mpar-cluster: applied algorithm of geo-selection for ... in an automated way. ... powerful framework...

Journal of Geodesy and Geomatics Engineering 1 (2016) 1-9 doi: 10.17265/2332-8223/2016.01.001

Mpar-Cluster: Applied Algorithm of Geo-Selection for

Optimization of the Credit Recovery of Electricity Supply

Augusto César da Silva Machado Copque and Mateus Prates de Andrade Rodrigues

Universidade Católica do SalvadorandCompanhia de Eletricidade do Estado da Bahia, Brazil

Abstract: This paper presents a study of optimization of operational recovery credit default with geoprocessing use through geoprocessing tools, developed in the Receivables Management Companhia de Eletricidade do Estado da Bahia-COELBA sector. The work was initially based on the application of Data Mining Tools for Software KNIME 2.9 and later use of the tool of GIS-ArcGIS 10.X/ESRI .Were evaluated and applied analytical processing algorithms, to improve the process of spatial selection and define the best sets logistical credit recovery. The focus study, based on geoprocessing use in cutting action is due to the fact that this process has the largest collection efficiency. It is understood that the efficiency of the cutting action, should the great importance that electricity has on modern life. The research was guided its evolution from analysis of algorithms agglutination and georeferenced database, whose focus was and is acting in the cutting action due to the fact that this process has the largest collection efficiency. For the implementation of the study, through some geoprocessing techniques, the Mpar-cluster, this optimization model was developed from a custom algorithm for spatial selection, that had with subsidy: information stored in geographic databases, database information alphanumeric, images (aerial photographs and satellite images), text files, and digital tables.

Key words: Geoprocessing, recovery credit, mpar-cluster.

1. Introduction

This article presents a study of operational and

financial optimization of the supply of electricity

cutting action from analysis of agglutination

algorithms and basic georeferenced data. For the

application of optimization modeling proposals were

used data mining platforms and Geography

Information System (GIS) or Geographic Information

Systems (GIS).

The focus of this study, based on geotechnology1

Corresponding author: Augusto César da S. M. Copque,

Geographer, master of urban environmental engineering, management expert planning and Professor at the Catholic University of Salvador-UCSAL, research fields: geoprocessing, geointeligence and geoscience. 1 Geprocessing/Geotechnology is a GIS operation used to manipulate spatial data. A typical geoprocessing operation takes an inputdataset, performs an operation on that dataset, and returns the result of the operation as an output dataset. Common geoprocessing operations include geographic feature overlay, feature selection and analysis, topologyprocessing, rasterprocessing, and data conversion. Geoprocessing allows for definition, management, and analysis of information used to form decisions.

[5, 6] use in the cutting action is due to the fact that

this process has the highest collection efficiency. It is

understood that the cutting action of efficiency, it

should be the great importance that the electricity has

in modern life, with the electricity directly related to

physiological items of the base of Maslow’s needs

pyramid [2].

However, the traditional model of disconnection of

the power supply is carried out by a specialized

workforce, the more expansive significantly the action.

Geoprocessing is a framework and set of tools for processing

geographic and related data. The large suite of geoprocessing tools can be used to perform spatial analysis or manage GIS data in an automated way. A typical geoprocessing tool performs an operation on a dataset (such as a feature class, raster, or table) and creates a resulting output dataset. For example, the Buffer tool takes a map layer as input, creates areas around the layer’s features to a specified distance, and writes those areas to a new output layer.

In addition to the suite of tools, geoprocessing also has a powerful framework that supports control of the processing environment and allows you to build custom tools that can further automate your work. You can use the geoprocessing tools included in ArcGIS as building blocks to create an infinite number of custom tools that automate repetitive tasks or solve complex problems.

D DAVID PUBLISHING

2

Thus, It has

an optimiza

operating co

For the

algorithms w

and Equa

optimization

selection alg

been “bapti

amount per

máximo de

map-cluster

in a GIS

information

of alphanu

photographs

digital files.

