CN105260815A - Electric power system payment channel evaluating method based on large data - Google Patents

Abstract

The invention provides an electric power system payment channel evaluating method based on large data and belongs to the field of electric power systems. The method comprises the steps of: firstly based on investigation questionnaire data, adopting a K-means clustering algorithm to carry out figure analysis on user payment behaviors, and identifying corresponding relations between various kinds of user attributes and the various kinds of user payment behaviors; utilizing a K nearest neighbor classification algorithm to carry out nearest neighbor clustering on electric power system payment channel evaluation factors, using a clustering result as a payment channel evaluation index, and establishing a payment channel evaluation system; utilizing a simulated genetic annealing algorithm to determine the weight of the payment channel evaluation index so as to form a payment channel evaluation module; and for new areas or new periods, utilizing electric power system practical payment data to judge whether the payment channel evaluation module is suitable, and if not, establishing a new evaluating module. The invention fills a technical void in the electric power system field, the quantitative evaluation of the electric power system payment channel is realized, and a scientific decision making basis are provided for the planning, construction and reconstruction of the payment channel.

Description

Based on the electric system payment channel evaluation method of large data

Technical field

The invention belongs to field of power, particularly a kind of payment of the electric system based on large data channel evaluation method.

Background technology

Along with the develop rapidly of sociometric technique, power consumption is sustainable growth always, Electricity customers quantity and electricity sales amount also continue significantly to improve, in order to serve users better, payment channel is always towards variation, the future development of diversification, as shown in Figure 1, but different payment channels is also existing drawback in varying degrees, such as outlet's layout is unreasonable, business window arranges dumb, human resources configuration is uneven, there is risk in tariff recovery, taking control agreement signs slow etc., how to utilize the quantitatively evaluating of large data realization to existing payment channel of Electric Power Marketing System, further for the pay the fees construction plan of channel and transformation provide scientific basis, still belong to technological gap.

Summary of the invention

For existing technological gap, the present invention proposes a kind of electric system based on large data payment channel evaluation method, concrete steps are as follows:

Step 1, obtain user property and subscriber payment behavioral data based on questionnaire, adopt K-means clustering algorithm to set up subscriber payment behavior portrait: namely to determine the corresponding relation between the classification of several user properties and several subscriber payment behaviour classification.

Step 1.1, by inquiry questionnaire obtain UAD and subscriber payment behavioral data, UAD comprises sex, age, occupation, schooling, subscriber payment behavioral data comprises pay charge way, payment frequency, cancel own business hall after payment channel, the user that may adopt think the payment channel of most convenient;

Step 1.2, exchange interrogate involve in that row is integrated, cleaning, change, reduction, obtain effective questionnaire, utilize SPSS statistical analysis software to form the sample space of user property and subscriber payment behavior;

Step 1.3, utilize K-means clustering algorithm to carry out cluster analysis to user property, obtain some class user properties;

Step 1.4, utilize K-means clustering algorithm to carry out cluster analysis to subscriber payment behavior, obtain the behavior of some class subscriber payments;

Step 1.5, K-means clustering algorithm is utilized to carry out cluster analysis respectively to a class user property and the behavior of some class subscriber payments, obtain the cluster result that several close degree are different, the subscriber payment behavior classification that the cluster result that wherein close degree is maximum is corresponding, be the paying behaviors that such user is inclined to employing most, draw the corresponding relation of such user property and such subscriber payment behavior, namely complete the subscriber payment behavior portrait of such user property, complete the subscriber payment behavior portrait of other class user properties by that analogy.

Step 2, K arest neighbors sorting algorithm is utilized to be close to cluster most to electric system payment channel factor of evaluation, be a class by affecting on payment channel that similar factor of evaluation gathers, polymerization result is as electric system payment channel evaluation index, and all kinds of evaluation index forms payment channel assessment indicator system.

Step 2.1, electric system payment channel factor of evaluation comprise coverage rate, the utilization factor of channel, cost, work efficiency, subscriber payment behavior portrait, convenience, customer satisfaction, the channel development trend of channel, the subscriber payment behavior portrait wherein step 1 obtained is as the initial value of subscriber payment behavior portrait, and the initial value of all the other each factors is determined according to expertise value;

Step 2.2, K arest neighbors sorting algorithm is utilized to be close to cluster most to electric system payment channel factor of evaluation, be a class by affecting on payment channel that similar factor of evaluation gathers, polymerization result is as electric system payment channel evaluation index, and all kinds of evaluation index forms payment channel assessment indicator system.

