using a logistic regression model based on the training set S of each type of merchants to obtain a group of disqualification rate prediction models { f }(1),f(2),...,f(M)}；
Wherein M is the number of unqualified prediction models;
s15, using the test set T to predict the model { f) of M failure rates(1),f(2),...,f(M)Performing performance test, taking a model with the highest prediction accuracy as a final failure rate prediction model of the merchant, and recording as F;
s16, calculating the predicted failure rate F (i) (i is 1,2, …, m), wherein m represents the total number of merchants in the market, of each merchant based on the final failure rate prediction model of each type of merchant;
the calculation formula of the predicted failure rate F (i) of each merchant is as follows:
Figure BDA0001704773970000033
wherein the content of the first and second substances,
Figure BDA0001704773970000034
weight of j in final failure prediction model F representing the category of merchant i, b*The bias term in F is represented as,
Figure BDA0001704773970000035
representing the value of attribute j for the current merchant i.
Further, the final inspection priority F of each merchant in the step S2*(i) The calculation formula of (2) is as follows:
Figure BDA0001704773970000041
wherein f (i) represents the predicted failure rate for each merchant;
ti represents the time of the last on-site inspection from the merchant i;
γ ∈ (0,1) is the failure rate threshold set manually by the supervisor.
Further, the method for updating the merchant check condition in step S3 specifically includes:
t of merchant who will have completed on-site inspectioniSet to 0 and will check F for a non-qualified merchant*(i) Set directly to γ, i.e., the merchant is added directly to the spot check merchant list when the next spot check merchant list is generated.
Further, the method for training the kth unqualified prediction model in step S14 specifically includes:
wherein k is 1,2.. M, and M is the number of unqualified prediction models;
s141, randomly initializing a group of weights for the logistic regression model
Figure BDA0001704773970000042
And its offset b(k)；
The logistic regression model is as follows:
Figure BDA0001704773970000043
wherein x isiRepresenting the ith sample in the training set;
Figure BDA0001704773970000044
a weight representing a vth attribute in the kth prediction model;
s142, adjusting the weight
Figure BDA0001704773970000045
And its offset b(k)And optimizing the logistic regression model by using the parameters to obtain a kth unqualified prediction model.
Further, in step S142, the method for optimizing the logistic regression model specifically includes:
measurement of prediction model result f by mean square error loss function(k)(xi) With true mark yiAnd minimizing the error in the logistic regression model by a batch gradient descent method to obtain a kth unqualified prediction model.
The calculation formula of the mean square error loss function Z is as follows:
Figure BDA0001704773970000051
where | S | represents the number of samples in the training set S.
The invention has the beneficial effects that: the method for determining the spot-check merchants based on the big data automatically generates a merchant list which needs to be checked on site on the same day every day for a supervisor according to the information data of events which occur in the recent market in the supervision platform in real time, and effectively solves the problems that the conventional market merchants are inaccurate in spot-check process, low in efficiency and incapable of knowing the quality safety of real-time food products in time.
Drawings
Fig. 1 is a platform interface for a supervisor to perform field inspection in the background art of the present invention.
FIG. 2 is a flow chart of a method for identifying a spot check merchant based on big data according to an embodiment of the present invention.
FIG. 3 is a flow chart of a method for calculating the predicted failure rate for each merchant in an embodiment of the present invention.
