Democratization of data science: Using machine learning to build credit risk models
Hello! I AM Moto Tohda I am the Vice President of Information Systems @ Tokyo Century (USA) Inc. 2
Tokyo Century (USA) Inc. Who we are Tokyo Century Corporation • Established: 1985 • Established: 1969 • Private • Tokyo Stock Exchange • 37 Countries 3
OUR 1 st STEP How did we start our journey on Big Data 4
1. Perception How do you change people’s perception of running the business using TECHNOLOGY ? 5
Technology = “Unknown” Disrupt The Comfort Disrupt The Business Flow (System Flow) • Unfamiliar product or process • Somehow the current system • New learning curve flow became the business flow • New process will/should not bring efficiency 6
Zero2Hero Stack from Bardess: Storage & Processing Visual Analytics Advanced Analytics Hadoop, Big Data processing Visualize the credit risk Built the credit risk model framework model in real-time. Run the prediction through Cazena – SaaS Data Lake Connects DataRobot, Qlik, give the answer back on Microsoft Azure Cloudera, and our SQL quickly servers. 7
Proof of Concept “POC” -Microsoft Azure/Cloudera/Trifacta/Qlik/DataRobot Customer 360 Model Credit Risk Model • Know your customer • Used our existing data only • Learned to visualize using • Tested to see what DataRobot existing data will bring • It can be used for target • Surprisingly good result! marketing, customer service, etc. 8
or HARD TRUTH What happens when technology is way over people’s head 9
2. Change Management How do you convince the business side to GET IT ? 10
HOW DO WE Use DATA ? 1. 2. 3. Historical Predictive Prescriptive 11
HOW CAN WE Compete USING DATA Reactive Proactive See what happened, and try to Forecast the future and plan an remedy the situation. action ahead of time. 12
HOW CAN WE Share DATA “Hunch” “Quantify” People “KNOW” what to do. Use DATA to put numbers to all the indicators. Cannot explain, because this comes from the years of People can speak the same experience and the knowledge. language, because of cosistency. 13
“Top - Down” or “Bottom - up” Top-Down Bottom-up • Strong leadership • Educate first • Uncomfortable surprises • Include in the process • May not be fully convinced • No surprises 14
PRODUCE Fast RAW is good VISUALIZE ○ We cannot and should not ○ Users will have to see it to wait for perfection. believe. ○ We can always “fail fast” and ○ More examples, the better. revise. ○ Small Qlik applications can go a long way. 15
3. Democratize How do you empower PEOPLE with data? 16
Giving “Tools & Data” = “Democratize” Traditional Way Democratized Way • IT or Reporting specialist work • IT make the data available on data • Users will proactively create • Users confirm requirements their own application • Go live • Creativity 17
Create Champions ○ Qlik Training for business users ○ Analytics Day ○ Create a common dataset for users to access Bring in business users to experience, and VISUALIZE . 18
Summary ○ Make the 1 st Step ○ Digital Transformation or disruption is NOT IT-only projects ○ Increase Data Literacy for the entire organization, change the perception 19
THANKS! Any questions? You can find me at mtohda@tokyocentury.com https://www.linkedin.com/in/moto-tohda-a520521/ 20
CREDITS ○ 09/24/19 FinData Day @ Strata Conference 2019 NYC ○ Presentation template by SlidesCarnival 21
Rate today ’s session O’Reilly Events App Session page on conference website
Recommend
More recommend
