Topic 9.1 - 9.5: Big Data & Machine Learning in the Financial Industry Flashcards

1
Q

Identify factors that may contribute to increases in third party dependencies
among financial institutions.

A
  • network effects

- scalability of new technologies

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Explain why unexpected forms of interconnectedness among institutions
could be created.

A

Institutions’ ability to make use of big data from new sources lead to dependencies on previously unrelated macro variables, financial prices

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Explain why new forms of macro-level risks could emerge.

A

Algorithms generating uncorrelated returns will be exploited on a scale whereby correlations will increase

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Describe the two recent developments that have contributed to increased
interest in AI.

A

Advances in the availability of financial sector data and infrastructure

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

List factors contributing to making the markets more efficient

A

More efficient processing of information (e.g. credit decisions, financial markets, insurance contracts) may contribute to a more efficient financial system.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Describe the relationship between AI, machine learning, and the three algorithms appearing in Figure 1.

A

machine learning is a sub-category of AI.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Describe the relationship between AI, machine learning, and the three algorithms
appearing in Figure 1.

A

Categories of Machine Learning:

  • Supervised learning (algo is fed labeled training data)
  • Unsupervised learning (algo is fed unlabeled training data)
  • Reinforcement learning (falls between sup and unsup).
  • Deep learning (works in layers)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Describe the role of machine learning algorithms in determining causality
vs. correlation.

A

Machine learning cannot determine causality but they can identify correlations.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Define ‘augmented intelligence.’

A

augmentation of human capabilities (rather than replacing them).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Discuss the supply factors related to advances in computing technologies
and changes in the financial sector.

A
  • computing power, lower hardware costs, and better access to computer power via cloud services.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Discuss the demand factors related to the search for higher profits, increased
competition, and changes in the regulatory environment.

A

Profitability: Potential for cost reduction, revenue gains, risk reduction

Competition: “arms-race” with other firms

Regulation: Prudential regulations, data reporting, AML

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Describe customer-focused uses, such as credit scoring, insurance, and client-facing chatbots.

A

Credit scoring - ML designed to speed up lending decisions, while limiting risk.

Insurance - analyze complex data to lower cost and improve profitability.

Client-facing chatbots - automated programs use NLP to interact with clients in natural language.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Describe operations-focused uses, such as optimal capital allocation,
risk management modeling, and market impact analysis.

A

Capital optimization - Maximisation of profits given scarce capital.

risk management modeling - back-testing and model validation. (banks use it to evaluate how well their risk models are performing).

market impact analysis - obtain more information from sparse historical models or help identify non-linear relationships in order flow.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Describe portfolio management and trading uses.

A
  • identify new signals

- make more use of vast amount of data available

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Describe regulatory compliance and supervision uses by financial institutions,
central banks, macroprudential authorities, and market regulators.

A
  • RegTech: facilitates compliance more efficiently and effectively (ML+NLP)
  • Monitoring behavior of traders
  • Interpret data inputs (e-mails, spoken word, IM)
  • To cope with new regulations
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

How is machine learning used in KYC

A
  1. evaluating images in identifying documents

2. calculate risk scores to determine who needs to receive additional scrutiny.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

Benefit of artificial intelligence and machine learning in information
gathering and processing their potential impacts on financial markets.

A

AI/ML can enhance the efficiency of information processing, reducing information asymmetries. This has the potential to strengthen the information function the financial system.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

Describe the uses of artificial intelligence and machine learning in information
gathering and processing their potential impacts on financial markets.

A
  • collect & analyze information on a greater scale to better understand the relationship between market prices and various factors.
  • Lower market participants trading costs
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

Describe the uses of artificial intelligence and machine learning in improving
the efficiency of financial institutions.

A

AI/ML can increase profitability and reduce risk:

  • enhance machine based processing of various operations
  • Earlier and more accurate estimation of risk
  • Encourage collaboration between FIs
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

Describe the uses of artificial intelligence and machine learning by financial
institutions and their potential impacts on customers and investors.

A
  • Lower fees
  • Wider access to financial services
  • Customized/personalized financial services
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

Describe economic growth and enhanced economic efficiency that could result
from artificial intelligence and machine learning applications to financial services.

A

ML/AI can contribute to efficiency and growth of the economy through:

  1. Enhancing the efficiency of financial services
  2. Facilitating collaboration and realizing new economies of scope
  3. Stimulating investments in AI and ML
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

Describe the implications of uses of artificial intelligence and machine learning
by financial institutions for market concentration and the systemic importance of
those institutions.

