CN109858716A - A kind of gas station's Benefit Calculation based on composite index performance analysis model - Google Patents
A kind of gas station's Benefit Calculation based on composite index performance analysis model Download PDFInfo
- Publication number
- CN109858716A CN109858716A CN201711204491.1A CN201711204491A CN109858716A CN 109858716 A CN109858716 A CN 109858716A CN 201711204491 A CN201711204491 A CN 201711204491A CN 109858716 A CN109858716 A CN 109858716A
- Authority
- CN
- China
- Prior art keywords
- data
- gas station
- composite index
- station
- analysis model
- 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 invention discloses a kind of gas station's Benefit Calculations based on composite index performance analysis model, the described method includes: data-optimized algorithm of the step 1) based on sample optimization and characteristic optimization, gas station's basic data is filtered out from data assay value, and three system scoring factors are filtered out according to correlation analysis;Step 2) determines the weighting coefficient of three system scoring factors;Construct composite index performance analysis model;Step 3) obtains the real time data of gas station, and the benefit of gas station is calculated according to composite index performance analysis model.Invention proposition method has effectively merged the data structure of conventional systems, and sample optimization and characteristic optimization are carried out to data, it is theoretical according to SPSS correlation analysis, a kind of composite index performance analysis model using stationary time series is realized, the acquisition and analysis processing of polynary isomeric data are finally reached.
Description
Technical field
The present invention relates to gasoline station management fields, and in particular to a kind of gas station based on composite index performance analysis model
Benefit Calculation.
Background technique
Breath is ceased as the stable development of social economy and the continuous improvement petroleum of living standards of the people become to live with people
A kind of relevant strategic resource.Whether enterprises and institutions are to the control of vehicle oil or oil marketer to the pipe of gas station
Reason all changes from extensive style to intensive style.Undoubtedly become gas station in worldwide computer technology today of information technology to advise
Model management, save the cost, a kind of effective means improved benefit.Current middle-size and small-size gas station due to fund, technology etc. its
The also relatively backward also rare water for resting on labor management mostly being managed using information technology of the management level of gas station
On flat, and process of IT application majority rests on the acquisition of simple oil-filling data, record.Common fuel station information library case
System function include oil product sales volume, inventory statistics, amount balance, statement analysis or marquee account unit use oil management and
Receipt, invoice are issued.Management, analysis, decision is all is accomplished manually, the various drawbacks of band, meanwhile, common gas station
Information system management platform can not be parsed for polynary isomeric data, can analyze data sheet one, the linearisation of processing, only
It is limited to the acquisition and displaying of data, reasonable decision opinion can not be provided to gasoline station management personnel.
Consider from entire oil sale company, since each gas station's scale is different, geographical location is different, periphery is competing
Difference is striven, then which comprehensive benefit between website and website how is judged using the model algorithm of science can be more preferable, how
It intuitively finds the advantage of gas station itself and insufficient and promote itself comprehensive operation ability in the later period, is oil sale in recent years
The urgent need of parent company.
Summary of the invention
It is an object of the present invention to which overcome can no longer meet at this stage for traditional fuel station information platform at present
Gas station's data magnanimity, the demand of diversification, isomerization, propose a kind of oiling based on composite index performance analysis model
Stand Benefit Calculation, this method can in conjunction with the road traffic of gas station, the vehicle quantity that enters the station, the vehicle that enters the station, Oil supplier,
The data factors such as refueling nozzle number, moon task index and every daily sales establish a kind of composite index performance analysis model, realization pair
Comprehensive benefit assessment between multiple and different gas stations instructs the resource of gas station to carry out reasonable disposition and excellent by analyzing result
Change, enhances comprehensive operation ability, greatly improve overall efficiency.
To achieve the goals above, the invention proposes a kind of gas station's benefits based on composite index performance analysis model
Calculation method, which comprises
Data-optimized algorithm of the step 1) based on sample optimization and characteristic optimization, filters out gas station from data assay value
Basic data filters out three system scoring factors according to correlation analysis;
Step 2) determines the weighting coefficient of three system scoring factors;Construct composite index performance analysis model;
Step 3) obtains the real time data of gas station, and the benefit of gas station is calculated according to composite index performance analysis model.