Thus, thi

better earn

operational

through the

the geo-refe

of state. The

of tools for D

Fig. 1 Estru

2 Algorithm management. collection cutoidentification problems in po

Mpar

been the cut

ation study,

osts action, pr

developmen

were applied

k-means[3

n model from

gorithm was d

ized” as ma

r radius-clus

seleçãoporra

were used so

environmen

stored in geo

umeric inf

s and satellit

is study dem

nings by re

processes i

use of geogr

erenced locati

e work was ini

Data Mining S

utura de um SI

developed b

The sea-clusteroff scores. Thecustomers suscower distributio

r-Cluster: Ap

tting action a

where an ef

rovides scalab

nt the stud

in Data Min

]. Posterior

m a custom

developed. Th

ap-cluster or

ster (aglome

aio)2. For th

ome geotechn

nt (Fig. 1)

ographical da

formation,

te images) t

monstrated th

eviewing the

in the orga

raphical know

ion of the cli

itially based o

Software KNI

IG. Fonte Med

y the author is used for geoe algorithm canceptible to newon

plied Algorith Credit Reco

a great setting

fficient and h

ble gains.

dy optimiza

ning as: K-me

rly, a sec

mized geogra

his algorithm

maximum

erado de v

he application

nology techniq

, with sub

atabases, datab

images (a

texts, tables

he possibility

e collection

anization stu

wledge, based

ents in the B

on the applica

IME 2.9 and l

deiros 2012.

rs to receivospatial selection also be applie

w debts, negated

hm of Geo-Seovery of Elect

g for

high

ation

eans

cond

aphic

m has

pick

valor

n of

ques

bsidy

base

erial

and

y of

n of

udied

d on

Bahia

ation

later

ables on of ed to d and

the

eva

to i

best

focu

effi

[4].

2. M

2.1

In

(CO

reco

Rec

com

plan

Me

(UR

adm

The

man

Com

The

pro

Res

T

cutt

of s

the

susc

Cur

prop

Col

by

3 ThlargcusttermConthe Bmuninhacomresid

election for Otricity Supply

use of GIS

aluated and ap

mprove the s

t logistics set

uses on demo

icient strategi

Materials a

Structure

n Companhi

OELBA)3, th

overy is plan

ceivables — C

mbat activitie

nning action

ssage Servic

RA), negativ

ministrative, b

e manageme

naged throug

mpany CCS

ese rules are

files and fol

solution No. 4

The focus of

ting action of

susceptibility

unit being

ceptible to p

rrently the l

prietary syst

llection (ICC

SAP CCS, he Companhia dest electricity tomers and the

ms, ranks first ntrolled by NeoeBrazilian electr

nicipalities in abitants in a co

mpany has ovedential custome

Optimization oy

S tools — A

pplied analyti

spatial selecti

s of credit rec

onstrate the r

ic and financ

and Method

a de Eletricid

he operationa

nned by the

CRGR that us

s, the expecte

ns such as

ce (SMS),

vity, power

billing adviso

nt of flow

gh the Custom

SAP, with “r

e parameteriz

llowing the

414/2010 of A

f this study

f electric energ

y to activity i

provided a

power supply

list of susce

tem called I

C). This syste

complementi

de Eletricidade distributor in

sixth in power among the N

energia Group, ricity sector, Co

Bahia, servinoncession areaer 5.6 millioners.

of the

ArcGIS 10.x/

ical processin

ion process an

covery. The s

results of ope

cial charges o

ds

dade do Esta

al process of

Management

ses different n

ed date of ele

warning no

Response U

cut, cut tr

ory and judici

collection o

mer Success S

rules” specifi

zed for diff

regulatory g

ANEEL [1].

was directe

gy. The regul

is managed b

daily list o

y suspension

eptible is h

Information

em treats the

ing financial do Estado da B

the country supplied volum

North dealers one of the larg

oelba is presentng more tha

a of 563,000 kn customers,

/ESRI. Were

ng algorithms

nd define the

tudy analysis

erations more

on portfolios

ado da Bahia

default loan

t Unit of the

non-payment

ectric energy,

otices, Short

Unit audible

racking, low

ial collection.

