![65
AI and Machine Learning in Production
https://www.itnews.com.au/news/hsbc-societe-generale-run-
into-ais-production-problems-477966
Kristy Roth from HSBC:
“It’s been somewhat easy - in a funny way - to
get going using sample data, [but] then you hit
the real problems,” Roth said.
“I think our early track record on PoCs or pilots
hides a little bit the underlying issues.
Matt Davey from Societe Generale:
“We’ve done quite a bit of work with RPA
recently and I have to say we’ve been a bit
disillusioned with that experience,”
“the PoC is the easy bit: it’s how you get that
into production and shift the balance”](https://image.slidesharecdn.com/mlmasterclass-cfapoland-201022125826/85/Ml-master-class-cfa-poland-65-320.jpg)
Uploaded byQuantUniversity
Ml master class cfa poland
Machine Learning and AI in Finance was presented by Sri Krishnamurthy. The presentation covered key trends in AI, machine learning and fintech. It provided an intuitive introduction to AI and ML with case studies. The agenda included an overview of key trends, an introduction to AI and ML, case studies and a Q&A session. The slides were available online for attendees to access.
More Related Content
Similar to Ml master class cfa poland
![제 23회 보아즈(BOAZ) 빅데이터 컨퍼런스 - [MBOAX] : ABSA를 활용한 소비자 반응 분석 기반 운영 효율화 대시보드 설계](https://cdn.slidesharecdn.com/ss_thumbnails/3-1boaz23rdconferencemboax-260203102709-9d519923-thumbnail.jpg?width=640&height=640&fit=bounds)
Ml master class cfa poland
- 1. Machine Learning andAI in Finance 2020 Copyright QuantUniversity LLC. Presented By: Sri Krishnamurthy, CFA, CAP sri@quantuniversity.com www.quantuniversity.com 10/22/2020 CFA Society Poland
- 2. 2 Speaker bio • Advisoryand Consultancy for Financial Analytics • Prior Experience at MathWorks, Citigroup and Endeca and 25+ financial services and energy customers. • Columnist for the Wilmott Magazine • Author of forthcoming book “The Model-Driven Enterprise” • Teaches AI/ML and Fintech Related topics in the MS and MBA programs at Northeastern University, Boston • Reviewer: Journal of Asset Management Sri Krishnamurthy Founder and CEO QuantUniversity
- 3. 3 QuantUniversity • Boston-based DataScience, Quant Finance and Machine Learning training and consulting advisory • Trained more than 1000 students in Quantitative methods, Data Science and Big Data Technologies using MATLAB, Python and R • Building a platform for AI and Machine Learning Exploration and Experimentation
- 4. 1. Key trendsin AI, Machine Learning & Fintech 2. An intuitive introduction to AI and ML 3. Case studies 4. Slides at: 5. https://academy.qusandbox.com/#/market/5f91612b99aa4a2469 1da7ef 6. Use Code: CFAPoland as registration code Agenda
- 5. AI and MachineLearning in Finance
- 6. 6 The 4th Industrialrevolution is Here! Source: Christoph Roser at AllAboutLean.com As per Wikipedia*, “The 4th Industrial Revolution ….. marked by emerging technology breakthroughs in a number of fields, including robotics, artificial intelligence, nanotechnology, quantum computing, biotechnology, the Internet of Things, the Industrial Internet of Things (IIoT), decentralized consensus, fifth-generation wireless technologies (5G), additive manufacturing/3D printing and fully autonomous vehicles.” * https://en.wikipedia.org/wiki/Fourth_Industrial_Revolution
- 7. 7 Scientists are disruptingthe way we live! Source: https://www.ladn.eu/tech-a-suivre/mobilite-2030-vehicules-volants-open-data/
- 8. 8 Interest in Machinelearning continues to grow https://www.wipo.int/edocs/pubdocs/en/wipo_pub_1055.pdf
- 9. 9 MACHINE LEARNING ANDAI IS REVOLUTIONIZING FINANCE
- 10. 10 Market impact atthe speed of light! 10
- 11. 11 • Machine learningis the scientific study of algorithms and statistical models that computer systems use to effectively perform a specific task without using explicit instructions, relying on patterns and inference instead1 • Artificial intelligence is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and animals1 Defining Machine Learning and AI 11 1. https://en.wikipedia.org/wiki/Machine_learning 2. Figure Source: http://www.fsb.org/wp-content/uploads/P011117.pdf
- 12. 12 Machine Learning &AI in finance: A paradigm shift 12 Stochastic Models Factor Models Optimization Risk Factors P/Q Quants Derivative pricing Trading Strategies Simulations Distribution fitting Quant Real-time analytics Predictive analytics Machine Learning RPA NLP Deep Learning Computer Vision Graph Analytics Chatbots Sentiment Analysis Alternative Data Data Scientist