The weight of each evaluation index in step 3, employing simulation genetic annealing algorithms determination assessment indicator system, set up payment channel evaluation model, obtain the evaluation of estimate of each payment channel, described payment channel evaluation model is the product first obtaining every evaluation index and weight, then ask each product and.

Step 4, utilize electric system actual payment data, judge whether payment channel evaluation model is applicable to new area or new period, if be suitable for, channel evaluation model of then paying the fees continues to use, otherwise adopt simulation genetic annealing algorithms to redefine the weight of each evaluation index, determine the payment channel evaluation model being applicable to new area or new period.

Step 4.1, utilize electric system actual payment data acquisition current power amount, by new area or new electric system in period actual payment historical data broken line graph determination electricity demand forecasting value;

Step 4.2, the proportionate relationship utilizing power consumption and pay the fees between channel evaluation of estimate, obtain the payment channel evaluation of estimate that electricity demand forecasting value is corresponding;

If the difference of the payment channel evaluation of estimate that payment channel evaluation of estimate corresponding to step 4.3 electricity demand forecasting value is corresponding with current power amount is less than 0.1, then continue to use current payment channel evaluation model, otherwise adopt simulation genetic annealing algorithms to redefine the weight of each evaluation index, return step 3 and set up new payment channel evaluation model.

Beneficial effect:

The present invention proposes a kind of electric system based on large data payment channel evaluation method, first to pay dues behavioral data based on the user property of questionnaire and user, adopt K-means clustering algorithm to carry out portrait analysis to subscriber payment behavior, identify the corresponding relation of all types of user attribute and all types of user paying behaviors.Utilize K arest neighbors sorting algorithm (KNN algorithm) to be close to cluster most to electric system payment channel factor of evaluation, polymerization result, as payment channel evaluation index, sets up payment channel appraisement system.Utilize simulation genetic annealing algorithms to determine the weight of payment channel evaluation index, thus form payment channel evaluation model.For new area or new period, whether payment channel evaluation model is suitable for, if inapplicable, sets up new payment channel evaluation model to utilize electric system actual payment data to judge.The present invention has filled up a technological gap of field of power, achieves the quantitatively evaluating to electric system payment channel, for payment channel planning construction and transformation provide the decision-making foundation of science.

Accompanying drawing explanation

Fig. 1 is the existing payment channel of electric system in an embodiment of the present invention;

Fig. 2 is that employing K-means clustering algorithm in an embodiment of the present invention is set up user and drawn a portrait process flow diagram;

Fig. 3 is KNN in an embodiment of the present invention and simulated annealing process flow diagram;

Fig. 4 is the payment channel evaluation method process flow diagram based on large data in an embodiment of the present invention.

Embodiment

Below in conjunction with accompanying drawing, the specific embodiment of the invention is elaborated.The concrete implementation step of electric system payment channel evaluation method based on large data is as follows, as shown in Figure 4:

Step 1, obtain user property and subscriber payment behavioral data based on questionnaire, adopt K-means clustering algorithm to set up subscriber payment behavior portrait: namely determine the classification of several user properties and several paying behaviors classify between corresponding relation, as shown in Figure 2.

Step 1.1, by inquiry questionnaire obtain UAD and subscriber payment behavioral data, UAD comprises sex, age, occupation, schooling, subscriber payment behavioral data comprises pay charge way, payment frequency, cancel own business hall after payment channel, the user that may adopt think the payment channel of most convenient.

Step 1.2, exchange interrogate involve in that row is integrated, cleaning, change, reduction, obtains 6000 parts of effective questionnaires, utilizes SPSS statistical analysis software formation user property and the sample space of subscriber payment behavior.

Step 1.3, utilize K-means clustering algorithm to carry out cluster analysis to user property, obtain some class user properties.

Step 1.3.1, using the user property in questionnaire and subscriber payment behavior as the first sample space, from the data set U comprising n sample, select k sample as the initial center of k class, be designated as in order to improve precision and the speed of cluster, be no longer random selecting to choosing of initial cluster center, but be optimized and choose.

Step 1.3.1.1, calculate the spacing of sample between two in data set U, these distances are formed dissimilarity matrix D as element.

Step 1.3.1.2, for the sample x in data set U, take x as the centre of sphere, formed a spherical territory with a certain positive number r for radius, in this spheroid, comprise the density of number as x of sample.According to Dens (x)=| { p|Dist (x, p)≤r, p ∈ U|}| calculates the density D ens (x of each sample _i), i=1,2 ... N, sample density being greater than average density value is placed in S set, and p is any one sample in data set U, and Dist (x, p) is the distance of sample x and sample p.

