book

Machine Learning for Financial Risk Management with Python

by Abdullah Karasan

December 2021

Intermediate to advanced

331 pages

8h 28m

English

O'Reilly Media, Inc.

Read now

Unlock full access

Conventions Used in This BookUsing Code ExamplesO’Reilly Online LearningHow to Contact UsAcknowledgements
RiskReturnRisk ManagementMain Financial RisksBig Financial CollapseInformation Asymmetry in Financial Risk ManagementAdverse SelectionMoral HazardConclusionReferences
Time Series ComponentsTrendSeasonalityCyclicalityResidualTime Series ModelsWhite NoiseMoving Average ModelAutoregressive ModelAutoregressive Integrated Moving Average ModelConclusionReferences
Recurrent Neural NetworksLong-Short Term MemoryConclusionReferences
ARCH ModelGARCH ModelGJR-GARCHEGARCHSupport Vector Regression: GARCHNeural NetworksThe Bayesian ApproachMarkov Chain Monte CarloMetropolis–HastingsConclusionReferences
Value at Risk (VaR)Variance-Covariance MethodThe Historical Simulation MethodThe Monte Carlo Simulation VaRDenoisingExpected ShortfallLiquidity-Augmented Expected ShortfallEffective CostConclusionReferences
Estimating the Credit RiskRisk BucketingProbability of Default Estimation with Logistic RegressionProbability of Default Estimation with the Bayesian ModelProbability of Default Estimation with Support Vector MachinesProbability of Default Estimation with Random ForestProbability of Default Estimation with Neural NetworkProbability of Default Estimation with Deep LearningConclusionReferences
Liquidity MeasuresVolume-Based Liquidity MeasuresTransaction Cost–Based Liquidity MeasuresPrice Impact–Based Liquidity MeasuresMarket Impact-Based Liquidity MeasuresGaussian Mixture ModelGaussian Mixture Copula ModelConclusionReferences

Getting Familiar with Fraud DataSupervised Learning Modeling for Fraud ExaminationCost-Based Fraud ExaminationSaving ScoreCost-Sensitive ModelingBayesian Minimum RiskUnsupervised Learning Modeling for Fraud ExaminationSelf-Organizing MapAutoencodersConclusionReferences
Stock Price Crash MeasuresMinimum Covariance DeterminantApplication of Minimum Covariance DeterminantLogistic Panel ApplicationConclusionReferences
Synthetic Data GenerationEvaluation of the Synthetic DataGenerating Synthetic DataA Brief Introduction to the Hidden Markov ModelFama-French Three-Factor Model Versus HMMConclusionReferences

Content preview from Machine Learning for Financial Risk Management with Python

Chapter 8. Modeling Operational Risk

...It’s not necessarily the biggest missteps that deliver the biggest blows; share prices can plummet as a result of even the smallest events.

Dunnett, Levy, and Simoes (2005)

Thus far, we have talked about three main financial risks: market, credit, and liquidity risks. Now it is time to discuss operational risk, which is more ambiguous than the other types of financial risks. This ambiguity arises from the huge variety of risk sources by which financial institutions may face huge losses.

Operational risk is the risk of direct or indirect loss resulting from inadequate or failed interval processes, people, and systems or from external events (BIS 2019). Please note that loss can be direct and/or indirect. Some direct losses would be:

Legal liability arising from judicial process
Write-downs due to theft or reduction in assets
Compliance emanating from tax, license, fines, etc.
Business interruption

Indirect cost is related to the opportunity cost in the way that a decision made by an institution may trigger a host of events resulting in a loss at an uncertain time in the future.

Normally, financial institutions allocate a certain amount of money to cover the loss emanating from operational risk, which is known as unexpected loss. However, allocating an appropriate amount of funds to cover unexpected loss is not as easy as it sounds. It is necessary to determine the right amount of unexpected loss; otherwise, either more funds are ...