- 13. 13 The Virtuous Circleof Machine Learning and AI 13 Smart Algorithms Hardware Data
- 14. 14 The rise ofBig Data and Data Science 14 Image Source: http://www.ibmbigdatahub.com/sites/default/files/infographic_file/4-Vs-of-big-data.jpg
- 15. 15 Smart Algorithms 15 Distributing ComputingFrameworks Deep Learning Frameworks 1. Our labeled datasets were thousands of times too small. 2. Our computers were millions of times too slow. 3. We initialized the weights in a stupid way. 4. We used the wrong type of non-linearity. - Geoff Hinton “Capital One was able to determine fraudulent credit card applications in 100 milliseconds”* * http://go.databricks.com/hubfs/pdfs/Databricks-for-FinTech-170306.pdf
- 16. 16 Hardware Speed up calculationswith 1000s of processors Scale computations with infinite compute power
- 17. 17 “Financial Technologies or“Fintech” is used to describe a variety of innovative business models and emerging technologies that have the potential to transform the financial services industry ” Technology drives finance! https://www.iosco.org/library/pubdocs/pdf/IOSCOPD554.pdf
- 18.
- 19.
- 20.
- 21.
- 22.
- 23. Risk Systems ThatRead® • Northfield uses machine learning based analysis of news text to describe how current conditions in financial markets are different than usual. • Typically, over 8000 articles per day containing more than 20,000 “topics” (companies, industries, countries) are processed. • The nature and magnitudes of these difference are used to revise expectations of financial market risks for all global equities and credit instruments on a daily basis.
- 24. 24 1. Leveraging largeand diverse datasets for Investment decision making at J.P. Morgan1 2. Improving Quantitative investing at AQR2 3. Using Sandboxes and labs to further innovation in fintech at Fidelity3 4. Use of AI and ML increasing in ssset management from idea generation to execution - Wells Fargo4 Additional Use cases 1. https://www.jpmorgan.com/global/cib/research/investment-decisions-using-machine-learning-ai 2. https://www.aqr.com/Learning-Center/Machine-Learning 3. https://www.fidelitylabs.com/ 4. https://www08.wellsfargomedia.com/assets/pdf/personal/investing/investment-institute/IG_Machines_Are_Coming_ADA.pdf
- 25.
- 26. 26 • Automation toincrease • Digital transformation and move to the cloud finally happening • Use of Synthetic data to increase • Edge cases of AI put to truth test! • Fintechs feeling the pressure to prove themselves! • Human-in-the-loop AI to regain focus! The changes have been drastic and sudden! What’s in store for the industry is yet to be seen! What does Covid2019 mean to adoption of AI and ML in Financial services?
- 27.
- 29. 29 Let’s get underthe hood 29 Source: https://www.pikrepo.com/fcsda/yellow-hot-rod-car-with-hood-open
- 30. Machine Learning Workflow DataScraping/ Ingestion Data Exploration Data Cleansing and Processing Feature Engineering Model Evaluation & Tuning Model Selection Model Deployment/ Inference Supervised Unsupervised Modeling Data Engineer, Dev Ops Engineer Data Scientist/QuantsSoftware/Web Engineer • AutoML • Model Validation • Interpretability Robotic Process Automation (RPA) (Microservices, Pipelines ) • SW: Web/ Rest API • HW: GPU, Cloud • Monitoring • Regression • KNN • Decision Trees • Naive Bayes • Neural Networks • Ensembles • Clustering • PCA • Autoencoder • RMS • MAPS • MAE • Confusion Matrix • Precision/Recall • ROC • Hyper-parameter tuning • Parameter Grids Risk Management/ Compliance(All stages) Analysts& DecisionMakers
- 31. 31 1. Data 2. Goals 3.Machine learning algorithms 4. Process 5. Performance evaluation Key steps involved
- 33. 33 Dataset, variable andObservations Dataset: A rectangular array with Rows as observations and columns as variables Variable: A characteristic of members of a population ( Age, State etc.) Observation: List of Variable values for a member of the population
- 34. 34 Variables A variable couldbe: ▫ Categorical – Yes/No flags – AAA,BB ratings for bonds ▫ Numerical – 35 mpg – $170K salary
- 35. 35 Longitudinal ▫ Observations aredependent ▫ Temporal-continuity is required Cross-sectional ▫ Observations are independent Datasets
- 36.