of default is

Story tool —

fic collection.

ferent clients

guidelines of

ed to supply

latory control

by SAP CSS,

of customers

action (cut).

andled in a

Control and

list provided

information

ahia is the thirdin number of

me. These same— Northeast.

gest investors int in 415 of 417n 14 million

km2. Today the88% of these

e

s

e

s

e

s

a

n

e

t

,

t

e

w

.

s

—

.

s

f

y

l

,

s

.

a

d

d

n

d f e . n 7 n e e

Mpar-Cluster: Applied Algorithm of Geo-Selection for Optimization of the Credit Recovery of Electricity Supply

3

and automatically applies the logistical limitations of

availability of Providers Outsourced Business Services

— EPS.

For the implementation of the concept of limitation

contractual logistics of EPS, the concession area

COELBA is divided into 43 centers of work. ICC

system (Fig. 2), these 43 work centers are divided into

target zones, these zones represent sets of

neighborhoods in the municipal seat or group of

municipalities — in the least populated areas. The

zones are further subdivided into cluster, which

represents sets of neighborhoods neighbors. These

definitions of zones and clusters are parameterized

empirically, based on the field knowledge of

operational coordinators of the respective work

centers.

Fig. 2 ICC. Source: COELBA, 2014.

The algorithm used by the ICC for selecting serial

succinctly defined by the following steps:

(1) Select a particular zone Work Center — CT to set

the selection;

(2) It makes up the summarization of the number of

clients per group;

(3) Sort up zones of decreasing order by the charged

total;

(4) Select first zones between the amount of service,

depending on the logistical limitation;

(5) Repeats 1 through 4 the process to be performed

to check banknotes for all areas of all TCs limited to

daily target set for each zone;

The selected amount is increased by a percentage of

15% compared to the target area in order to facilitate

the composition of sets of notes from classes by EPS.

In addition, it adds up over 10% from the percentage of

notes identified as paid during field service.

As an example, stands out in Table 1 the selection of

the groups 22, 23, 1 and 2 of the sample under study

referring to the day 05/20/2014. In Table 1, were

summarized customers likely to cut according to the

groups being classified the clusters in descending order

of sum of debt collect. In this example are highlighted

1007 client where the program selected 1,000

customers, representing: 800 customers regarding the

4

cutting targe

paid account

This selec

generates th

the cutoff sc

experience

dispatches t

dispatched b

subsequently

In Fig. 3

distribution

Earth softwa

classes on 0

of blue show

susceptible

perceives in

representing

concentratio

In a close

overlap of ac

dispatch no

displacemen

Analyzing

improvemen

the debt sta

greater conc

this universe

for optimiza

of statistica

algorithms,

algorithm, im

Table 1 Th

Praça da Sé. Grouping

22

23

1

2

16

13

12

7

Mpar

et this area of

t and 15% of

ction of the IC

he service and

core for its E

brings toget

to their field

by EPS are

y made availa

3, there is

of the notes b

are, relative th

05/20/2014. T

w the orders

portfolio sel

n the sameFig

g, respective

ons of debts.

er view of th

ctivities of the

ote orders t

nt in group 2,

g the curre

nt points from

ains with sm

centration and

e of study, w

ation and debt

l tools for D

and anothe

mplemented i

e first lines of

Deb

R$ 47'3

R$ 42'0

R$ 39'3

R$ 32'5

R$ 29'2

R$ 28'2

R$ 27'6

R$ 25'7

r-Cluster: Ap

f the work cen

f routing.