FIG. 4 is a flowchart of a method for training a k-th merchant failure rate prediction model in an embodiment of the invention.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
As shown in fig. 2, a method for determining a spot check merchant based on big data includes the following steps,
s1, classifying all merchants in the market according to the inspection item data in the market supervision platform, and calculating the prediction reject ratio F (i) of each merchant;
wherein i is 1,2, …, m, m represents the total number of merchants in the market;
as shown in fig. 3, the step S1 specifically includes:
s11, classifying all merchants in the market according to whether the inspection items of the sold products are the same or not;
s12, establishing a data set D ═ D { D } based on the historical check records of each type of merchants according to the data in the market platform1,D2,…DL}；
Wherein the subscript L represents the number of sales merchants belonging to the same class of products;
Di＝{(x1,y1),(x2,y2),…(xn,yn) Denotes the check record, x, for merchant inIndicates the nth examinationThe conditions of the attributes of the merchant are recorded as
Figure BDA0001704773970000061
Figure BDA0001704773970000062
Value, y, representing attribute v at the time of the nth checknIndicates the result of the nth inspection, if the inspection is qualified yn0, otherwise yn＝1；
The attributes are corresponding inspection items when a supervisor carries out field inspection;
s13, according to the data set D of each type of merchant as N: 1 is randomly divided into a training set S and a test set T;
wherein the number of samples in the training set S is N times of the number of samples in the T;
s14, learning for each type of merchants by using a logistic regression model based on the training set S of each type of merchants to obtain a group of disqualification rate prediction models { f }(1),f(2),...,f(M)}；
Wherein M is the number of unqualified prediction models;
as shown in fig. 4, the method for training the kth unqualified prediction model in step S14 specifically includes:
wherein k is 1,2, … M;
s141, randomly initializing a group of weights for the logistic regression model
Figure BDA0001704773970000063
And its offset b(k)；
The logistic regression model is as follows:
Figure BDA0001704773970000064
wherein x isiRepresenting the ith sample in the training set;
Figure BDA0001704773970000071
is shown inA weight of a vth attribute in a kth prediction model;
s142, adjusting the weight
Figure BDA0001704773970000072
And its offset b(k)And optimizing the logistic regression model by using the parameters to obtain a kth unqualified prediction model.
The method for optimizing the logistic regression model specifically comprises the following steps:
measurement of prediction model result f by mean square error loss function(k)(xi) With true mark yiAnd minimizing the error in the logistic regression model by a batch gradient descent method to obtain a kth unqualified prediction model.
The calculation formula of the mean square error loss function Z is as follows:
Figure BDA0001704773970000073
where | S | represents the number of samples in the training set S.
Repeating the above steps S141-S142, i.e. repeating the random setting, for each unqualified predictive model training
Figure BDA0001704773970000074
And b(k)Obtaining M unqualified prediction models.
S15, using the test set T to predict the model { f) of M failure rates(1),f(2),...,f(M)Performing performance test, taking a model with the highest prediction accuracy as a final failure rate prediction model of the merchant, and recording as F;
s16, calculating the predicted failure rate F (i) (i is 1,2, …, m), wherein m represents the total number of merchants in the market, of each merchant based on the final failure rate prediction model of each type of merchant;
the calculation formula of the predicted failure rate F (i) of each merchant is as follows:
Figure BDA0001704773970000075
wherein the content of the first and second substances,
Figure BDA0001704773970000076