A
  • Emergence of small number of advanced third-party providers which increases concentration
  • Access to big data might be of systemic importance/only big firms can afford large investments for ML/AI.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

Describe how financial institutions’ uses of artificial intelligence and machine
learning could be sources of greater instability and vulnerability in financial markets

A
  • trading algos might be less predictable
  • HFT might increase market volatility through large sales or purchases
  • Minimizing of margins may increase risks
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
24
Q

Describe how the uses of artificial intelligence and machine learning by the
insurance industry could affect both moral hazard and adverse selection problems.

A
  • using AI/ML to continuously adjust insurance fee in accordance with behavior of policy holder reduces moral hazard.
  • Offer customized insurance policy reduces adverse selection.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
25
Q

Algorithm

A

set of computational rules to be followed to solve a mathematical problem

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
26
Q

Artificial intelligence

A

the theory and development of computer systems able to perform tasks that traditionally have required human intelligence

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
27
Q

Augmented intelligence

A

augmentation of human capabilities with technology

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
28
Q

Cluster analysis

A

Statistical technique where by data are classified into distinct groups

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
29
Q

Deep learning

A

algorithms inspired by the structure and function of the brain (neural networks)

30
Q

FinTech

A

technology enabled financial innovation

31
Q

IOT

A

the inter-networking of physical devices, vehicles and other items embedded with electronics and software.

32
Q

Machine Learning

A

method of designing a sequence of actions to solve a problem that optimize automatically through experience (limited/no human intervention)

33
Q

Natural language processing

A

focus on programming computers to parse, process, and understand human language

34
Q

Reinforcement learning

A

algo receives unlabeled data, chooses an action, receives feedback on the result of action and learns.

35
Q

Supervised learning

A

algo receives training data that contains labels on the observations

36
Q

Tonality analysis

A

gauge the negativity of a piece of text by counting terms with a negative connotation

37
Q

Topic modeling

A

unsupervised learning lets the data define key themes in text

38
Q

Unsupervised learning

A

Data provided to algo does not contain labels.

39
Q

Describe the difference between machine programming and machine learning.

A

machine programming starts from the inputs, uses human understanding to explicitly define the rules that turn the inputs to outputs.

ML starts from examples of how inputs and outputs relate and seek to learn the rules via estimation.

40
Q

Explain how tree models can be used for sorting stocks.

A
  1. Select sorting variables (MC, B/M).
  2. Form groups based on MC size.
  3. Within the groups further sort on B/M.

Leaves of the trees are comprised of groups of stocks most similar based on sorting variables.

41
Q

Describe the difference between “regression” and “regression with transformed X’s.”

A

Transformed X’s add a layer of complexity (denoted Z) and rather than regress y on x it first processes x into z and measures how y relates to z.

42
Q

Explain why finance is different when it comes to the applications of machine learning.

A
  • Low signal-to-noise ratios due to noisy markets and is kept low.
  • Evolving markets (market efficiency).
  • Short samples and unstructured data (not really big data)
  • Need for interpretability (fiduciary duty)
43
Q

Explain why it is important to combine economic theory and machine learning

A

It can tackle the low signal-to-noise ratio by bringing economic theory to describe some aspects of the data.

44
Q

Describe features such as autocorrelation, jumps, structural breaks, and
volatility clustering.

A

autocorrelation: correlation of same variable between successive time-intervals.

Jumps: jumps in individual stocks due to informational or liquidity shocks.

Breaks: sudden changes in behavior or economic events lead to structural breaks in the financial time series data.

Volatility clustering: When its high its likely to stay high and vice-versa.

45
Q

Differentiate time-series dependence from cross-sectional dependence.

A

time-series dependence: residuals are correlated across firms and time.

cross-sectional dependence: in a particular time period residuals might be correlated across firms.

46
Q

List and explain the computational approaches from statistics found to have
wide applications in investment management.

A
  • Linear models -> Inference, Attribution, Prediction
  • Std. error adjustments -> Attribution
  • Time-series models -> Prediction
  • Bayesian models -> Inference, attribution, prediction
  • Mean-Varians optimization -> diversification, prediction.
  • Nonparametric models -> Inference, Attribution, Prediction
  • Semiparametric models -> Inference, Attribution, Prediction
47
Q

List the steps in training a machine learning model to make predictions.

A
  1. Categorize data (supervised or unsupervised)
  2. Exploratory data analysis.
  3. Choose algorithm
  4. Cross validation
48
Q

List the advantages of deep learning over traditional approaches.

A
  • incorporate large and diverse sources of variables
  • capture complex interactions and nonlinearity
  • avoid overfitting
49
Q

List the advantages of machine learning algorithms

A
  • incorporate non-linearities
  • add out-of-sample prediction tasks
  • create large out-of-sample gains.
50
Q

Explain how to choose between machine learning and financial econometric
approaches for specific applications, including which model features influence
model selection.