CC is directe

d provides th

EPS. In EPS,

ther the note

d technicians

handled by

able to a new

an example

being display

he generation

The icons (ba

per group, c

lection proce

g. 3 some stai

ely, the sma

he EPS of rou

e groups, as w

hat impact

see Fig. 4.

ent model

m a selection:

maller radius

d better order

were evaluate

t assemblage

Data mining

r by the m

in ArcGIS/ES

f the zone grou

Abt

R$314.43

R$042.14

R$326.92

R$577.55

R$218.35

R$277.90

R$669.40

R$705.06


nter, plus the

ed SAP CCS,

he following

a staff with f

es by class

s. The notes

y SAP CCS

selection of I

e of the sp

yed on the Go

n and dispatch

alloons) in sh

coming from

ess. In addit

ins yellow to

aller the hig

uting, there i

well as an erro

on an incr

selection, o

more in line w

of action, w

r per group. F

ed two propo

, one for anal

using cluste

mpar-cluster

SRI 10.

upings 1 do C

QTAverage

33$ 142.51

23$ 178.90

21$ 185.50

22$ 142.88

13$ 210.20

12$ 235.65

16$ 172.93

15$ 163.73


10%

that

day

field

and

not

and

ICC.

atial

ogle

h the

ades

m the

tion,

red,

gher

is an

or of

rease

ccur

with

with

From

osals

lysis

ering

own

CT of

TD

32

35

12

28

39

20

60

57

Fig.debt

Fig.

cuto

2.2

2

T

part

(kn

kno

data

199

F

in th

sou

ease

bloc

algo


. 3 Satellite imt stain. Prepar

. 4 Overlappi

off scores for g

Analysis of S

2.2.1 Optimiz

The term data

t of Knowle

owledge disc

owledge disco

a (Fayyad, Pia

96).

For the study

he free versio

urce). The cho

e of use, wi

cks drag an

orithms, there

Optimization oy

mage containinred by the auth

ing classes and

group 2. Prepar

Selection of O

zation in Data

a mining (min

edge Discove

covery in da

overy of met

atetsky-Shap

in question w

on No 2.9 of th

oice of the sof

ith an exten

nd drop type

e is the K-m

of the

ng the selectionhors, 2014.

d error in the d

red by the aut

Optimization P

a Mining

ning and min

ery in Datab

atabases) tha

thods in large

iro, Smyth &

was used KNI

he open sourc

ftware in ques

nsive pre-inst

e. Among t

eans. An agg

n of notes and

distribution of

hors, 2014.

Proposals

ing data) is a

bases (KDD)

at deals with

e amounts of

Uthurusamy

IME software

ce type (open

stion is due to

talled library

the available

glutination or

d

f

a

)

h

f

y,

e

n

o

y

e

r

clustering

observations

to the closes

a division o

type of sp

segments fro

it can be no

colors, wher

performed o

points are as

between ind

The geog

previously

South Amer

1969 in d

normalize la

to the large v

the draft de

coordinates,

action was

distance calc

Even with

shaft, it can

by the distan

by this value

In a secon

the size debt

dimensions

Fig. 5 GraphSource: dartm

Mpar

algorithm t

s “k” sets wh

st cluster aver

of the data sp

patial decom

om distance o

oted “n” poi

re for definin

one distance c

ssigned a col

dividuals.

graphic coord

treated in re

rica — Sout

decimal degr

atitude dimen

variation in th

ecimal variati

when conve

taken to mi

culation for th

h the normal

be seen in th

nce of debt, g

e due.

nd stage of th

t and applied

latitude and l

hical representmouth.edu, 201

r-Cluster: Ap

that aims

here each obs

rage, see Fig.

pace on a Vo

mposition wh

of relational a

ints being di

ng the color o

calculation b

lor according

dinates used

egional geod

th American

rees. Was

nsions longitu

he data of the

ion of latitud

erted to decim

itigate agglut

he dimension

lization of th

he studies a gr

getting more

he study, it w

the K-means

longitude. Du

tation of the re12.


to break

servation belo

5. This resul

oronoi diagram

here the sam

analysis. In Fi

ivided into t

of each point

between the o

to the proxim

in KNIME w

detic system

Datum — S

made neces

ude and debt

e latter comp

de and longi

mal degrees.

tination trend

n debt.