represents the weight of j in the final disqualification prediction model F of the category of the merchant i, b represents the bias term in F,
Figure BDA0001704773970000077
representing the value of attribute j for the current merchant i.
S2, calculating the final inspection priority F (i) of each merchant according to the predicted failure rate of each merchant;
final inspection priority F of each merchant in the above step S2*(i) The calculation formula of (2) is as follows:
Figure BDA0001704773970000081
wherein f (i) represents the predicted failure rate for each merchant;
ti represents the time of the last on-site inspection from the merchant i;
γ ∈ (0,1) is the failure rate threshold set manually by the supervisor.
S3, according to F, multiple merchants in the market*(i) In descending order, arrange F*(i) Automatically generating a merchant list by the K merchants with the maximum value, updating the checking condition of the merchants after the on-site checking is finished, and storing corresponding data in a market supervision platform;
and K is the number of merchants needing on-site inspection on the day and manually set by the supervisor.
The method for updating the merchant check condition in step S3 specifically includes: t of merchant who will have completed on-site inspectioniSet to 0 and will check F for a non-qualified merchant*(i) The direct setting is gamma, namely the merchant does not calculate the priority of the unqualified merchant at the next spot check, and the merchant directly enters the spot check merchant list.
In one embodiment of the invention, when the method is applied to the edible agricultural product market, a specific process of generating a spot check for a market merchant is provided:
when classifying merchants in the market according to inspection items of agricultural products sold, the following classifications are included, but not limited to: in a data set D established on the basis of historical inspection records of each type of merchants, the attributes and attribute values of each data sample are shown in a table 1:
TABLE 1 Attribute List Explanation for each data sample
Figure BDA0001704773970000082
Figure BDA0001704773970000091
When the merchants in the market are subjected to spot inspection generation, the merchants in the market are divided into meat, seafood, aquatic products, vegetables, eggs and fruits according to statistical data in the market supervision platform, wherein 8 merchants are related to meat products, data D of the meat merchants are established on the basis of historical inspection record data in the market supervision platform and are shown in a table 2, unqualified prediction models of the meat merchants are trained, inspection priorities of all the merchants are calculated, and inspection lists of the meat merchants are automatically generated. It should be noted that in practice, the inspection priority is calculated for each merchant, and the sampling list is automatically generated based on all categories of merchants.
TABLE 2 data set created with meat Merchant historical exam records
Figure BDA0001704773970000092
Figure BDA0001704773970000101
First, a data set D is created based on historical inspection records of 8 meat product merchants, as shown in Table 2, e.g., attribute vector x for sample 11＝(1 1 0 0 1 1 1 2 4.3)，y10; then, randomly dividing the data set D into a training set S and a testing set T according to a ratio of 3:1, wherein the number of samples contained in S is 24 samples {1,2,3,5,6,7,9,10,11,13,14,15,17,18,19,21,22,23,25,26,27,29,30,31} and the number of samples contained in S is 8 samples {4,8,12,16,20,24,28,32 };
secondly, learning on the training set S by using a logistic regression model to obtain 1 failure rate prediction model f(1)Where f is(1)Model parameter ω of(1)＝(-0.11,-1.11,-0.47,0.21,-0.45,-1.01,0.001,0.80,0.51)，b(1)＝0.01；
Resetting the initial values of the model parameters to obtain the 2 nd failure rate prediction model f(2)，ω(2)＝(-0.08,-0.9,0.2,0.1,-0.2,-1.1,0.2,0.82,-0.48)，b(2)＝0.2；