As a kind of improvement of the above method, the step 1) includes:
Step 101) carries out correlation analysis to data assay value, removes influence degree phase according to calculated related coefficient
Same data;
The data assay value include: each gas station attendant's number, each gas station include fuel oil gun number,
Road traffic, gas station progress vehicle number, date, moon sales volume, moon oil product sales volume where each station gasoline storage capacity, gas station
Operational indicator, progression rates use rifle number, accumulative sales volume per capita and complete ratio;
Step 102) optimizes data sample using feature space samples selection method, selects most in feature space
Representative sample characterizes entire sample set;
Step 103) carries out the principal component analysis based on kernel function to the sample after optimization, obtains principal component feature breath;
Step 104) screens principal component characteristic information using Estimation of Distribution Algorithm, is guaranteeing state characteristic information not
Really under the premise of, more identification informations is selected to enter subsequent index computation model, realizes characteristic optimization;
Step 105) levies vector selection algorithm according to kernel optimization distribution and filters out oiling website basic data;
Step 106) carries out correlation analysis to oiling website basic data, and combination obtains the system scoring factor.
As a kind of improvement of the above method, oiling website basic data that the step 105) filters out are as follows:
Oiling staff's number;The number of whole refueling nozzles in gas station;The number of vehicles that daily inbound is refueled;It refuels
The vehicle number crossed daily outside standing;Oily tonnage and oil gas moon sale tonnage index are sold daily.
As a kind of improvement of the above method, the system of the step 106) scores the factor are as follows: people Qiang Bi ﹑ enters the station rate and complete
At than;
People's rifle ratio is " rifle number/number " ratio, reflects the proportion situation of oiling staff and refueling nozzle quantity;
The rate that enters the station is " amount of pulling in/vehicle flowrate " ratio, and reflection gas station attracts clients the ability of oiling of entering the station,
The completion is than dynamically reflecting a moon index performance for " day sales volume/moon index " ratio.
As a kind of improvement of the above method, the step 2) includes:
Step 201) is standardized input data:
Since the unit of the different factors is different, so data are normalized, it is public using general normalization
Formula:
Newvalue=(oldvalue-min)/(max-min)
Wherein Newvalue indicates dimensionless number after normalization, and max, min respectively indicate the maximum value in this group of numerical value
And minimum value, oldvalue are raw value;Then data normalization is handled;
Step 202) does canonical correlation analysis to the data after standardization and carries out significance test, obtains phase relation
Number;
Related coefficient is respectively as follows: 0.039,0.546,0.560, three values be respectively people's rifle ratio ﹑ enter the station rate, complete the power of ratio
Weight coefficient;
Step 203) constructs composite index performance analysis model;
Enable a=39;B=546;C=560 obtains the calculation formula of composite index:
A=a × P+b × Q+c × R
Wherein: A indicates composite index, and P indicates that people's rifle ratio, Q indicate the rate that enters the station, and R indicates the ratio that hits the target day, above-mentioned public affairs
Formula is composite index performance analysis model.
Present invention has an advantage that
The present invention proposes that method has effectively merged the data structure of conventional systems, and carries out sample optimization to data
And characteristic optimization, it is theoretical according to SPSS correlation analysis, it is realized and a kind of is referred to using the synthesis of stationary time series using R language
Number performance analysis model is finally reached the acquisition and analysis processing of polynary isomeric data, and can be according to the performance analysis proposed
Model provides gas station's composite index score, and administrative staff can score according to system understands gas station's all data and profit shape
State, this method really realize fuel station information, intelligence.
Detailed description of the invention
The flow chart of Fig. 1 gas station's Benefit Calculation of the invention based on composite index performance analysis model;
Fig. 2 is Mare Tranquillitatis Jing Fu highway Dongcheng gas station basic data figure;
Fig. 3 is Mare Tranquillitatis oiling center gas station basic data figure;
Fig. 4 is Jinghai County, gas station, west of a city basic data figure;
Fig. 5 is the gas station Mare Tranquillitatis Tuan Bowa Jin Wang basic data figure;
Fig. 6 is the gas station Mare Tranquillitatis Xia Guantun basic data figure;
Fig. 7 is scatter plot;
Fig. 8 is the comprehensive benefit schematic diagram of five gas stations;
Fig. 9 is that the comprehensive benefit of five gas stations compares figure.