- 38. 38 • Descriptive Statistics ▫Goal is to describe the data at hand ▫ Backward-looking ▫ Statistical techniques employed here • Predictive Analytics ▫ Goal is to use historical data to build a model for prediction ▫ Forward-looking ▫ Machine learning & AI techniques employed here Goal 38
- 39. 39 • Given adataset, build a model that captures the similarities in different observations and assigns them to different buckets- Clustering • Given a set of variables, predict the value of another variable in a given data set- Prediction ▫ Predict salaries given work experience, education etc. ▫ Predict whether a loan would be approved given fico score, current loans, employment status etc. Predictive Analytics : Cross sectional datasets 39
- 40. 40 Goal Descriptive Statistics Cross sectional Numerical Categorical Numerical vs Categorical Categoricalvs Categorical Numerical vs Numerical Time series Predictive Analytics Cross- sectional Segmentation Prediction Predict a number Predict a category Time-series Summary 40
- 42. 42 Machine Learning Unsupervised Supervised ReinforcementSemi-Supervised Machine Learning
- 43. 43 Goal Descriptive Statistics Cross sectional Numerical Categorical Numerical vs Categorical Categoricalvs Categorical Numerical vs Numerical Time series Predictive Analytics Cross- sectional Segmentation Prediction Predict a number Predict a category Time-series Machine Learning Algorithms 43
- 44. 44 Supervised Algorithms ▫ Givena set of variables 𝑥!, predict the value of another variable 𝑦 in a given data set such that ▫ If y is numeric => Prediction ▫ If y is categorical => Classification ▫ Example: Given that a customer’s Debt-to-Income ratio increased 20%, what are the chances he/she would default in 3 months? Machine Learning 44 x1,x2,x3… Model F(X) y
- 45. 45 Unsupervised Algorithms ▫ Givena dataset with variables 𝑥!, build a model that captures the similarities in different observations and assigns them to different buckets => Clustering ▫ Example: Given a list of emerging market stocks, can we segment them into three buckets? Machine Learning 45 Obs1, Obs2,Obs3 etc. Model Obs1- Class 1 Obs2- Class 2 Obs3- Class 1
- 46. 46 • Parametric models ▫Assume some functional form ▫ Fit coefficients • Examples : Linear Regression, Neural Networks Supervised Learning models - Prediction 46 𝑌 = 𝛽! + 𝛽" 𝑋" Linear Regression Model Neural network Model
- 47. 47 • Non-Parametric models ▫No functional form assumed • Examples : K-nearest neighbors, Decision Trees Supervised Learning models 47 K-nearest neighbor Model Decision tree Model
- 48.
- 49. 49 Machine Learning moversand shakers Deep Learning Automatic Machine Learning Ensemble Learning Natural Language Processing Data Robot H20.ai Autosklearn autokkeras Tensorflow Pytorch NLTK HuggingFace Bagging Boosting DNN CNN LSTM GAN
- 50.
- 52.
- 53. 53 • What transformationsdo I need for the x and y variables ? • Which are the best features to use? ▫ Dimension Reduction – PCA ▫ Best subset selection – Forward selection – Backward elimination – Stepwise regression Feature Engineering 53
- 54.