3 rd failure rate prediction model f(3)，ω(3)＝(-0.2,-0.8,0.24,0.05,-0.18,-0.9,-0.2,0.7,-0.6)，b(3)＝-0.1；
In the present embodiment, it is assumed that M is 3, that is, only 3 failure rate prediction models are trained for meat merchants; testing the accuracy of the 3 failure rate prediction models by using the test set T, wherein in the embodiment, the classification threshold of the accuracy of the failure rate prediction models of meat merchants is set to be 0.5, and the judgment result is unqualified when the predicted value is greater than or equal to 0.5; if the predicted value is less than 0.5, judging that the result is qualified; according to this setting, the prediction results shown in table 3 can be obtained;
TABLE 33 comparison of results for rejection prediction models
Business company a b c d e f g h
f(1) Fail to be qualified Fail to be qualified Fail to be qualified Fail to be qualified Fail to be qualified Fail to be qualified Fail to be qualified Fail to be qualified
f(2) Qualified Fail to be qualified Fail to be qualified Qualified Fail to be qualified Fail to be qualified Qualified Fail to be qualified
f(3) Qualified Fail to be qualified Fail to be qualified Qualified Fail to be qualified Fail to be qualified Qualified Qualified
Can calculate to obtain f(1)、f(2)、f(3)The accuracy of the model is 100%, 50% and 40%, respectively, so the f with the highest accuracy is selected(1)The failure rate prediction model is marked as F as a meat merchant failure rate prediction model; substituting the current attribute value of each merchant into the failure rate prediction model F to obtain the predicted failure rate of each merchant, wherein the table 4 shows the current attribute condition of 8 merchants;
Figure BDA0001704773970000112
the predicted failure rates of 8 merchants obtained by the same method are respectively as follows:
0.034,0.0562,0.833,0.0787,0.0787,0.722,0.955,0.214；
table 48 current attribute profiles for merchants
Figure BDA0001704773970000111
Figure BDA0001704773970000121
Then, the inspection priority F of each merchant is calculated using formula (4)*(i) Assume in this embodiment a merchant failure rate thresholdSince γ is 0.7 and the predicted failure rates of the merchants F and g are both greater than the threshold value, F*(c)＝0.833，F*(f)＝F(f)＝0.722，F*(g) 0.955; the predicted failure rates of the merchants a, b, d, e, h are all less than the threshold, so the merchants need to make corrections according to the time interval from the last inspection, and the inspection priority is calculated as follows:
F*(a)＝F(a)(1-2-2)＝0.034×0.75＝0.026
the same can be obtained:
F*(b)＝F(b)(1-2-4)＝0.053，F*(d)＝0.006，F*(e)＝0.008，F*(h)＝0.207
thirdly, 8 merchants are ranked from high to low according to inspection priority and are sequentially (g,0.955), (c,0.833), (f,0.722), (h,0.207), (b,0.053), (a,0.026), (e,0.008), (d,0.006), and in the embodiment, assuming that only 3 merchants are inspected on the day, the automatically generated list of spot inspection merchants is { (g,0.955), (c,0.833), (f,0.722) };
finally, the data record information in the market supervision platform is updated according to the inspection result of the current day, in this embodiment, assuming that the inspection result is that the merchant g is unqualified and the merchants c and d are qualified, the time interval between the last inspection and the 3 merchants is changed to 0, and the F of the merchant g is changed to 0*(g) And setting the reject rate threshold to be 0.7, and directly adding the spot check merchant list when generating a next spot check list.
The method for determining the spot-check merchants based on the big data automatically generates a merchant list which needs to be checked on site on the same day every day for a supervisor according to the information data of events which occur in the recent market in the supervision platform in real time, and effectively solves the problems that the conventional market merchants are inaccurate in spot-check process, low in efficiency and incapable of knowing the quality safety of real-time food products in time.