he values of

rouping tende

clusters grou

was removed f

the algorithm

uring the stud

esult of a cluste


“n”

ongs

lts in

m, a

mple

ig. 5

three

was

other

mity

were

for

SAD

ssary

due

ared

tude

This

d of

debt

ency

uped

from

m for

dy it

ering.

was

link

is n

lim

F

algo

indi

imp

Pier

orie

use

K-m

dist

algo

lim

in l

long

the

will

2

In

Dat

solu

algo

clie

of

high

T

nec

each

pos

of s

CT

foll

rega

T

mad

Eng

(RO

(CO

app


s ratified a

ked to the dev

not possible

its on the num

From that p

orithms that

ividuals by h

plemented a

rre-David, ti

ented program

s the concept

means, calcu

tance, but l

orithm for c

ited to two d

line with the

gitude dimen

algorithm EK

l be shown in

2.2.2 Geotech

n addition to

ta Mining, w

ution focusing

orithm develo

ent more limi

a group acti

her note valu

The use of GI

essary due th

h customer or

t. The locatio

susceptible to

’s were fund

low and effici

ard to higher

The implemen

de with the

gineer, postg

ODRIGUES,

OPQUE, A.).

plication of

Optimization oy

limitation of

velopment of

to create cl

mber of subje

oint, we tri

t allow to

himself. Thu

new mining

itled EKmea

mming langua

ts implemente

ulated accor

limiting the

clustering. Th

dimensions, w

e idea of usi

nsions for ag

Kmeans found

n the compara

hnology Optim

the challengi

we sought to d

g on geotechn

oped and has

ted debt amo

ion and enh

ues to the limi

IS through Ar

he search of ge

r the proximit

on of the custo

o charging by

damental to

iency in the c

values.

ntation of the

e partnership

graduate in

M.) and Geo

Thus, the use

f mpar-clust

of the

f the k-mean

f the same fun

lustering with

ects per group

ied to searc

define the

us, it was id

algorithm d

ans. This too

age object —

ed in heuristic

rding to th

number of

his Java app

which was alr

ing only the

gglutination.

d interesting r

ative topic bet

mization

ing modeling

develop and

nology tool. T

s had a focus

ount to a max

ance the clu

it of notes per

rcGIS/ESRI s

eographical p

ty of this with

omers and the

zone or neigh

define the b

collection, es

e designed al

p between a

Business

ographer, spec

e of GIS tool

ter algorith

5

ns algorithm

nctionality; it

h predefined

p.

ch clustering

amount of

dentified and

developed by

ol developed

— Java, which

cs expression

e Euclidean

f individuals

plication was

ready studied

latitude and

By applying

results which

tween models

developed in

implement a

The proposed

: identify the

ximum radius

ustering with

r group.

software was

positioning of

h the adjacent

e relationship

hborhoods of

best route to

pecially with

lgorithm was

an Electrical

Management

cialist in GIS

s enabled the

hm in the

5

m

t

d

g

f

d

y

d

h

n

n

s

s

d

d

g

h

s

n

a

d

e

s

h

s

f

t

p

f

o

h

s

l

t

S

e

e

6

Company’s

exemplified,

below:

(1) For ea

and zone li

susceptible t

(2) Amon

highest debt

(3) Consi

zone if list t

its maximum

(4) Select

form the am

(5) Repe

concerned;

(6) Repea

centers by ar

In Fig. 6,

mpar-cluster

being displ

referring to

05.20.2014

dark blue rep

action. The

debts to be

radius of act

of the field a

representam

proximidade

representing

Fig. 6 Appliby the author

Mpar

business mo

, succinctly,

ach Work Cen

ist is georef

to cutting acti

ng the list o

t value;

idering the p

to all custome

m radius of ac

ts the highest

mount of notes

eat step 2-4

at step 2-4 t

rea;

, there is the

r with the s

layed on th

customers su

for CT Praça

present all cu

icons (points

collected. T

tion of group

action history

m os clientes s

e da maior dív

g the likely cu

cation of the ars, 2014.