Step 1.3.1.3, choose the maximum sample x of S set Midst density as first initial cluster center point i.e. Dens (x)=Max{Dens (x) x ∈ S}, and will be put in initial cluster center point set M.

Step 1.3.1.4, find next initial cluster center the following condition of demand fulfillment:

① $Dist(x,m_1^{\left(0\right)})>\mathrm{\β};$

② $m_2^{\left(0\right)}=Max\left\{Dens\left(x\right)\right|x\∈S\backslash M\}.$

Wherein, when β rounds the 1/k of the first sample space (k is the number of cluster), initial cluster center is chosen relatively more accurate, can avoid choosing the close point of same class Midst density.

Step 1.3.1.5, by point add initial cluster center point set M, namely

Step 1.3.1.6, repetition step 1.3.1.4, step 1.3.1.5, until find out k initial cluster center.

Step 1.3.2, the data sample x in data set U to be sorted is allocated to a certain class in k class according to minimal distance principle one by one j ∈ 1,2 ... k}; Wherein t represents iterations, if distance then x assigns to class in, l ∈ 1,2 ..., k}.

Each cluster centre after step 1.3.3, calculating reclassify:

$m_j^{(t+1)}=\frac{1}{n_j^{(t+1)}}\underset{x\∈C_j^{(t+1)}}{\Σ}x,j=1,2,\mathrm{...},k$

In formula, for class the number of middle data sample.Using the mean value of data sample as new cluster centre, square error criterion function can be made minimum, wherein m _ithe mean value of all samples, C _iit is the i-th class.

If step 1.3.4 j=1,2 ..., k, then algorithm terminates; Otherwise, if j ∈ 1,2 ..., k}, then return step 1.3.2, iterates, the cluster analysis of completing user attribute.

Step 1.4, utilize K-means clustering algorithm to carry out cluster analysis to subscriber payment behavior, obtain the behavior of some class subscriber payments, concrete implementation step is as shown in step 1.3.

Step 1.5, K-means clustering algorithm is utilized to carry out cluster analysis respectively to a class user property and the behavior of some class subscriber payments, obtain the cluster result that several close degree are different, the subscriber payment behavior classification that the cluster result that wherein close degree is maximum is corresponding, be the paying behaviors classification that such user is inclined to employing most, draw the corresponding relation of such user property and such subscriber payment behavior, namely the subscriber payment behavior portrait of such user property is completed, complete the subscriber payment behavior portrait of other class user properties by that analogy, concrete implementation step is as shown in step 1.3.

Step 2, K arest neighbors sorting algorithm is utilized to be close to cluster most to electric system payment channel factor of evaluation, be a class by affecting on payment channel that similar factor of evaluation gathers, polymerization result is as electric system payment channel evaluation index, all kinds of evaluation index forms payment channel assessment indicator system, as shown in Figure 3.

Step 2.1, electric system payment channel factor of evaluation comprise coverage rate, the utilization factor of channel, cost, work efficiency, subscriber payment behavior portrait, convenience, customer satisfaction, the channel development trend of channel, the subscriber payment behavior portrait wherein step 1 obtained is as the initial value of subscriber payment behavior portrait, and the initial value of all the other each factors is determined according to expertise value.

The coverage rate of channel: refer to certain channel on-hand quantity and account for the ratio of all payment channels in region.

The utilization factor of channel: refer to the payment stroke count of existing channel, turnover and turnover.

Cost: comprise construction cost, operating cost and maintenance cost.

Work efficiency: refer to that the averaging time of every fee payment service handled by different payment channel, wherein need to consider the queuing situation of entity business hall, the working condition etc. being deferred to account situation and other financial channels of Alipay.

Subscriber payment behavior is drawn a portrait: i.e. subscriber payment preference, draws from step 1.

Convenience: when referring to subscriber payment, time used and the electricity charge are to the number of account time.

Customer satisfaction: refer to the satisfaction of user to a certain payment channel.

Channel development trend: the situation of change of the utilization factor of a certain channel of primary concern, becomes better and better or worse and worse.

Step 2.2, K arest neighbors sorting algorithm is utilized to be close to cluster most to electric system payment channel factor of evaluation, be a class by affecting on payment channel that similar factor of evaluation gathers, polymerization result is as electric system payment channel evaluation index, and all kinds of evaluation index forms payment channel assessment indicator system.