Specific embodiment
Method of the invention is described in detail in the following with reference to the drawings and specific embodiments.
As shown in Figure 1, a kind of gas station's Benefit Calculation based on composite index performance analysis model, which comprises
Data-optimized algorithm of the step 1) based on sample optimization and characteristic optimization filters out six from data assay value and adds
Petrol station basic data filters out three system scoring factors according to correlation analysis;It specifically includes:
Data assay value is imported SPSS and carries out correlation analysis by step 101), removes shadow according to calculated related coefficient
The identical data of the degree of sound;
The data assay value include: each gas station attendant's number, each gas station include refueling nozzle rob it is ripe, every
Road traffic, gas station progress vehicle number, date, moon sales volume, moon oil product pin where a station 95# gasoline storage capacity, gas station
Measure operational indicator, progression rates, per capita using rifle number, accumulative sales volume and completion ratio.
The related coefficient of calculating includes: typical load coefficient: a canonical variable organizes the simple correlation of all variables with this
Coefficient;Intersect load coefficient: the simple correlation coefficient of a canonical variable and another group of each variable of set of variables.
Step 102) optimizes data sample using feature space samples selection method, selects most in feature space
Representative sample characterizes entire sample set;
The purpose of sample optimization is to eliminate similar sample, can further decrease computational complexity.
Step 103) carries out the principal component analysis based on kernel function to the sample after optimization, obtains principal component feature breath;
Sample inner product in feature space is replaced using kernel function.
Step 104) screens principal component characteristic information using Estimation of Distribution Algorithm, is guaranteeing state characteristic information not
Really under the premise of, more identification informations is selected to enter subsequent index computation model as far as possible, realizes characteristic optimization;
Step 105) levies the oiling website basic data that vector selection algorithm filters out according to kernel optimization distribution;
Since the feature vector selection of kernel optimization is a discrete process, it is difficult to be carried out with general analysis means
It solves, and Estimation of Distribution Algorithm can then solve this discrete optimization process and be avoided by using discrete type coding form
The problem of traditional genetic algorithm easily falls into locally optimal solution, therefore, it is proposed to based on special using the distribution of kernel optimization
Vector selection algorithm is levied, algorithm steps are as follows:
Firstly, generating initial population Dl(l=0).Monitoring data are carried out in feature space using kernel optimization algorithm
Principal component analysis obtains N number of feature vector arAnd corresponding eigenvalue λrWherein r=1,2,3...N arranges characteristic value in descending order
Column select the corresponding feature vector of several biggish characteristic values as way principal component is waited, remember K=(a1,...an)。
Stochastic variable is done with feature vector a, generates the initial sample D that population number is M0.Using binary coding mode pair
Feature vector arIt is encoded, each in coding represents whether corresponding feature vector is selected, and indicates to choose the spy for 1
Vector is levied, indicates unselected for 0.
Then, fitness value, selection individual are calculatedBy comparing the adaptation of individual each in population's fitness function
It spends to select excellent individual to carry out individual information statistics, to generate new sample.
Fitness function setting are as follows:
Fit=J (KPi)×F(KPi) (1)
J(KPi) indicate selected loud recognition capability, F (KPi) indicate constraint function.
Wherein J (KPi)=tr (Sb)/tr(Sw), tr (Sb) indicate shatter value matrix in class, J (KPi) value is bigger, indicate one
Distance is bigger between class sample, and similar sample illustrates that selected feature combined effect is good apart from small.
The oiling filtered out from data assay value based on the distributed nature vector selection algorithm using kernel optimization
Website basic data are as follows:
Oiling staff number (being abbreviated as number);The number (being abbreviated as rifle number) of whole refueling nozzles in gas station;Often
The number of vehicles (being abbreviated as the amount of pulling in) that its inbound is refueled;The vehicle number (being abbreviated as vehicle flowrate) crossed daily outside gas station;Often
It sells oily tonnage (being abbreviated as a day sales volume) and oil gas moon sale tonnage index (being abbreviated as a moon index).