r-Cluster: Ap

odel, this ap

according

nter — CT (o

ferenced form

ion;

of susceptible

previous stud

ers around th

ction;

t debts in the

s, the group w

4 to form a

to all classes

e application

selection of

he Google

usceptible to

a da Sé. The

ustomers susc

s) in red repr

The red circl

, defined from

y. Os ícones (p

suscetíveis e d

vida. The ora

ustomers, with

lgorithm mpar


pplication can

to the seque

of the 43 exist

m all custom

e, you select

dy of a radiu

his greater deb

e action radiu

will act;

all zone gro

s of all the w

of the algori

two (2) clus

Earth softw

cutting action

icons (points

ceptible to cut

resent the lar

le represents

m the best res

pontos) azul c

dentro do rai

ange icons (ite

h larger value

r-cluster. Prep


n be

ence

ting)

mers

t the

s by

bt to

us to

oups

work

ithm

sters

ware,

n on

s) in

tting

rgest

the

sults

claro

o de

ems)

es

pared

and

com

W

algo

foll

3. R

3.1

GIS

F

wer

usin

the

com

glob

S

pres

ima

rati

con

O

a st

than

mad

EKm

to c

Sé.

ord

EKm

susc

the

In

can

com

gen

betw

still

EKm

trig

In t


d debt within

mplete the tot

With the app

orithm furthe

lows concerni

Results and

Comparison

S

For the compa

re developed

ng the gadge

software Goo

mparative Ch

bal day of act

Selecting the v

sented in “2.1

age was ob

fied by diver

ncentrations.

On the optimi

tudy that the E

n the study o

de to Fig. 7

means algorit

cutting action

The icons (b

ers per clas

means algor

ceptible to cu

same spot du

n a visual co

n be seen grap

mpared to mo

nerated by E

ween classes

l possible to

means earnin

ggered and the

his case, it w

Optimization oy

the proximit

al customer t

plication of m

er significant

ing the comp

d Discussio

n between M

arative evalu

geographic re

et ARCGIS 1

ogle Earth. Fu

art summariz

tion group.

view of the IC

1 The structu

bserved opt

rgent selectio

ization in Dat

EKmeans alg

f application

7, which de

thm, referring

n on day 05.20

balloons) in

s, resulting

rithm in the

utting action

ue detailed in

omparison bet

phically EKm

odel the ICC.

EKmeans ha

and the deb

o see the ap

ng opportunit

e dispersion

was found for t

of the

ty distance h

triggered by a

mpar-cluster

results and th

parison betwe

ns

Models: Curr

uation of resu

epresentation

10.x/ESRI an

urthermore, i

zing the infor

CC cutting po

ure” through F

timization o

on in relation

ta Mining, it

gorithm had a

of K-means.

epicts the ap

g to customer

0.2014 for CT

shades of bl

from the ap

e database o

n. Furthermor

Fig. 3.

tween Fig. 7

means better

The clusteri

ave a greate

bt stains. How

pplication of

ties tied to de

of the class a

the cluster nu

igher debt to

a group.

were found

hat will be as

en models.

rent, Mining,

lts of models

ns of selection

nd displaying

it was made a

rmation for a

ortfolio it was

Fig. 3. In that

opportunities,

to debt stain

was found in

a better result

Thus, it was

pplication of

rs susceptible

T of Praça da

lue show the

pplication of

of customers

re, it appears

and Fig. 3 it

performance

ng algorithm

er alignment

wever, it was

f the model

ebt stains not

action radius.

umber 15 one

o

d

s

s

n

g

a

a

s

t

,

n

n

t

s

f

e

a

e

f

s

s

t

e

m

t

s

l

t

.

e

distance gre

cluster numb

Fig. 8 w

application o

image depict

in customer

Fig. 7 Appli

Fig. 8 App

By visual

validates gr

mpar-cluster

(Fig. 7) and

by mpar-clu

between gro

Mpar

eater than K

ber 13.

was made fo

of Geotechno

ts the applica

base suscept

ication of the a

plication of the

lly comparing

raphically th

r algorithm (F

the ICC (Fig

uster algorith

oup and the

Maop

r-Cluster: Ap

Km 3, and a

or a graphic

logy optimiza

ation of mpar-

tible to cuttin

algorithm EKm

e algorithm mp

g Figs. 8, 7, 3

he best perf

Fig. 8) compa

g. 3). The clu

hm have a s

debt stains.

aximizing the ctimization and


a light scatte

c display of

ation model.