Claims (3)

1. A method for determining a spot check merchant based on big data is characterized by comprising the following steps,
s1, classifying all merchants in the market according to the inspection item data in the market supervision platform, and calculating the prediction reject ratio F (i) of each merchant;
wherein i ═ 1,2.. m, m denotes the total number of merchants in the market;
s2, calculating the final inspection priority F of each merchant according to the predicted failure rate of each merchant*(i)；
S3, according to F, a plurality of merchants in the market*(i) In descending order, arrange F*(i) Automatically generating a merchant list by the K merchants with the maximum value, updating the checking condition of the merchants after the on-site checking is finished, and storing corresponding data in a market supervision platform;
k is the number of merchants needing on-site inspection on the day and manually set by a supervisor;
the step S1 specifically includes:
s11, classifying all merchants in the market according to whether the inspection items of the sold products are the same or not;
s12, establishing a data set D ═ D { D } based on the historical check records of each type of merchants according to the data in the market platform1,D2,…DL}；
Wherein the subscript L represents the number of sales merchants belonging to the same class of products;
Di＝{(x1,y1),(x2,y2),…(xn,yn) Denotes the check record, x, for merchant inThe condition of each attribute of the merchant at the nth check is represented and recorded as
Figure FDA0003167492240000011
Figure FDA0003167492240000012
Value, y, representing attribute v at the time of the nth checknIndicates the result of the nth inspection, if the inspection is qualified yn0, otherwise yn＝1；
The attributes are corresponding inspection items when a supervisor carries out field inspection;
s13, according to the data set D of each type of merchant as N: 1 is randomly divided into a training set S and a test set T;
wherein the number of samples in the training set S is N times of the number of samples in the T;
s14, learning for each type of merchants by using a logistic regression model based on the training set S of each type of merchants to obtain a group of disqualification rate prediction models { f }(1),f(2),...,f(M)}；
Wherein M is the number of unqualified prediction models;
s15, using the test set T to predict the model { f) of M failure rates(1),f(2),...,f(M)Performing performance test, taking a model with the highest prediction accuracy as a final failure rate prediction model of the merchant, and recording as F;
s16, calculating the predicted failure rate F (i) (i is 1,2, …, m), wherein m represents the total number of merchants in the market, of each merchant based on the final failure rate prediction model of each type of merchant;
the calculation formula of the predicted failure rate F (i) of each merchant is as follows:
Figure FDA0003167492240000021
wherein the content of the first and second substances,
Figure FDA0003167492240000022
weight of j in final fail prediction model F representing category to which merchant i belongs, b*The bias term in F is represented as,
Figure FDA0003167492240000023
a value representing attribute j of the current merchant i;
final inspection priority F of each merchant in said step S2*(i) The calculation formula of (2) is as follows:
Figure FDA0003167492240000024
wherein f (i) represents the predicted failure rate for each merchant;
ti represents the time of the last on-site inspection from the merchant i;
gamma belongs to (0,1) as a failure rate threshold value manually set by a supervisor;
the method for updating the merchant check condition in step S3 specifically includes:
t of merchant who will have completed on-site inspectioniSet to 0 and will check F for a non-qualified merchant*(i) Set directly to γ, i.e., the merchant is added directly to the spot check merchant list when the next spot check merchant list is generated.
2. The big-data-based method for determining spot-checked merchants according to claim 1, wherein the method for training the kth unqualified prediction model in step S14 specifically comprises:
wherein k is 1,2.. M, and M is the number of unqualified prediction models;
s141, randomly initializing a group of weights for the logistic regression model
Figure FDA0003167492240000031
And its offset b(k)；
The logistic regression model is as follows:
Figure FDA0003167492240000032
wherein x isiRepresenting the ith sample in the training set;
Figure FDA0003167492240000033
a weight representing a vth attribute in the kth prediction model;
s142, adjusting the weight
Figure FDA0003167492240000034
And its offset b(k)And optimizing the logistic regression model by using the parameters to obtain a kth unqualified prediction model.
3. The big-data-based method for determining the spot-checked merchants according to claim 2, wherein in the step S142, the method for optimizing the logistic regression model specifically comprises:
measurement of prediction model result f by mean square error loss function(k)(xi) With true mark yiMinimizing the error in the logistic regression model by a batch gradient descent method to obtain a kth unqualified prediction model;
the calculation formula of the mean square error loss function Z is as follows:
Figure FDA0003167492240000035
where | S | represents the number of samples in the training set S.
CN201810650948.XA 2018-06-22 2018-06-22 Method for determining spot-check merchants based on big data Expired - Fee Related CN108830492B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810650948.XA CN108830492B (en) 2018-06-22 2018-06-22 Method for determining spot-check merchants based on big data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810650948.XA CN108830492B (en) 2018-06-22 2018-06-22 Method for determining spot-check merchants based on big data