Step 2.2.1, using all factors of evaluation that will analyze as the second sample space R, the data of first random selecting 10% are as initial training collection: X={x ₁, x ₂..., x _n, x _i∈ R, corresponding class is Y={y ₁, y ₂..., y _m, m≤n, all the other 90% data are as test set.

Step 2.2.2, establish x _ifrom y _ii-th training sample of class, N _yithe training sample number of every class, O ^yiy _ithe central sample of class, is calculated by following formula:

Step 2.2.3, definition y _ieach sample x in class _ito central sample O ^yidistance be: d (x _i, O ^yi)=|| x _i-O ^yi||, if d is (x _i, O ^yi)≤ε (wherein ε is the maximum deflection difference value set in advance), then by x _icluster is to y _iin class; Otherwise, by x _iwith other non-y _icentral sample in class is carried out distance and is calculated, and repeats step 2.2.2, step 2.2.3 until by complete for equal for all indexs cluster, obtains payment channel evaluation index.

Step 2.2.4, in order to verify that certain single sample is gathered the rationality of certain class, evaluate with individual silhouette coefficient S, S larger expression cluster is more reasonable, and the expression formula of silhouette coefficient S is:

Wherein individual silhouette coefficient:

$S_i=\frac{b\left(i\right)-a\left(i\right)}{max\[a\left(i\right),b\left(i\right)\]}$

$a\left(i\right)=\frac{1}{n_{y_i}}\underset{i,j\∈y_i}{\Σ}d(i,j)$

$b\left(i\right)={\min }_{q,q\∉y_i}\[\frac{1}{n_{yi}}\underset{i\∈y_i,j\∈y_i}{\Σ}d(i,j)\]$

N represents total sample number, supposes sample x _igathered y _iclass, a (i) represents sample x _iwith belong to y _imean distance between other all samples of class, b (i) represents sample x _iwith non-y _ithe minimum value of the mean distance of all samples in each class of class, max () and min () represents max function and minimum value function, n respectively _yirepresent y _ithe number of class, q represents and does not belong to y _ithe sample of class.

The weight of each evaluation index in step 3, employing simulation genetic annealing algorithms determination assessment indicator system, set up payment channel evaluation model, obtain the evaluation of estimate of each payment channel, described payment channel evaluation model is the product first obtaining every evaluation index and weight, then ask each product and.

Step 3.1, initialization: according to the parameter variation range of each index in tentatively given 3rd sample space of expertise value, Stochastic choice initial weight value m within the scope of this ₀, and calculate corresponding target function value E (m ₀).

Step 3.2, disturbance is carried out to present weight value, produce a new value m, calculate corresponding target function value E (m), obtain Δ E=E (m)-E (m ₀).

If step 3.3 Δ E < 0, then new weight m is accepted; If Δ E > 0, then new weight m presses probability accept, T is the given time.When weight is accepted, put m ₀=m, E (m ₀)=E (m).

Step 3.4, in given time T under, repeat disturbance and the receiving course of certain number of times, namely repeat step 3.2, step 3.3.

Step 3.5, slowly reduction T value preset time.

Step 3.6, repetition step 3.2-step 3.5, till the condition of convergence meets, finally obtain optimum weight parameter, sets up payment channel evaluation model y=a ₁x ₁+ a ₂x ₂+ ... + a _nx _n, wherein x _irepresent the evaluation index in the appraisement system formed after KNN algorithm cluster, a _irepresent the weight of corresponding corresponding index, be through the optimal value repeatedly simulated genetic annealing algorithms iteration and obtain, y represents the evaluation of estimate that a certain channel is final.For different regions, different decision maker and different demands, x _i, a _ican be different.

Step 4, utilize electric system actual payment data, judge whether payment channel evaluation model is applicable to new area or new period, if be suitable for, channel evaluation model of then paying the fees continues to use, otherwise adopt simulation genetic annealing algorithms to redefine the weight of each evaluation index, determine the payment channel evaluation model being applicable to new area or new period.

Step 4.1, utilize electric system actual payment data acquisition current power amount, by new area or new electric system in period actual payment historical data broken line graph determination electricity demand forecasting value.

Step 4.2, the proportionate relationship utilizing power consumption and pay the fees between channel evaluation of estimate, obtain the payment channel evaluation of estimate that electricity demand forecasting value is corresponding.

