Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

of

Data Science Applications in Finance and Investing Slide 1 Data Science Applications in Finance and Investing Slide 2 Data Science Applications in Finance and Investing Slide 3 Data Science Applications in Finance and Investing Slide 4 Data Science Applications in Finance and Investing Slide 5 Data Science Applications in Finance and Investing Slide 6 Data Science Applications in Finance and Investing Slide 7 Data Science Applications in Finance and Investing Slide 8 Data Science Applications in Finance and Investing Slide 9 Data Science Applications in Finance and Investing Slide 10 Data Science Applications in Finance and Investing Slide 11 Data Science Applications in Finance and Investing Slide 12 Data Science Applications in Finance and Investing Slide 13 Data Science Applications in Finance and Investing Slide 14 Data Science Applications in Finance and Investing Slide 15 Data Science Applications in Finance and Investing Slide 16 Data Science Applications in Finance and Investing Slide 17 Data Science Applications in Finance and Investing Slide 18 Data Science Applications in Finance and Investing Slide 19 Data Science Applications in Finance and Investing Slide 20 Data Science Applications in Finance and Investing Slide 21 Data Science Applications in Finance and Investing Slide 22 Data Science Applications in Finance and Investing Slide 23 Data Science Applications in Finance and Investing Slide 24
Upcoming SlideShare
What to Upload to SlideShare
Next
Download to read offline and view in fullscreen.

0 Likes

Share

Download to read offline

Data Science Applications in Finance and Investing

Download to read offline

Presented at Bethesda Data Science Meetup October 2019

Chris Conlan shares his perspective on when and how data science methods ought to be applied in financial services organizations.

Related Books

Free with a 30 day trial from Scribd

See all

Related Audiobooks

Free with a 30 day trial from Scribd

See all
  • Be the first to like this

Data Science Applications in Finance and Investing

  1. 1. Data Science in Finance How We Can Help & How We Can Hurt Bethesda Data Science Meetup – October 2019 Presented by Conlan Scientific
  2. 2. Chris Conlan – President chris@conlan.io
  3. 3. Types of Financial Assets • Stocks • Bonds • Currencies • Debt • Derivatives • Commercial Real Estate • Residential Real Estate • Businesses • Accounts Receivable • Automotive • Boats • Freight • Bank Deposits • Cash Equivalents • Inventory • Funds • Insurance • Employee Benefit Plans
  4. 4. Types of Financial Services Businesses • Fund Managers • Trusts • Banks • Family Offices • Mortgage Lenders • Mortgage Originators • Mortgage Servicers • Auto Lenders • Specialty Lenders • Insurers • Credit Unions • Custody Managers • Brokerages • Consultants • Researchers • Wealth Managers • Payday Advance Services • Invoice Factoring Services • Check Cashing Services • Credit Card Companies • Venture Capitalists • Payment Processors • Appraisers • Crypto Services
  5. 5. The Analytic Lifecycle at Finance Companies #1 Manual Review #2 Scoring #3 Machine Learning
  6. 6. The Analytic Timeline at Finance Companies #1 Manual Review #2 Scoring #3 Machine Learning Given an investment, process, or event that evolves over an amount of time, 𝝉, companies should evolve their decision-making process at about the following pace. 0 2𝝉 5𝝉
  7. 7. #1 Manual Review #2 Scoring #3 Machine Learning 0 2𝝉 5𝝉 Any Finance Company You just started a new finance business or entered a new investment market.
  8. 8. #1 Manual Review #2 Scoring #3 Machine Learning 0 2𝝉 5𝝉 Any Finance Company You hire some experts and let them call the shots.
  9. 9. #1 Manual Review #2 Scoring #3 Machine Learning 0 2𝝉 5𝝉 Any Finance Company You’ve been doing this a while and you know what to look for.
  10. 10. #1 Manual Review #2 Scoring #3 Machine Learning 0 2𝝉 5𝝉 Any Finance Company You build a scoring formula in Excel based on characteristics of your previously successful investments.
  11. 11. #1 Manual Review #2 Scoring #3 Machine Learning 0 2𝝉 5𝝉 Any Finance Company You have a ton of data on your investment performance and many investments on the table.
  12. 12. #1 Manual Review #2 Scoring #3 Machine Learning 0 2𝝉 5𝝉 Any Finance Company You build a machine learning model to make sure all your new investments are quantitatively similar to your best investments.
  13. 13. #1 Manual Review #2 Scoring #3 Machine Learning 0 1 year 5𝝉 Credit Card Company You take a bunch of applications and accept the best-looking customers. You are overly conservative. 𝝉 ≈ 6 months
  14. 14. #1 Manual Review #2 Scoring #3 Machine Learning 0 1 year 2.5 years Credit Card Company You have a big book of active clients. You score them with a homemade formula and drop the bottom 20% to improve your portfolio. You use your formula to score your incoming clients. Your formula looks something like this. rank(cash_flow/limit) * 0.33 + rank(equity/limit) * 0.33 + rank(late_payments) * 0.34;
  15. 15. #1 Manual Review #2 Scoring #3 Machine Learning 0 2𝝉 2.5 years Credit Card Company You now know how your clients interact with your service, but you also know which clients make the most money for your business. You build a machine learning model optimized for client profitability.
  16. 16. Don’t Skip Steps #1 Manual Review #2 Scoring #3 Machine Learning 0 2𝝉 5𝝉
  17. 17. What if 𝝉 is small? #1 Manual Review #2 Scoring #3 Machine Learning 0 8 months 2 years Good, use this system.
  18. 18. What if 𝝉 is huge? #1 Manual Review #2 Scoring #3 Machine Learning 0 40 years 100 years 𝝉 ≈ 20 years in Commercial Real Estate
  19. 19. What if you’re in the stock market? #3 Machine Learning 0 0 years 0 years Investments have a known outcome. Go straight to machine learning.
  20. 20. Predict Something Predictable • Accounting formulas introduce too much noise to ML processes • Predict something important but self-contained Bad Target Variables • Date of repayment • Net Profit • Years until founder can retire Good Target Variables • Likelihood of repayment • Proportion of repayment • Capital appreciation
  21. 21. Predicting Profits • Plug your predictions into larger accounting formula to predict profits • Don’t predict profits directly Investment Data Accounting Logic Accounting Logic Accounting Logic Your Machine Learning Model Accounting Logic Profit Estimate Your Prediction Accounting Logic
  22. 22. Predicting Profits • This yields the highest accuracy when predicting profits • Why? Accounting practices aren’t as strict as you think • Accounting practices can introduce lots of noise Investment Data Accounting Logic Accounting Logic Accounting Logic Your Machine Learning Model Accounting Logic Profit Estimate Your Prediction Accounting Logic
  23. 23. Don’t Do This Your neural network doesn’t have a CPA Investment Data Your Machine Learning Model Profit Estimate
  24. 24. Conclusion 1. There’s a reason to take it slow when it comes to ML adoption. 2. Some companies need it, some companies will never need it. 3. Predict small things for big outcomes.

Presented at Bethesda Data Science Meetup October 2019 Chris Conlan shares his perspective on when and how data science methods ought to be applied in financial services organizations.

Views

Total views

142

On Slideshare

0

From embeds

0

Number of embeds

1

Actions

Downloads

4

Shares

0

Comments

0

Likes

0

×