This document provides an abridged sample risk report analyzing defaulted mortgage loans in Turkey from 2012-2014. It uses graphical data mining techniques like density plots, violin plots, and facet plots in R to visualize loan default trends by region, city, sector, and time period. The report aims to identify financial anomalies and risk profiles according to geographic and temporal factors. It finds significant differences in risk profiles affected by these factors. Only summary information from the full report covering all Turkish regions from 2010-2015 is presented.
1. Fatma ÇINAR, MBA Capital Markets Board of Turkey
e-mail: fatma.cinar@spk.gov.tr fatma.cinar@datalabtr.com
@fatma_cinar_ftm @DataLabTR
C. Coşkun KÜÇÜKÖZMEN, PhD e-mail: coskun.kucukozmen@ieu.edu.tr
@ckucukozmen @RiskLabTurkey
Kutlu MERİH, PhD e-mail: kutlu.merih@datalabtr.com
kutmerih@gmail.com @cortexien
www.datalabtr.com
https://www.riskonomi.com
RISK MANAGEMENT WITH DATA VISULISATION
RISK REPORT (abridged)
2. Description of The Report
This a abridged Sample Report of Defaulted
Mortgage Loans for the 12 NUTS Sectors and
81 cities of Turkey with 2012-2014 time span
Covers only Whole Turkey and West Anatolia
Region
Real Report covers all the 12 NUTS Regions and
81 Cities of Turkey
With 2010-2015 time span
3. Purpose of the Report
• With this study we investigate NUTS 12
Regions credit loans performations by
Graphical Datamining Analysis technique
with a suitable software developed by us.
• This technique is submitted various OR and
Finance congress
• Dataset are factorized according to cities and
years, sectorals and financial periods
factors.
• Sample Report Covers Periods: 2012-2014
accounts.
4. Source of The Data
• We downloaded FINTURK dataset from the
site of BRSA and anotated it by NUTS factors.
• Our software read this data from an excel file
with the name of “dataset”
• From now on “dataset” means our improvised
NUTS Credit Loans FINTURK data
5. Description of The Analytics
• Data: BRSA* and NUTS of Turkey
(Nomenclature of Territorial Units for Statistics, NUTS)
• Dataset: NUTS Region Investment Promotion and 3
account period Graphical Datamining Analysis of
FINTURK of BRSA
• Period: 2012-2014 Accounts
• Dataset are factorized according to year, sector,
quarter and region factors.
• Graphical Datamining and Data Visualisation applied on
this factorized data.
*BRSA: Banking Regulations and Supervisison Agency
8. NUTS-1:12 Regions of Turkey
• NUTS-1: 12 Regions
• NUTS-2: 26 Subregions
• NUTS-3: 81 Provinces
1. MEDITERRANEAN
2. SOUTHEAST ANATOLIA
3. EAGEAN REGION
4. NORTHEAST ANATOLIA
5. MIDDLE ANATOLIA
6. WEST BLACK SEA
7. WEST ANATOLIA
8. EAST BLACK SEA
9. WEST MARMARA
10. MIDDLE EAST ANATOLIA
11. ISTANBUL
12. EAST MARMARA
9. Wednesday, February
3, 2016
İstanbul
Region
West
Marmara
Region
Aegean
Region
East
Marmara
West
Anatolia
Region
Mediterranean
Region
Anatolia
Region
West Black
Sea Region
East Black
Sea Region
Northeast
Anatolia
Region
East
Anatolia
Region
Southea
st
Anatoli
a
İstanbul
(Subregion)
Tekirdağ
(Subregion)
İzmir
(Subregion)
Bursa
(Subregion)
Ankara
(Subregion)
Antalya
(Subregion)
Kırıkkale
(Subregion)
Zonguldak
(Subregion)
Trabzon
(Subregion)
Erzurum
(Subregion)
Malatya
(Subregion)
Gaziant
ep
(Subreg
ion)
Edirne
Aydın
(Subregion)
Eskişehir
Konya
(Subregion)
Isparta Aksaray Karabük Ordu Erzincan Elazığ
Adıyam
an
Kırlareli Denizli Bilecik Karaman Burdur Niğde Bartın Giresun Bayburt Bingöl Kilis
Balıkesir
(Subregion)
Muğla
Kocaeli
(Subregion)
Adana
(Subregion)
Nevşehir
Kastamonu
(Subregion)
Rize
Ağrı
(Subregion)
Dersim
Şanlıurf
a
(Subreg
ion)
Çanakkale
Manisa
(Subregion)
Sakarya Mersin Kırşehir Çankırı Artvin Kars
Van
(Subregion)
Diyarba
kır
A.Karahisar Düzce
Hatay
(Subregion)
Kayseri
(Subregion)
Sinop Gümüşhane Iğdır Muş
Mardin
(Subreg
ion)
Kütahya Bolu Kahramanmaraş Sivas
Samsun
(Subregion)
Ardahan Bitlis Batman
Uşak Yalova Osmaniye Yozgat Tokat Hakkari Şırnak
Çorum Siirt
Amasya
1 Province 5 Province 8 Province 8 Province 3 Province 8 Province 8 Province 10 Province 6 Province 7 Province 8 Province
9
Provinc
e
10. Graphical DataMining Analysis with
FINTURK Sectoral Loans Dataset
• Real Time Interactive Data Management for
• Effect and Response Analysis
Technique:
• Graphical DataMining using #ggplot2
Graphical Package of #R Software
• Graphical DataMining will be the dominant
anlysis technique of the future
11. Styles of Graphs
• For brevity we apply three types off ggplot2
graphical styles with ggplot2 geoms with this
report
1. Densityplot with geom_density()
2. Violinplot with geom_violin()
3. Facetplot with facet_grid()
• Logarithmic scale leads a more stable density
formations for financial data.