To sum up, oiling website basic data is the number ﹑ rifle number ﹑ amount of pulling in ﹑ days sales volumes of ﹑ vehicle stream amount and moon index.Such as figure
Shown in 2, Fig. 3, Fig. 4, Fig. 5 and Fig. 6.
Step 106) carries out correlation analysis to oiling website basic data, and combination obtains the system scoring factor;
Firstly, carrying out data dependence analysis according to different types of statistic.Correlation analysis is to utilize generalized variable
Correlativity between is come the Multielement statistical analysis method of the overall relevancy reflected between two groups of indexs.
The basic principle of correlation is: in order to hold the correlativity between two groups of indexs on the whole, respectively at two groups
It is extracted in variable representational two generalized variables U1 and V1 (linear combination of each variable in respectively two set of variables), benefit
Reflect the overall relevancy between two groups of indexs with the correlativity between the two generalized variables.
It is dependent variable that we, which choose total sales volume, and other factors are independent variable, and it is related to be programmed into row data using R language platform
Property analysis, obtain the related coefficient between different data, and result verification is carried out using statistics software SPSS.
Wherein, the related coefficient that we calculate includes typical load coefficient: a canonical variable organizes all variables with this
Simple correlation coefficient;Intersect load coefficient: the simple correlation coefficient of a canonical variable and another group of each variable of set of variables.
Finally, according to the size of different factor related coefficients, different factors shared weight in composite index again is determined.
6 groups of data are subjected to combination of two, filter out direct relation and important data combination, in which:
Rifle number combined with number in " rifle number/number " ratio be oiling staff and refueling nozzle quantity proportion situation,
Abbreviation people rifle ratio;
The amount of pulling in is combined with vehicle flowrate, and " amount of pulling in/vehicle flowrate " this ratio, which represents gas station and attracts clients, to enter the station
The ability of oiling, referred to as enter the station rate;
Day sales volume and moon indicator combination, day sales volume add up be known as day sales volume day by day, when being exactly the moon to last day of the place moon
Sales volume, in order to embody the dynamic process of sales volume in one month, ratio " day sales volume/moon index " Dynamically Announce moon index completes feelings
Condition referred to as completes ratio.
Eventually by above-mentioned processing, 6 groups of basic datas have obtained 3 groups of data, and for the system scoring factor: people Qiang Bi ﹑ enters the station
Rate and completion ratio, these three impact factors are exactly the scoring element between station and station, pass through this three groups of data at each station, calculate
Obtain the comprehensive evaluation index at the station, abbreviation composite index.
Step 2) determines the weighting coefficient of three system scoring factors;Obtain composite index performance analysis model;
The determination of weighting coefficient will ensure reasonability and accuracy, since gas station concerns profit and sales volume, can incite somebody to action
" people's rifle ratio " " enter the station rate " " completing ratio " passes through proportional tune to the correlation score of same class data (selecting day sales volume herein)
After whole, as weighting coefficient.Weighting coefficient determination process is provided with example operation.
Performance analysis system uses R language development platform, and R language is a kind of flexible programming language, aims at promotion and explores
Property data analysis, the classical theory of statistics test and advanced figure and design.R possess it is abundant, still in ever-expanding data packet
Library, the forward position in statistics, data analysis and data mining development.R has been integrated into multiple commercial packets, such as IBM SPSS
With InfoSphere, Mathematica.