- 56. 56 Evaluating Machine learning algorithms Supervised - Prediction R-squareRMS MAE MAPE Supervised- Classification Confusion Matrix ROC Curves Evaluation framework 56
- 57. 57 • Fit measuresin classical regression modeling: • Adjusted 𝑅! has been adjusted for the number of predictors. It increases only when the improve of model is more than one would expect to see by chance (p is the total number of explanatory variables) 𝐴𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑅! = 1 − ⁄∑"#$ % (𝑦" − 0𝑦")! (𝑛 − 𝑝 − 1) ∑"#$ % 𝑦" − 4𝑦" ! /(𝑛 − 1) • MAE or MAD (mean absolute error/deviation) gives the magnitude of the average absolute error 𝑀𝐴𝐸 = ∑"#$ % 𝑒" 𝑛 Prediction Accuracy Measures
- 58. 58 ▫ MAPE (meanabsolute percentage error) gives a percentage score of how predictions deviate on average 𝑀𝐴𝑃𝐸 = ∑!"# $ 𝑒!/𝑦! 𝑛 ×100% • RMSE (root-mean-squared error) is computed on the training and validation data 𝑅𝑀𝑆𝐸 = 1/𝑛 2 !"# $ 𝑒! % Prediction Accuracy Measures
- 59. 59 1. Data 2. Goals 3.Machine learning algorithms 4. Process 5. Performance Evaluation Recap
- 60. Machine Learning Workflow DataScraping/ Ingestion Data Exploration Data Cleansing and Processing Feature Engineering Model Evaluation & Tuning Model Selection Model Deployment/ Inference Supervised Unsupervised Modeling Data Engineer, Dev Ops Engineer Data Scientist/QuantsSoftware/Web Engineer • AutoML • Model Validation • Interpretability Robotic Process Automation (RPA) (Microservices, Pipelines ) • SW: Web/ Rest API • HW: GPU, Cloud • Monitoring • Regression • KNN • Decision Trees • Naive Bayes • Neural Networks • Ensembles • Clustering • PCA • Autoencoder • RMS • MAPS • MAE • Confusion Matrix • Precision/Recall • ROC • Hyper-parameter tuning • Parameter Grids Risk Management/ Compliance(All stages) Analysts& DecisionMakers
- 61.
- 62.
- 63. 63 Claim: • Machine learningis better for fraud detection, looking for arbitrage opportunities and trade execution Caution: • Beware of imbalanced class problems • A model that gives 99% accuracy may still not be good enough 1. Machine learning is not a generic solution to all problems
- 64. 64 Claim: • Our modelswork on datasets we have tested on Caution: • Do we have enough data? • How do we handle bias in datasets? • Beware of overfitting • Historical Analysis is not Prediction 2. A prototype model is not your production model
- 65. 65 AI and MachineLearning in Production https://www.itnews.com.au/news/hsbc-societe-generale-run- into-ais-production-problems-477966 Kristy Roth from HSBC: “It’s been somewhat easy - in a funny way - to get going using sample data, [but] then you hit the real problems,” Roth said. “I think our early track record on PoCs or pilots hides a little bit the underlying issues. Matt Davey from Societe Generale: “We’ve done quite a bit of work with RPA recently and I have to say we’ve been a bit disillusioned with that experience,” “the PoC is the easy bit: it’s how you get that into production and shift the balance”
- 66. 66 Claim: • It works.We don’t know how! Caution: • It’s still not a proven science • Interpretability or “auditability” of models is important • Transparency in codebase is paramount with the proliferation of opensource tools • Skilled data scientists who are knowledgeable about algorithms and their appropriate usage are key to successful adoption 3. We are just getting started!
- 67. 67 Claim: • Machine Learningmodels are more accurate than traditional models Caution: • Is accuracy the right metric? • How do we evaluate the model? RMS or R2 • How does the model behave in different regimes? 4. Choose the right metrics for evaluation
- 68. 68 Claim: • Machine Learningand AI will replace humans in most applications Caution: • Beware of the hype! • Just because it worked sometimes doesn’t mean that the organization can be on autopilot • Will we have true AI or Augmented Intelligence? • Model risk and robust risk management is paramount to the success of the organization. • We are just getting started! 5. The Robots are coming! https://www.bloomberg.com/news/articles/2017-10-20/automation- starts-to-sweep-wall-street-with-tons-of-glitches
- 69.
- 70. #Disrupt19 Alternative investments: Interestrate predication for Peer-to-Peer Market places using ML techniques
- 71. 71 1. Case Intro 2.Data Exploration of the Credit risk data set 3. Problem Definition and Machine learning 4. Performance Evaluation 5. Deployment Case study
- 72. 72 Credit decisions Credit-scoring modelsand techniques assess the risk in lending to customers. Typical decisions: • Grant credit/not to new applicants • Increasing/Decreasing spending limits • Increasing/Decreasing lending rates • What new products can be given to existing applicants ?
- 73. 73 How Lending clubworks? https://www.lendingclub.com/public/how-peer-lending- works.action
- 74. 74 • How muchshould I expect as interest? • Is my borrower credit worthy? • How much interest would a similar borrower pay? • What is the repayment and default rate for a similar borrower? Investor’s big decisions
- 75.
- 76. 76 Credit Risk pipeline DataIngestion from Lending Club Pre-Processing Feature Engineering Model Development and Tuning Model Deployment Stage 1 Stage 2 Stage 3 Stage 4 Stage 5
- 77.
