AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or...

23
AI/ML for Operational Data Management Masood Khatri VP, Product Owner Xoriant CDi Apnav Agrawal Director, Data Management Xoriant CDi A conversation with

Transcript of AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or...

Page 1: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

AI/ML for Operational Data Management

Masood KhatriVP, Product Owner Xoriant CDi

Apnav AgrawalDirector, Data ManagementXoriant CDi

A conversation with

Page 2: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Moderated by Mike MeritonCo-Founder & COO, EDM Council

• Joined EDM Council full-time 2015 to lead Industry Engagement

• EDM Council Co-Founder & First Chairman (2005-2007)

• EDM Council Finance Board Chair (2007-2015)

• Former CEO GoldenSource (2002-2015)

• Former Executive for D&B Software and Oracle

• FinTech Innovation Lab – Executive Mentor (2011 – Present)

Page 3: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Speaker Profiles

Masood KhatriVP, Head of Digital Practice, Product Owner

Apnav AgarwalDirector, Data Management and Governance

Technology Executive & Recognized Thought Leaderwith passion for building and leading transformationaldigital products and bringing businesses to theforefront of technology and innovation. Masood hasextensive experience in strategically designing,building products, roadmaps & bringingideas/concepts to life to drive implementation.Masood has developed and deployed cutting-edgecross-functional digital capabilities to needs of fortune100 enterprises, which exceeded business andcustomer expectations.

A DCAM certified professional having over 15 years ofexperience in Data Management & Governance withproven ability to offer strategic direction/guidance toenterprises for enterprise data strategy developmentand management. Apnav has experience in guidingthe team related to the data warehouse, businessrules, data modelling, architecture creation, executivecollaborations, and technical requirements. Apnav isleading a team of data management SMEs whiledeveloping and implementing SaaS products accordingto client specifications.

Page 4: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Traditional Methods- Operational Data Management

Transformation &Business Rules

Manual Reviews,4 Eye ChecksSubscribe to Feeds Source Information

Manually

Page 5: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

POLL – Challenges

Page 6: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Challenges - Operational Data Management

Labor Intensive Human Errors ImpactData Quality

Lack of Consistency -Variety of Data

Sources and Formats

Delays - Volume of Data to Manage

Page 7: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Operational Efficiencies in Data Management

AI/ML to improve Operational Efficiency

Data Sources

Ope

ratio

nal D

ata

Man

agem

ent

PLATFORM

PROCESS

PEOPLE

Clean and Up-to-date data to fulfill business objectives and regulatory compliance

}

Page 8: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

AI for Operational Data Management

DQ Framework to Enhance Data Quality

Enrichment and Standardization of Addresses

Automating HierarchyData Extraction

Natural Language Processing Machine Learning Deep Learning

AI Techniques Used

Page 9: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

POLL – AI/ML

Page 10: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

www.cdi.xoriant.com

DQ FRAMEWORK TO ENHANCE DATA QUALITY

Page 11: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

DQ Framework to Enhance Data Quality

� Increase in throughput for QC review of the data elements.

Sourced DataAttribute : Party TypeNew Value : Other Financial

Old Value : Investment Advisor

Blackrock Group Ltd Apply Data Science

• Natural Language Processing• Machine Learning Models :

Feature Extraction, Feature Engineering, Classification

CDi DQ Framework Recommendation

Accept

CDi Gold Database

Page 12: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Results

SampleResults

Details of Xoriant Solution

Models

# Attribute Old Value New Value Prediction Probability Action

1 IMMEDIATE_PARENT REINSURANCE GROUP OF AMERICA INC RGA REINSURANCE COMPANY 0.9529 Accept

2 IMMEDIATE_PARENT BANK OF HAWAII First Hawaiian Bank 0.5129 Review

3 IMMEDIATE_PARENT BNY MELLON FUND MANAGEMENT LUXEMBOURG SA BNY MELLON GLOBAL FUNDS PLC 0.0202 Reject

• Hashing Vectorizer

Data Processing

• Fuzzy Logic• Co-occurrence

Matrix • Reference

Sources

Feature Selection

Deep Learning

Our DQ Framework automated exception

resolution saving about 60% of the manual effort.

Business Benefit

Page 13: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

www.cdi.xoriant.com

ADDRESS ENRICHMENT AND STANDARDIZATION

Page 14: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Address Enrichment and Standardization

Apply Data Science

• Information Extraction• Natural Language

Processing• Machine Learning Models

:Feature Extraction, Feature Engineering, Classification

Xoriant Corporation

343

Thor

nall

Stre

et, S

uite

720

, Ed

ison

NJ

0883

7

Standardize Address Fields

Address 1: Suite 720

Address 2: 343 Thronall Street

City: Edison

State/Province: NJ

Country: USA

Postal Code: 08837

Cleansed Data• Registration Authority• Corporate Site• Stock Exchange• SEC• Regulator Site

Models

Deep Learning

Word Embeddings

Pre-trained Models Labeled Data

Page 15: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Edit Address Enter One Line Address and Select the Country – Click Validate

Page 16: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Address StandardizationAddress Standardization using ML/AI provides enriched address in separate fields for user to review

Page 17: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Address StandardizationAddress automatically applied to respective attributes.

Page 18: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Address Standardization AccuracyNorth America

Canada US

90 93

76 97

76 99

100 100

100 99

100 99

76 93

Europe

Germany FranceUnited

Kingdom

84 92 92

78 84 87

94 84 85

100 98 99

100 100 95

100 100 95

78 84 85

Australia

Australia

96

96

96

100

98

100

96

Fiel

dLe

velA

ccur

acy

Address 1(Building, Unit/Suite, Floor)

Address 2(Street)

Address 3(Landmark)

City

State or Province

Zip or Postal Code

Address Level Accuracy

Asia

China Hong Kong

96 92

82 90

84 90

100 94

100 100

100 100

82 90

Handling volume of 1 million entities having 3 addresses per records would have taken 1,000 person-days. Supported by 90%+ accuracy through AI/ML this effort is reduced by 900 person-days.

Page 19: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

www.cdi.xoriant.com

AUTOMATING HIERARCHYDATA EXTRACTION

Page 20: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Automating Hierarchy Data Extraction

Apply Data Science

• Information Extraction• Natural Language

Processing• OCR• Machine Learning

Models :Feature Extraction, Feature Engineering, Classification

Immediate ParentIdentification

• Kabouter Management LLC• Royal London Asset

Management• Henderson Group PLC• Aberforth Partners• Blackrock, Inc.• FIL Limited• Kames Capital

Ultimate Parent Identification

• Kabouter Management LLC: 6.0% (Holding)

• Royal London Asset Mgmt:5.95% (Holding)

• Henderson Group : 5.09% (Holding)With more than 50% holding, Brewin Dolphin Limited is Ultimate Parent.

Cleansed Data

Repeat

N steps

• Registration Authority

• Corporate Site• Stock Exchange• SEC• Regulator Site

Brewin Dolphin Limited

Page 21: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Benefits

Improved DataQuality

IncreasedProductivity Speed Reduction in Cost

Page 22: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

Questions?

Page 23: AI/ML for Operational Data Management · 2020. 6. 16. · (Landmark) City State or Province Zip or Postal Code Address Level Accuracy Asia China Hong Kong 96 92 82 90 84 90 100 94

FOR MORE INFORMATION:

MASOOD KHATRI

VP, Head of Digital Practice, Product Owner [email protected]: 732.718.9547

APNAV AGRAWAL

Director, Data Management & [email protected]: 646.895.3538

Visit us at https://cdi.xoriant.com to learn more