It is determining model coefficient that this programme, which establishes the premise of model using R language, and detailed process is as follows:
(1) it is loaded into initial data: data.frame
(2) initial data standardizes: scale
(3) canonical correlation analysis: cancor
(4) test of significance of coefficient of correlation: corcoef.test.R
Now carry out sampling analysis to 20 groups of gas station's performance indicators: gas station's number (X1), is sold every day the vehicle number that enters the station (X2)
It measures (X3);
Three training quotas: gas station's rifle number (Y1), single channel flow (Y2), odd-numbered day need to complete oiling amount (Y3);
The selection of index meets composite index set forth above, suits gas station's actual conditions, examination this group of data of analysis
Correlation.Steps are as follows:
1) the form input data of data frame is used:
test<-data.frame(
X1=c (191,193,189,211,176,169,154,193,176,156,189,162,182,167,154,
166,247,202,157,138),
X2=c (36,38,35,38,31,34,34,36,37,33,37,35,36,34,33,33,46,37,32,33),
X3=c (50,58,46,56,74,50,64,46,54,54,52,62,56,60,56,52,50,62,52,68),
Y1=c (5,12,13,8,15,17,14,6,4,15,2,12,4,6,17,13,1,12,11,2),
Y2=c (162,101,155,101,200,120,215,70,60,225,110,105,101,125,25 1,210,
50,210,230,110),
Y3=c (60,101,58,38,40,38,105,31,25,73,60,37,42,40,250,115,50,1 20,80,43)
Tables of data is obtained, since the unit of the different factors is different, so data are normalized, use is general
Normalization formula:
Newvalue=(oldvalue-min)/(max-min)
Wherein Newvalue indicates dimensionless number after normalization, and max, min respectively indicate the maximum value in this group of numerical value
And minimum value, oldvalue are raw value.Then scale function, test <-scale are called into data normalization processing
(test)
2) canonical correlation analysis is done to standardized data, and checks analysis result wherein:
#cor is canonical correlation coefficient;
The coefficient that #xcoef corresponds to data x is also known as the loads typical i.e. sample canonical variable overall coefficient of heat transfer square about data x
The transposition of battle array A;
#xcenter is the center i.e. sample average of data X of data X;
The coefficient that #y corresponds to data x is also known as the loads typical i.e. sample canonical variable V coefficient matrix B about data y
Transposition;
#ycenter is the center i.e. sample average of data Y of data Y.
Therefore, it can analyze and obtain from program operation result, corresponding first group of x value, y value;Second group of x value, y value;The
Three groups of x values, y value related coefficient be respectively
Calculate score U=AX V=BY of the data under canonical variable
U <-as.matrix (test [, 1:3]) %*%ca $ xcoef
V <-as.matrix (test [, 4:6]) %*%ca $ ycoef
Scatter plot is drawn by taking first group of variable as an example
Plot (U [, 1], V [, 1], xlab=" U1 ", ylab=" V1 ")
Scatter plot is shown in Fig. 7;By scatter plot it is found that the first canonical correlation variable is distributed near straight line.
3) significance test of canonical correlation coefficient
Purpose as correlation analysis is exactly to select how many pairs of canonical variables;Therefore need to do the significant of canonical correlation coefficient
Property examine.If thinking, related coefficient k is 0, it is not necessary to consider kth to canonical variable.
This programme shows first pair of canonical correlation variable of selection according to final program operation result.
Operation result shows that the selected three groups of dependent variables of model are reasonable dependent variable.
Here it three after preservation decimal point, obtains related coefficient and is respectively as follows: 0.039:0.546:0.560.
Expand 1000 times simultaneously according to uniformity principle, obtaining ratio: 39:546:560 can enable a=39;B=546;C=
560, to obtain the calculation formula of composite index:
A=a × P+b × Q+c × R
Wherein: A indicates composite index, and P indicates that people's rifle ratio, Q indicate the rate that enters the station, and R indicates the ratio that hits the target day.
A ﹑ b ﹑ c is respectively the weighting coefficient of multiplied item.
Analysis gained weight coefficient is brought into formula to obtain:
A=39 × P+546 × Q+560 × R
The moon composite index of each website can be calculated according to formula, and oil station Mare Tranquillitatis ﹑ is added with Jing Fu highway Dongcheng, Mare Tranquillitatis
For oiling center adds the oil station Jinghai County ﹑ west of a city to add the oil station Mare Tranquillitatis ﹑ Tuan Bowa saliva prosperous plus the oil station Mare Tranquillitatis ﹑ gas station Xia Guantun, it
Moon composite index result be detailed in Fig. 8 and Fig. 9.
4) data distribution and modeling
In practical modeling process, all data are not used to all be trained model by this programme, because comparing
Compared with performance of the model data collection in training, what is be more concerned about is the training set of model, that is, in the data that do not encounter of model
Prediction performance.