- 78. #Disrupt19 Synthetic data generationusing Neural Networks
- 79. 79 All scenarios haven’t playedout • Stress scenarios • What-if scenarios Challenges with real datasets Figure ref: http://www.actuaries.org/CTTEES_SOLV/Documents/StressTestingPaper.pdf
- 80. 80 Missing values • Missingat random • Missing sequences • Need data to fill frames Challenges with real datasets
- 81. 81 • Access ▫ Hardto find ▫ Rare class problems ▫ Privacy concerns making it difficult to share Challenges with real datasets
- 82. 82 Imbalanced • Need moresamples of rare class • Need proxies for data points that were not observed or recorded Challenges with real datasets
- 83. 83 Labels • Human labelingis hard • Synthetic label generators Challenges with real datasets
- 84.
- 85.
- 86.
- 87. 87 Demo: Synthetic datageneration Extreme scenario generation
- 88.
- 89.
- 90. 90 1. Case Intro 2.Data Exploration of WIG20 stock data 3. Problem Definition and Machine learning 4. Deployment Case study
- 91. 91 Clustering stocks • Whichstocks are like each other? • Are growth stocks behaving like growth stocks or value stocks? • Does the time series of prices & returns reveal which stocks are close to each other?
- 92. 92 Clustering workflow Data Ingestion Pre- Processing Clustering Visualization & analysis Model Deployment Stage1 Stage 2 Stage 3 Stage 4 Stage 5
- 93. 93 • Clustering ▫ Howdo we define distance between stocks? – Correlation – 1- Correlation2 (https://arxiv.org/abs/cond-mat/9802256) ▫ Hierarchical Clustering Methodology
- 94. 94 • Clustering ▫ Howdo we define distance between stocks? – Covariance Estimation - https://scikit- learn.org/stable/modules/covariance.html#sparse-inverse-covariance ▫ Clustering: Affinity propagation - https://scikit- learn.org/stable/modules/clustering.html#affinity-propagation ▫ Visualizing relationships with manifold learning – https://scikit-learn.org/stable/modules/manifold.html#manifold Methodology
- 95.
- 96.
- 97.
- 98.
- 99.
- 100.
- 101. 101 • Understanding sentimentsin Earnings call transcripts Goal
- 102. 102 • Interpreting emotions •Labeling data Challenges
- 103. 103 What is NLP? AI Linguistics Computer Science
- 104. 104 • Q/A • Dialogsystems - Chatbots • Topic summarization • Sentiment analysis • Classification • Keyword extraction - Search • Information extraction – Prices, Dates, People etc. • Tone Analysis • Machine Translation • Document comparison – Similar/Dissimilar Sample applications
- 105.
- 106. 106 • If computerscan understand language, opens huge possibilities ▫ Read and summarize ▫ Translate ▫ Describe what’s happening ▫ Understand commands ▫ Answer questions ▫ Respond in plain language Language allows understanding
- 107. 107 • Describe rulesof grammar • Describe meanings of words and their relationships • …including all the special cases • ...and idioms • ...and special cases for the idioms • ... • ...understand language! Traditional language AI https://en.wikipedia.org/wiki/Formal_language
- 108. 108 What is NLP? Jumping NLP Curves https://ieeexplore.ieee.org/document/6786458/
- 109. 109 Q: What’s hardabout writing programs to understand text?
- 110. 110 • Ambiguity: ▫ “ground” ▫“jaguar” ▫ “The car hit the pole while it was moving” ▫ “One morning I shot an elephant in my pajamas. How he got into my pajamas, I’ll never know.” ▫ “The tank is full of soldiers.” “The tank is full of nitrogen.” Language is hard to deal with
- 111.
- 112. 112 • Many waysto say the same thing ▫ “the same thing can be said in many ways” ▫ “language is versatile” ▫ “The same words can be arranged in many different ways to express the same idea” ▫ … Language is hard to deal with
- 113. 113 • APIs • HumanInsight • Expert Knowledge • Build your own Options?
- 114. 114 NLP pipeline Data Ingestion fromEdgar Pre-Processing Invoking APIs to label data Compare APIs Build a new model for sentiment Analysis
- 115.
- 116.
- 117. Thank you! Sri Krishnamurthy,CFA, CAP Founder and CEO QuantUniversity LLC. srikrishnamurthy www.QuantUniversity.com Contact Information, data and drawings embodied in this presentation are strictly a property of QuantUniversity LLC. and shall not be distributed or used in any other publication without the prior written consent of QuantUniversity LLC. 117