Publications (2)

Publication Number Publication Date
CN108830492A CN108830492A (en) 2018-11-16
CN108830492B true CN108830492B (en) 2022-02-08

Family

ID=64137393

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810650948.XA Expired - Fee Related CN108830492B (en) 2018-06-22 2018-06-22 Method for determining spot-check merchants based on big data

Country Status (1)

Country Link
CN (1) CN108830492B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109871978B (en) * 2018-12-28 2023-07-18 广州兴森快捷电路科技有限公司 PCB order qualification rate prediction method and device and readable storage medium
CN110826855A (en) * 2019-10-09 2020-02-21 广州供电局有限公司 Method and system for testing network access performance of intelligent power distribution room state monitoring sensor
CN112101819A (en) * 2020-10-28 2020-12-18 平安国际智慧城市科技股份有限公司 Food risk prediction method, device, equipment and storage medium
CN115660260A (en) * 2022-12-28 2023-01-31 深圳市四格互联信息技术有限公司 Method and system for dynamically generating inspection task of property management

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6675149B1 (en) * 1998-11-02 2004-01-06 International Business Machines Corporation Information technology project assessment method, system and program product
CN105096151A (en) * 2014-05-15 2015-11-25 中国移动通信集团公司 Information recommendation method, device, and server
CN106897883A (en) * 2017-01-23 2017-06-27 武汉奇米网络科技有限公司 One kind automation commodity sampling observation method and system
CN107103487A (en) * 2017-03-02 2017-08-29 浙江兰德纵横网络技术股份有限公司 A kind of advertisement sending method based on big data

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6675149B1 (en) * 1998-11-02 2004-01-06 International Business Machines Corporation Information technology project assessment method, system and program product
CN105096151A (en) * 2014-05-15 2015-11-25 中国移动通信集团公司 Information recommendation method, device, and server
CN106897883A (en) * 2017-01-23 2017-06-27 武汉奇米网络科技有限公司 One kind automation commodity sampling observation method and system
CN107103487A (en) * 2017-03-02 2017-08-29 浙江兰德纵横网络技术股份有限公司 A kind of advertisement sending method based on big data

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
一种基于结构化学习的排序算法;程凡等;《计算机工程与应用》;20110421;第16-19页 *

Also Published As

Publication number Publication date
CN108830492A (en) 2018-11-16

Similar Documents

Publication Publication Date Title
CN108830492B (en) Method for determining spot-check merchants based on big data
CN108564286B (en) Artificial intelligent financial wind-control credit assessment method and system based on big data credit investigation
CN104077306B (en) The result ordering method and system of a kind of search engine
US20190236497A1 (en) System and method for automated model selection for key performance indicator forecasting
CN108629436B (en) Method and electronic equipment for estimating warehouse goods picking capacity
CN109034483B (en) Detection planning method based on quality function configuration
CN109146611A (en) A kind of electric business product quality credit index analysis method and system
CN107545038A (en) A kind of file classification method and equipment
CN107808346A (en) A kind of appraisal procedure and apparatus for evaluating of potential target object
Mazurkiewicz et al. Universal methodology for the innovative technologies assessment
CN110738565A (en) Real estate finance artificial intelligence composite wind control model based on data set
CN113538021B (en) Machine learning method for store duration prediction
CN111330871B (en) Quality classification method and device
CN108399545B (en) Method and device for detecting quality of electronic commerce platform
CN116485279B (en) Equipment information processing method and device based on water management platform
CN111212434A (en) WIFI module quality prediction method, device, equipment and storage medium
CN116912016A (en) Bill auditing method and device
Deodhar Motivation for and cost of HACCP in Indian food processing industry
CN108805603A (en) Marketing activity method for evaluating quality, server and computer readable storage medium
CN115081921A (en) ERP e-commerce management system based on big data
Maryam et al. Evaluate implementation of enterprise resource planning (ERP) system
CN114693428A (en) Data determination method and device, computer readable storage medium and electronic equipment
CN115062687A (en) Enterprise credit monitoring method, device, equipment and storage medium
JP7281708B2 (en) Manufacturing condition calculation device, manufacturing condition calculation method, and manufacturing condition calculation program for identifying equipment that contributes to the generation of defective products
Uddin et al. Comparison of some statistical forecasting techniques with GMDH predictor: A case study

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
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20220208