Therefore, 70% data of data set are used to training pattern, remaining 30% is used to the knot of testing model prediction
Fruit.
Adopt CSV format storage initial data such as data includes 15 groups of variables altogether, last group of aggregation of variable index is to need
Obtained variable, before 14 variables be all predictive factor, be each for describing gas station's attribute.setwd("C:/
Users/shi hui/Desktop/ sample) ")
It is loaded into analysis data packet
requi re(readr)
requi re(ggplot2)
requi re(dplyr)
requi re(tidyr)
requi re(caret)
requi re(corrplot)
requi re(Hmisc)
requi re(parallel)
requi re(doParallel)
requi re(ggthemes)
Carry out population processing
n_Cores<-detectCores()
cluster_Set<-makeCluster(n_Cores)
registerDoParallel(cluster_Set
Since the data of gas station's collection in worksite are there are missing values, influences will cause on follow-up data processing, it should to scarce
Mistake value carries out interpolation processing:
random_Number<-sample(1:3168,30)
Oil_Original1<-Oil_Original
Oil_Original[random_Number,1]<-NA
describe(Oil_Original)
Then by be by be set as missing values original value and interpolation after value be integrated in a data frame
compare_Imputation<-data.frame(Oil_Original1[random_Number,1],Oil_
Original[random_Number,1])
Dimension is gone with PreProcess function now
Pp <-preProcess (Oil_Train, method=c (" scale ", " center ", " pca ")) Oil_Train
<-predict(pp,Oil_Train)Oil_Test<-predict(pp,Oil_Test)
First the numeric type factor is standardized, it is ensured that all factors are in a dimension, then to standard
The data of change carry out principal component analysis.
5) Logistic regression model is introduced
Consider that there are 14 groups of Independent Vector models, Logistic recurrence is carried out to all variables first with glm function and is built
Mould, glm are assigned to the description of a symbolistic description linear prediction error distribution for meeting generalized linear model.
Fit <-glm (y~, train, family=" binomial ") summary (fit)
Models fitting effect is carried out using total regression variable, wherein the date, the p value of memory capacity fails through verifying, can
Directly to reject, y is returned using surplus variable.
Second all variable of regression model has all passed through inspection or even AIC value (red pond information criterion) smaller, Suo Youmo
The fitting effect of type is better.Then evidence weight conversion is carried out, Logistic regression model can be changed into scale
Card format.The purpose for introducing WOE conversion is not intended to improve model quality, and only some variables should not be included into model,
This is not either because they can increase model value, or because error related with its model related coefficient is larger, in fact
Establishing standard credit scorecard can not also use WOE to convert.In this case, the processing of Logistic regression model needs is bigger
The independent variable of quantity.Even now will increase the complexity of modeling program, but finally obtained scorecard is the same.To surplus
Under variable carry out WOE conversion.
Step 3) obtains the real time data of each gas station, and the effect of gas station is calculated according to composite index performance analysis model
Benefit, and score.
The creation and implementation of composite index.The creation of composite index be derived from scale card system, it is popular for be exactly to comment
Divide and needs oneself to preset a threshold values.
If the sales volume of this gas station today has reached 0.8 (completing 80%) of day amount, this gas station is set
Basic score value is 500 points;
If the sales volume of this gas station today has reached 0.9 (completing 90%) of day amount, this gas station is set
Basic score value is 600 points;
The setting of threshold values need to be according to the continuous tracking adjustment of industry experience, and score setting below only represents project Initial experience
Value, with the propulsion of project, the update of data, threshold value will be constantly adjusted.
Start to set up scoring below, it is assumed that comparing 15 by quality is 600 points, and every high 20 points of quality calculate P than doubling, Q.Such as
Fruit anaphase is unobvious, can the fine or not Bizets of high 30-50 point double.
Score=q-p*log (odds)
There is equation:
620=q-p*log (15)
600=q-p*log (15/2)
TrainWOE $ y=1-train $ y
Glm.fit=glm (y~, data=trainWOE, family=binomial (link=logit))
summary(glm.fit)
Coe=(glm.fit $ coefficients)
Wherein basis point calculation method is
base<-q+p*as.numeric(coe[1])
base
It gives a mark to each variable, by taking one group of progression rates as an example:
Construction calculates score value function:
Ultimately generate scorecard.