-cluster algori

ng action, on

means. Prepare

par-cluster. Pr

3 and Table

formance of

ared to EKme

stering gener

strong alignm

Besides hav

ollection, distrd performance

Group conce


ering

f the

This

ithm

day

05.2

reds

app

cust

sim

ed by the autho

repared by the

2, it

the

eans

rated

ment

ving

mor

avo

In

met

algo

sele

ribution e of ray

entrated


20.2014 for C

s show ord

plication of mp

tomers susce

milar stain deta

ors, 2014.

e authors, 2014

re control ov

oiding cluster

n Table 2 it

thods, the

orithm was

ection. This

Optimization oy

CT of Praça d

ders per gro

par-cluster al

eptible to cut

ailed debt is sh

4.

ver the radius

dispersion.

t is possible

total charge

61% higher

gain was re

of the

da Sé. The ico

oup, resultin

lgorithm in th

tting action.

hown in Fig.

of action of

to identify t

ed by the

than the cu

ecorded even

7

ons (points) in

ng from the

he database of

Moreover, a

3.

f classes, thus

the measured

mpar-cluster

urrent model

n though the

7

n

e

f

a

s

d

r

l

e

8

amount of no

ICC selectio

can be seen

provides a r

and may eve

day ahead to

3.2 Ratifica

Gains

Aiming to

mpar-cluster

application

Canavieiras

a gain of ove

the group a

effective sev

collection of

over the prop

The tabul

CT Ilhéus,

question and

financial gai

Fig. 10 Comfor CT Ilheu2014.

Table 2 Comodels.

Modelagem

Atual

Ekmeans

Algoritmo GIS

Mpar

otes 1.47% lo

on. Regarding

graphically th

reduction of

en increase th

o earnings by

ation of Ge

o verify the g

r algorithm w

of simulatio

area. He was

er 60% compa

action radius

ven (7) classe

f R$ 187,900.

posal generat

ation of the g

demonstrates

d that this ten

ins for the sec

mparison betwes zone Canavi

mparative sum

QTD Turmas

QN

23 74

23 73

23 73

r-Cluster: Ap

ower than that

g the perform

hat the mpar-

the radius of

he amount of

reducing shi

eotechnologic

gains from th

was made a n

on in Ilhéus

s noticed in Fi

ared to the av

s. Moreover,

es it was poss

.56, represent

ted by the IC

geostatistical

s the validity

nds to bring or

ctor and thus f

een the currentieiras. Prepare

mmary of the

QTD Notas

Total Cobrado

47 R$135.0

32 R$164.5

36 R$217.7


t generated by

mance of radiu

-cluster algori

f action of gr

f notes execut

ifts.

ca Optimiza

he applicatio

new model of

work cente

ig. 9 and Tab

verage decreas

, with the s

ible to trigger

ting a gain of

C algorithm.

analysis used

y of the stud

rganizational

for the Comp

t selection anded by the auth

court of selec

o % Cobra

095,03 100%

590,76 122%

704,03 161%


y the

us, it

ithm

roup

ted a

ation

on of

f the

er in

le 3,

se of

same

r the

f 440%

d for

dy in

l and

pany.

d GIS hors,

ction

ado%

%

%

TabIlhe

M

Sele

Sel

The

Tab

sub

and

R

exa

com

lead

para

loca

stru

4. C

T

of

retu

of t

sim

opti

ene

also

org

D

the

whe

mpa

trig

a pr

In a

prov

the

circ

A

Geo

imp

opti

prac


ble 3 Compaeus.

