🧠

FinTech - Shaping the Financial World mit course finance technology fintech AI machine learning

Artificial Intelligence and machine learning in financial services

Background and definitions

  • Categories of machine learning
    • Supervised learning
      • Contains labels
      • Algorithm predicts labels for observations without labels
    • Unsupervised learning
      • Data has no labels
      • Algorithm predicts patterns based on underlying characteristics
    • Reinforcement learning
      • Falls between supervised and unsupervised
      • Algorithm is fed unlabelled set of data --> chooses an action for each data point --> gets feedback based on the action
    • Deep learning
      • Utilizes neural networks
  • Machine learning algorithms are used to identify patterns
    • The patterns are correlation

Drivers

  • Faster processor speeds
  • Lower hardware costs
  • Access to computing power via the cloud
  • Storage is cheap for large amounts of data
  • The financial sector benefits from AI/ML growth in other areas
    • Search engines, self-driving cars, etc.

Selected use cases

  • Examples:
    • Sentiment analysis
    • Trading signals
    • Fraud detection

Customer-focused use cases

Credit scoring

  • Speed up lending decisions
  • Sources for unstructured data used to create a credit score:
    • Social media
    • Mobile phone use
    • Test messages
  • Goals:
    • Lowering the cost of assessing credit risks
    • Increase the number of people whose credit risk is being assessed
  • Problems
    • False negatives can be very harmful for people
    • Black box decision making may not be fair
      • Biases built into the AI models?

Insurance

  • Using data like online shopping or sensors in your car
    • I'm actually pro-car telemetrics for insurance; drive irresponsibly --> pay more insurance
      • What about speeding to the hospital because your wife is pregnant?
      • Hard to deal with edge cases
  • Used to make it easier for humans to process claims
    • Highlight key considerations

Chatbots

  • Serve more customers with automated responses

Operations-focused

Capital optimisation

  • Maximization of profits given scarce capital
  • Relies heavily on math

Market Impact Analysis

  • Evaluating the effect of a firm's trading on market prices
  • Attempt to minimize trading execution costs

Trading and portfolio management

  • Execution --> sell-side
  • Portfolio management --> buy-side

Trading execution

  • Using data to improve ability to sell
    • Eg. using past trading behaviour to predict a client's next order
  • Taking advantage of the massive amounts of data trading creates
  • Used to pro-actively manage risk exposure
    • Eg. flag an account if it's risk profile has increased

Portfolio management

  • Identify new signals on price movements
  • Mostly used by quant funds
  • Drives a small amount of quant funds' trades due to lack of trust
    • They want insights on the decisions made by models

Micro-financial analysis

Impact on markets

  • Enabling certain participants to collect and analyze data on a bigger scale
    • Help to build understanding on the relationship between market prices and various factors
  • Lower participants trading costs

Impact on financial institutions

  • Enhancing operations
    • Increasing revenue, decreasing costs
  • Better risk management
    • May miss new types of fraud due to lack of historical data
  • Reducing fraud
  • Increase collaboration between financial institutions and other industries
    • Other industries are driving AI/machine learning
  • New set of issues due to black box decision-making
    • Hard for humans to grasp AI decisions
    • Who is responsible for bad decisions?
  • Increased dependance on third-parties
    • Developers of the technology

Impact on consumers and investors

  • Lower fees and borrowing costs
  • Wider access to financial services
    • Could also be the opposite if a machine learning model determines you aren't fit for particular services
  • More personalized services

Macro-financial analysis

  • More efficient risk management for individual loans may benefit the larger system
  • New economies of scope (collaboration between industries) could help economic growth
  • Economy-wide investment in AI/machine learning could stimulate economic growth

Market concentration

  • Those who own the data benefit the most
  • Those who develop the technology benefit the most
  • Those who can afford the technology benefit the most
    • Big banks might edge out smaller firms simply due to their ability to pay for better tech
    • This is probably already the case?

Backlinks

Class 2 - Artificial Intelligence, Machine Learning, and Deep Learning