It should be noted last that the above examples are only used to illustrate the technical scheme of the present invention and are not limiting.Although ginseng
It is described the invention in detail according to embodiment, those skilled in the art should understand that, to technical side of the invention
Case is modified or replaced equivalently, and without departure from the spirit and scope of technical solution of the present invention, should all be covered in the present invention
Scope of the claims in.
Claims (5)
1. a kind of gas station's Benefit Calculation based on composite index performance analysis model, which comprises
Data-optimized algorithm of the step 1) based on sample optimization and characteristic optimization filters out gas station basis from data assay value
Data filter out three system scoring factors according to correlation analysis;
Step 2) determines the weighting coefficient of three system scoring factors;Construct composite index performance analysis model;
Step 3) obtains the real time data of gas station, and the benefit of gas station is calculated according to composite index performance analysis model.
2. gas station's Benefit Calculation according to claim 1 based on composite index performance analysis model, feature
It is, the step 1) includes:
Step 101) carries out correlation analysis to data assay value, and it is identical to remove influence degree according to calculated related coefficient
Data;
The data assay value includes: the fuel oil gun number, each that each gas station attendant's number, each gas station include
It stands gasoline storage capacity, road traffic, gas station progress vehicle number, date, moon sales volume, moon oil product sales volume business where gas station
Index, progression rates use rifle number, accumulative sales volume per capita and complete ratio;
Step 102) optimizes data sample using feature space samples selection method, most represents in feature space selection
The sample of property characterizes entire sample set;
Step 103) carries out the principal component analysis based on kernel function to the sample after optimization, obtains principal component feature breath;
Step 104) screens principal component characteristic information using Estimation of Distribution Algorithm, is guaranteeing that state characteristic information is not certain
Under the premise of, it selects more identification informations to enter subsequent index computation model, realizes characteristic optimization;
Step 105) levies vector selection algorithm according to kernel optimization distribution and filters out oiling website basic data;
Step 106) carries out correlation analysis to oiling website basic data, and combination obtains the system scoring factor.
3. gas station's Benefit Calculation according to claim 2 based on composite index performance analysis model, feature
It is, the oiling website basic data that the step 105) filters out are as follows:
Oiling staff's number;The number of whole refueling nozzles in gas station;The number of vehicles that daily inbound is refueled;Outside gas station
The vehicle number crossed daily;Oily tonnage and oil gas moon sale tonnage index are sold daily.
4. gas station's Benefit Calculation according to claim 3 based on composite index performance analysis model, feature
It is, the system of the step 106) scores the factor are as follows: people Qiang Bi ﹑, which enters the station, rate and completes ratio;
People's rifle ratio is " rifle number/number " ratio, reflects the proportion situation of oiling staff and refueling nozzle quantity;
The rate that enters the station is " amount of pulling in/vehicle flowrate " ratio, and reflection gas station attracts clients the ability of oiling of entering the station,
The completion is than dynamically reflecting a moon index performance for " day sales volume/moon index " ratio.