A

Data should be the starting point of this analysis and choosing between financial econometric approaches and ML depends on its context.

51
Q

List common errors made when applying machine learning techniques
to economics and financial data sets.

A
  • epistemological errors
  • data processing errors
  • classification errors
  • evaluation errors
52
Q

List and explain the guidelines for applying computational algorithms to
research process

A
  • Start project with scientific caution
  • Understand assumptions, methodology decisions
  • Balance underfitting with overfitting (bias-variance trade-off)
  • Understand theoretical/empirical properties of techniques used
  • Recreate the ML analysis with a statistical procedure and reconcile differences
  • Compare approaches to different computational techniques
  • Test computational approach using simulated data.
53
Q

Describe artificial intelligence (AI).

A

ability of machines to exhibit human-like intelligence and a degree of autonomous learning

54
Q
Describe five applications of AI in the areas of portfolio management and
client enablement:
• Automated insights
• Relationship mapping
• Alternative data sets
• Growth opportunities
• Client outreach
A
  • Automated insights: reading earnings transcripts to gauge sentiment.
  • Relationship mapping: identify nonintuitive relationships between securities/indicators.
  • Alternative data sets: study weather/container ship movements to structure hedging strategies.
  • Growth opportunities: website traffic to gauge future growth.
  • Client outreach: smart client outreach and demand generation (social media data)
55
Q
Describe five applications of AI in the areas of front, middle and back-office efficiency
• Operations intelligence
• Powering risk performance
• Reporting and servicing
• On-demand reporting
• Employee insights
A
  • Operations intelligence: using machine learning to automate functions.
  • Powering risk performance: AI/ML to monitor for suspicious transactions.

• Reporting and servicing:
generating reports for clients.

  • On-demand reporting: chatbots and ML used to respond to queries from customers or employees.
  • Employee insights: monitor employee conduct
56
Q

Four Pillars for Transformation of Investment Management Firms: Describe pillar #1: Generating alpha

A

The goal for every active IM is to generate alpha, as technology/data evolves firms that keep pace, or lead, will outperform.

57
Q

Describe the four points that should be considered in adopting alternative data sets.

A
  • Identify the right alternative data type
  • Have integrated data analytics platform
  • Establish fluid data architecture
  • Build a collaborative insights team
58
Q

Four Pillars for Transformation of Investment Management Firms: Describe pillar #2: Enhancing operational efficiency.

A

Managing the cost of operations remains critical in order to survive the waves of new regulations, fee pressures.

59
Q

Four Pillars for Transformation of Investment Management Firms: Describe pillar #3: Improving product content and distribution.

A

AI in IM allows extending distribution models into new markets and customer segments that have been underserved. (AI enabled analytical tools on top of CRM)

60
Q

Describe how intelligent machines can help in

• Creating a seamless client experience

A

Deliver personalized, consistent, and efficient client experience.

61
Q

Describe how intelligent machines can help in

• Optimizing marketing and sales

A

Insights/tools sales team need to more efficiently engage with their clients

62
Q

Describe how intelligent machines can help in

• Creating effective content

A

produce and distribute relevant, high-quality and timely content on demand.

63
Q

Four Pillars for Transformation of Investment Management Firms: Describe pillar #4: Managing risk.

A

AI can bolster compliance and risk management.

64
Q

List six areas of risk management that intelligent machines can help
• Investment compliance management

A
  • identify investment guidelines from source documents
65
Q

List six areas of risk management that intelligent machines can help
• Liquidity risk management

A
  • Identify liquidity events, and based on defined triggers, automate response protocols.
66
Q

List six areas of risk management that intelligent machines can help
• Operational risk management

A
  • Create exception-based dashboards to identify processing errors
67
Q

List six areas of risk management that intelligent machines can help
• Conduct risk

A

Used bad behavior models to identify bad employees.

68
Q

List six areas of risk management that intelligent machines can help
• Regulatory reporting

A

Extract information from regulations to identify new or updated requirements

69
Q

List six areas of risk management that intelligent machines can help
• Reputational risk mgt.

A

Scan horizon to sense potential threats/seize opportunities within a 72 hour window.

70
Q

Contrast the risks faced by early adopters vs. late adopters before implementation.

A

Early adopters: Learning fee

Late adopters: run risk of being left behind

71
Q

Describe the three considerations that can reduce implementation risk.

A
  • engage compliance and risk mgt. framework early to ensure firm/regulatory compliance.
  • identify mature, well-defined, rules based process as candidates for automation and AI.
  • Establish robust testing program.
72
Q

Describe the three risks that require ongoing monitoring

A
  • Operational
  • Regulatory
  • Technological