Proportionate relationship between power consumption and payment channel evaluation of estimate is:

If the difference of the payment channel evaluation of estimate that payment channel evaluation of estimate corresponding to step 4.3 electricity demand forecasting value is corresponding with current power amount is less than 0.1, then continue to use current payment channel evaluation model, otherwise adopt simulation genetic annealing algorithms to redefine the weight of each evaluation index, return step 3 and set up new payment channel evaluation model.

Claims (4)

1., based on the electric system payment channel evaluation method of large data, it is characterized in that: comprise the following steps:

Step 1, obtain user property and subscriber payment behavioral data based on questionnaire, adopt K-means clustering algorithm to set up subscriber payment behavior portrait: namely to determine the corresponding relation between some class user properties and the behavior of some class subscriber payments;

Step 2, K arest neighbors sorting algorithm is utilized to be close to cluster most to electric system payment channel factor of evaluation, be a class by affecting on payment channel that similar factor of evaluation gathers, polymerization result is as electric system payment channel evaluation index, and all kinds of evaluation index forms payment channel assessment indicator system;

The weight of each evaluation index in step 3, employing simulation genetic annealing algorithms determination assessment indicator system, set up payment channel evaluation model, obtain the evaluation of estimate of each payment channel, described payment channel evaluation model is the product first obtaining every evaluation index and weight, then ask each product and;

Step 4, utilize electric system actual payment data, judge whether payment channel evaluation model is applicable to new area or new period, if be suitable for, channel evaluation model of then paying the fees continues to use, otherwise adopt simulation genetic annealing algorithms to redefine the weight of each evaluation index, determine the payment channel evaluation model being applicable to new area or new period.

2., according to claim 1 based on the electric system payment channel evaluation method of large data, it is characterized in that: described step 1 concrete steps are as follows:

Step 1.1, by inquiry questionnaire obtain UAD and subscriber payment behavioral data, UAD comprises sex, age, occupation, schooling, subscriber payment behavioral data comprises pay charge way, payment frequency, cancel own business hall after payment channel, the user that may adopt think the payment channel of most convenient;

Step 1.2, exchange interrogate involve in that row is integrated, cleaning, change, reduction, obtain effective questionnaire, utilize SPSS statistical analysis software to form the sample space of user property and subscriber payment behavior;

Step 1.3, utilize K-means clustering algorithm to carry out cluster analysis to user property, obtain some class user properties;

Step 1.4, utilize K-means clustering algorithm to carry out cluster analysis to subscriber payment behavior, obtain the behavior of some class subscriber payments;

Step 1.5, K-means clustering algorithm is utilized to carry out cluster analysis respectively to a class user property and the behavior of some class subscriber payments, obtain the cluster result that several close degree are different, the subscriber payment behavior classification that the cluster result that wherein close degree is maximum is corresponding, be the paying behaviors that such user is inclined to employing most, draw the corresponding relation of such user property and such subscriber payment behavior, namely complete the subscriber payment behavior portrait of such user property, complete the subscriber payment behavior portrait of other class user properties by that analogy.

3., according to claim 1 based on the electric system payment channel evaluation method of large data, it is characterized in that: described step 2 concrete steps are as follows:

Step 2.1, electric system payment channel factor of evaluation comprise coverage rate, the utilization factor of channel, cost, work efficiency, subscriber payment behavior portrait, convenience, customer satisfaction, the channel development trend of channel, the subscriber payment behavior portrait wherein step 1 obtained is as the initial value of subscriber payment behavior portrait, and the initial value of all the other each factors is determined according to expertise value;

Step 2.2, K arest neighbors sorting algorithm is utilized to be close to cluster most to electric system payment channel factor of evaluation, be a class by affecting on payment channel that similar factor of evaluation gathers, polymerization result is as electric system payment channel evaluation index, and all kinds of evaluation index forms payment channel assessment indicator system.

4., according to claim 1 based on the electric system payment channel evaluation method of large data, it is characterized in that: described step 4 concrete steps are as follows:

Step 4.1, utilize electric system actual payment data acquisition current power amount, by new area or new electric system in period actual payment historical data broken line graph determination electricity demand forecasting value;

Step 4.2, the proportionate relationship utilizing power consumption and pay the fees between channel evaluation of estimate, obtain the payment channel evaluation of estimate that electricity demand forecasting value is corresponding;

If the difference of the payment channel evaluation of estimate that payment channel evaluation of estimate corresponding to step 4.3 electricity demand forecasting value is corresponding with current power amount is less than 0.1, then continue to use current payment channel evaluation model, otherwise adopt simulation genetic annealing algorithms to redefine the weight of each evaluation index, return step 3 and set up new payment channel evaluation model.