Metodo Tur

ecaoAtual 0

lecao GIS 0

e implementa

ble 3 — Sele

stantial gains

d percentage s

Regarding the

ample, there

mpared with t

ds geospatia

allel with thi

ation of the

uctured base m

Conclusion

The developed

this study a

urns, organiza

he use of GIS

mulated clum

imizing logis

ergy supply s

o informed b

anized base m

During develo

applying var

en the resu

ar-cluster alg

ggered increas

roportional in

addition, the

vides an incr

forward

cumvented by

After this stu

ographer/GIS

plementation

imization lo

ctical gains

Optimization oy

ring the Cour

rmas Notas R

07 122 1

07 126 5

ation of the g

ection GIS) r

s, as the amou

susceptible to

e reduction o

is a signific

the present si

al/entail sm

is it is also p

client maps

map.

ns

d research sou

and the main

ational and f

S. To this end

mping algo

stics selecting

suspension ac

by concepts

map.

opment of the

rious clusteri

ults achieved

gorithm. The

sed value for

ncrease of the

focus on de

rease in the a

group d

y reducing the

udy, it was h

S Analyst, wh

of a pilo

gistics. Duri

of mpar-clu

of the

rt of model se

Raio Va

15 km R$42.6

5 km R$187.

geographical

research was

unt of notes,

o collection.

of the radius o

cant logistica

ituation and t

aller displac

possible to ha

(Digital Prin

ught to guide

n geographic

financial, to t

d, we have be

orithms, fo

g customers s

ction. All th

of GIS anl

e research we

ing algorithm

d in implem

possible gai

the charge, w

e total amoun

efining the a

amount of no

displacement

e dispersion o

hired by this

hich collabora

ot project t

ing deploym

uster algorith

election in CT

alor %

665,84 100%

.900,56 440%

concept (see

identified in

action radius

of action, for

al gain when

this selection

cements. In

ave the exact

nting) from a

e the potential

cal nature of

the detriment

en tested and

ocusing on

susceptible to

is study was

d use of an

e were tested

ms, especially

menting the

ins presented

which enables

nt recovered.

action radius,

otes made by

economy

of notes.

s company a

ated with the

to use this

ment, already

hm has been

T

e

n

s

r

n

n

n

t

a

l

f

t

d

n

o

s

n

d

y

e

d

s

.

,

y

y

a

e

s

y

n

Mpar-Cluster: Applied Algorithm of Geo-Selection for Optimization of the Credit Recovery of Electricity Supply

9

validated, and other gains related to: basic treatment

georeferenced customers; Studies of best access routes;

Redefinition of areas of workplaces, based on customer

sensitivity, secondary mesh network of energy and

support roads;

The optimization pilot project selection via GIS was

fully deploying in September 2014.

References

[1] Agência Nacional de Energia Elétrica – ANEEL, Resolução nº 414, Disponível em: http://www.aneel.gov.br/biblioteca/downloads/livros/REN_414_2010_atual_REN_499_2012.pdf, Accessed March 01, 2014.

[2] Maslow, A. 1962. Introdução à psicologia do ser. Rio de Janeiro: Eldorado.

[3] Jain, A. K., Murthy, M. N., and Flynn, P. J. 1999. “Data Clustering: A Review.” ACM Computing Survey 31 (3): 264-323.

[4] Oliveira, D. De P. R. 2005. Planejamento estratégico – conceitos, metodologia, práticas (22nd ed.). São Paulo:

Atlas.

[5] Davis, C., and Câmara, G. 2014. “Arquitetura de Sistemas de Informações Geográficas.” Accessed: http://www.dpi.inpe.br/gilberto/livro/introd/cap3-arquitetura.pdf, July 01.

[6] Câmara, G., and Monteiro, A. M. V. 2014. “Conceitos Básicos em Ciência da Geoinformação.” http://www.dpi.inpe.br/gilberto/livro/introd/cap2-conceitos.pdf, August 15.

mpar-cluster: applied algorithm of geo-selection for ... in an automated way. ... powerful framework...

Documents