12. Description of Density Graphs
• Density Graphs are the
continuous version of
Histograms They plot a
single numerical variable
against their frequency.
• We can detect single or
multiple peaks of density
graphs and pinpoint the
effective factors.
• On the other hand
soperposing density graphs
acording the factors with
different colors provide us
with information of the
effect of the factors
13. Description of Violin Graphs
• Violin Graphs can be
seen as two-
dimensional density
graphs
• Axis of the violin
represents the median
of the free variable
• Through the median of
X-axis Y-density graph
occurs with mirror copy
14. Mushroom, Pottery and Bottle Risk
Profiles
• Violin Graphs comes
with Mushroom, Pottery
and Bottle formations
• Mushroom formation
represents a risk
concentration on hig
order values of financial
data
• Pottery means risk on
the medium order
• and the bottle menas risk
on the lower orders
15. Description of Power Law Graphs
• When double Log scale
applied Power Law
analysis is a by product
• LogY = a.LogX + b
• a is the Risk Measure
• And it is the same for
every level of X and Y
• Power Law means that
risk is scale free
16. Description of Facet Graphs
• Facet graphs of ggplot2
package can show us
3-dimensional graphs
distributed according 3
factors in matrix form.
• In which we can see
the anomalies occurs on
which year and which
region and which
period.
17. Risk Profiles
Graphical Dataminig of the Profiles of
Default Mortgage Risks for Whole
Turkey, 12 NUTS Regions and 81 Cities
(This version covers only Whole Turkey and
West Anatolia Region)
18. Risk Profiles for Whole Turkey
Here we investigate Mortgage Loans
versus Default Mortgage Loans for
Whole Turkey according to region, year
and period factors.
X and Y Scales used is Log10
23. Power Law Risk Analysis of Mortgage Loans
by Nuts Regions of Turkey
Wednesday, February
3, 2016
24. Faceted Power Law Analysis of Mortgage Loans
by Quarters
Wednesday, February
3, 2016
25. Risk Profiles for
W. Anatolia Region
Here we investigate Mortgage Loans
versus Default Mortgage Loans for
West Anatolian Region according to
cities, year and period factors.
X and Y Scales used is Log10
26. Density Graphs of Mortgage Loans
by Cities of West Anatolia
Wednesday, February
3, 2016
32. Conclusion
• Graphical Datamining applied on this
factorized data and financial anomalies
detected acording to time and space factors.
• We observes apparently obvious differences
of risk profiles affected by these factors
• It is quite clear that pictures tells more stories
than numbers
• This is only an abridged version of the whole
report for 12 Nuts Regions
34. Resources
• Küçüközmen, C. C. Ve Çınar F., (2014). “Finansal Karar Süreçlerinde Grafik-
Datamining Analizi”, TROUGBI/DW SIG, Nisan 2014 İstanbul,
http://www.troug.org/?p=684
• Küçüközmen, C. C. ve Çınar F., (2014). “Görsel Veri Analizinde Devrim” Söyleşi,
Ekonomik Çözüm, Temmuz 2014, http://ekonomik-cozum.com.tr/gorsel-veri-
analizinde-devrim-mi.html.
• Küçüközmen, C. C. ve Merih K., (2014). “Görsel Teknikler Çağı" Söyleşi, Ekonomik
Çözüm, Temmuz 2014, http://ekonomik-cozum.com.tr/gorsel-teknikler-cagi.html
• Küçüközmen, C. C. and Çınar F., (2014). “Banking Sector Analysis of Izmir
Province: A Graphical Data Mining Approach”, Submitted to the 34th National
Conference for Operations Research and Industrial Engineering (YAEM 2014), Görükle
Campus of Uludağ University in Bursa, Turkey on 25-27 June 2014.
• Küçüközmen, C. C. and Çınar F., (2014). “New Sectoral Incentive System and
Credit Defaults: Graphic-Data Mining Analysis”, Submitted to the ICEF 2014
Conference, Yıldız Technical University in İstanbul, Turkey on 08-09 Sep. 2014.
• Küçüközmen, C. C. and Çınar F., (2015). “Visual Anaysis of Electricity Demand
Energy Dashboard Graphics” Submitted to the 5th Multinational Energy and Value
Conference May 7-9, 2015 Kadir Has University in İstanbul, Turkey
• Merih, K. C. and Çınar F., (2015). “Sectoral Loans Default Chart of Turkey ”, Submitted to
35th National Operations Research and Industrial Engineering Congress (ORIE 2015) 09-
11,September, 2015,Middle East Technical University, Ankara, Turkey