5. gas station's Benefit Calculation according to claim 4 based on composite index analysis model, which is characterized in that
The step 2) includes:
Step 201) is standardized input data:
Since the unit of the different factors is different, so data are normalized, using general normalization formula:
Newvalue=(oldvalue-min)/(max-min)
Wherein Newvalue indicates dimensionless number after normalization, and max, min respectively indicate maximum value in this group of numerical value and most
Small value, oldvalue are raw value;Then data normalization is handled;
Step 202) does canonical correlation analysis to the data after standardization and carries out significance test, obtains related coefficient;
Related coefficient is respectively as follows: 0.039,0.546,0.560, three values be respectively people's rifle ratio ﹑ enter the station rate, complete the weight system of ratio
Number;
Step 203) constructs composite index performance analysis model;
Enable a=39;B=546;C=560 obtains the calculation formula of composite index:
A=a × P+b × Q+c × R
Wherein: A indicates composite index, and P indicates that people's rifle ratio, Q indicate the rate that enters the station, and R indicates that the ratio that hits the target day, above-mentioned formula are
Composite index performance analysis model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711204491.1A CN109858716A (en) | 2017-11-27 | 2017-11-27 | A kind of gas station's Benefit Calculation based on composite index performance analysis model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711204491.1A CN109858716A (en) | 2017-11-27 | 2017-11-27 | A kind of gas station's Benefit Calculation based on composite index performance analysis model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109858716A true CN109858716A (en) | 2019-06-07 |
Family
ID=66887504
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711204491.1A Pending CN109858716A (en) | 2017-11-27 | 2017-11-27 | A kind of gas station's Benefit Calculation based on composite index performance analysis model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109858716A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110570089A (en) * | 2019-08-09 | 2019-12-13 | 中国科学院南京地理与湖泊研究所 | construction method for evaluating river ecological condition by aquatic organism community multi-parameter index |
CN113148936A (en) * | 2021-04-02 | 2021-07-23 | 湖北新强信息技术工程有限公司 | Gas station equipment integration safety management system |
CN113219889A (en) * | 2021-04-02 | 2021-08-06 | 湖北新强信息技术工程有限公司 | Integrated operation management system of filling station equipment |
-
2017
- 2017-11-27 CN CN201711204491.1A patent/CN109858716A/en active Pending
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110570089A (en) * | 2019-08-09 | 2019-12-13 | 中国科学院南京地理与湖泊研究所 | construction method for evaluating river ecological condition by aquatic organism community multi-parameter index |
CN113148936A (en) * | 2021-04-02 | 2021-07-23 | 湖北新强信息技术工程有限公司 | Gas station equipment integration safety management system |
CN113219889A (en) * | 2021-04-02 | 2021-08-06 | 湖北新强信息技术工程有限公司 | Integrated operation management system of filling station equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Rico-Juan et al. | Machine learning with explainability or spatial hedonics tools? An analysis of the asking prices in the housing market in Alicante, Spain | |
TWI712981B (en) | Risk identification model training method, device and server | |
CN110837931B (en) | Customer churn prediction method, device and storage medium | |
US20210073283A1 (en) | Machine learning and prediction using graph communities | |
JP6034890B2 (en) | Specification, estimation, causal driver discovery and market response elasticity or lift coefficient automation | |
Rothaermel et al. | Balancing vertical integration and strategic outsourcing: effects on product portfolio, product success, and firm performance | |
JP4358475B2 (en) | Credit evaluation system | |
Carrillo et al. | Can tightness in the housing market help predict subsequent home price appreciation? Evidence from the United States and the Netherlands | |
Silva et al. | Cross country relations in European tourist arrivals | |
CN111784486A (en) | Construction method and device of business risk prediction model and computer equipment | |
CN109858716A (en) | A kind of gas station's Benefit Calculation based on composite index performance analysis model | |
CN112419030B (en) | Method, system and equipment for evaluating financial fraud risk | |
CN105303447A (en) | Method and device for carrying out credit rating through network information | |
EP2880601A2 (en) | Asymmetrical multilateral decision support system | |
CN115841345B (en) | Cross-border big data intelligent analysis method, system and storage medium | |
CN107844874A (en) | Enterprise operation problem analysis system and its method | |
Brzezicka et al. | The applicability of the speculative frame method for detecting disturbances on the real estate market: evidence from Poland | |
Weerathunga et al. | Firm-specific and contextual determinants of Sri Lankan corporate hotel performance | |
CN114548620A (en) | Logistics punctual insurance service recommendation method and device, computer equipment and storage medium | |
Sutter et al. | Measuring entrepreneurship and knowledge capital: Metropolitan economic efficiency in the USA | |
JP4358498B2 (en) | Credit evaluation system | |
Aroul et al. | The impact of brand affiliation on asset values: the case of UK hotels | |
KR102590232B1 (en) | System for providing day trading stock picking service using bigdata based artificial intelligence | |
US20230196485A1 (en) | After-repair value ("arv") estimator for real estate properties | |
Shang et al. | Estimation and Explanation of Time-Varying Weights in Chinese CPI |
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: 20190607 |
|
RJ01 | Rejection of invention patent application after publication |