A New Philosophy of Risk Model Implementation at HSBC


Accelerating Model Development Across Risk-Aware Institutions


The financial services industry has been a hub for new technologies, but also an industry notable in recent decades for devastating modelling and implementation errors. In 2016, old technologies continue to blend with the new in computational finance. On one hand, bastions of conservatism maintain slow, bureaucratic, and risky development. On the other hand, industry disruptors excite about the proliferation of tools offering seemingly free and easy opportunities, but these tools carry a multitude of risks, legal problems, and inefficiencies. Meanwhile, increasing numbers of practitioners on buy and sell sides and in insurance, financial technology, and government effectively manage and take risk using MATLAB® based analytics support systems.

This keynote session explores why and how financial innovators, aided by MATLAB enhancements, are taking MATLAB further and faster into their technology stacks.

Jos Martin, MathWorks

Machine Learning and Visualisation in the Context of a Large Enterprise


In this presentation, Arjun discusses specific, practical, and fun ways to get benefits out of machine learning in your organisation. As well as discussing his experiences with machine learning and visualisation, Arjun talks about gamification and organisational structures, and briefly considers the impact of AI on human society.

Disclosure: This presentation uses public datasets and original methods. No Citi-proprietary techniques or datasets are detailed here.

Portfolio Allocation with Machine Learning and MATLAB Distributed Computing Server on Microsoft Azure Cloud


Emilio and James present the approach developed at Aberdeen Asset Management for the practical implementation of machine learning to analyse financial market trends, in order to generate tactical trades on multi-asset class portfolios. The intensive use of computing power to build solid tests about the validity of the design involved high-performance computing. Microsoft® Azure™ and MATLAB® were the tools of choice to produce and accelerate the process.

Emilio Llorente-Cano, Aberdeen Asset Management


James Mann, Aberdeen Asset Management

LoPeZ: A MATLAB Application to Optimise and Visualise Gazprom's LNG Portfolio


In this presentation, Denis Zuev discusses how Gazprom Marketing & Trading uses modelling and optimization to search for new trading opportunities and manage risk in a constantly changing environment. Denis describes the variables and constraints that Gazprom operates under, how MATLAB® is used to model different scenarios and perform portfolio optimisation, and what conclusions can be drawn. Denis presents recent developments to this approach including interactive visualisations and the use of parallel computing.

Denis Zuev, Gazprom Global LNG

Microeconometrics with MATLAB


There are many domains in which a sophisticated understanding of human behaviour is required for optimal decision making. Yet, while we increasingly have access to vast amounts of data on the choices made by individuals and firms, it remains a challenge to make statements about the underlying causal mechanisms. Microeconometrics is concerned with the analysis of microdata to recover the causal incentive structures that individuals face. These techniques are applied by economists to answer a wide range of questions about behaviour in a diverse set of circumstances: from customers’ choosing breakfast cereals and applying for mortgages to multinationals’ entering new product markets. This talk introduces some key concepts in microeconometrics and demonstrates the capabilities of MATLAB® in this area, focussing especially on discrete choice modelling and nonparametric estimation techniques. The talk builds on Abigail Adams’ new OUP textbook, Microeconometrics and MATLAB: An Introduction, originally conceived while teaching a computational econometrics course for graduate students at the University of Oxford.

Abigail Adams, University of Oxford

MATLAB: The Engine of Goal Monitor


In this presentation, Trevin Lam discusses how Rabobank chose MATLAB® as the engine for Goal Monitor, a portfolio monitoring and forecasting application servicing retail and private clients. Trevin discusses the simulation tasks for which Rabobank uses MATLAB, then how they implemented the routines for external use with MATLAB Production Server™, saving €2m.

Trevin Lam, Rabobank

Supporting Data Science Workflows with MATLAB


You cannot fail to notice the excitement around big data and data science. Do you consider these terms to represent marketing hype, technological revolution, or modest evolution? Your thoughts likely depend on your role, perhaps as a researcher prototyping algorithms, as part of a supporting IT project team scaling models, or as a developer implementing models into your (big) data environment. Maybe you are an analyst, executive, project manager, or other stakeholder, creating or applying “insight” from those models. As a team, you likely care that your (big data) analytics are reliable, tested, and optimized. In this presentation, we examine how MATLAB® can be used by multiple personas within a collaborative data science workflow, and on the way introduce and demonstrate popular themes such as deep learning.

Optimisers Everywhere


Optimisation is equation-fuelled and integral to modelling and model calibration. It is also important to quant and data science workflows. However, optimisation technologies are often overlooked, particularly within data science literature. In this session, Sarah shares her thoughts on how and which optimisation techniques can add significant benefits to research and production workflows in risk, trading, macroeconomics, portfolio management, and financial technology (“FinTech”) scheduling and operations management. Sarah also discusses the applicability of robust optimisation to asset allocation applications.

Sarah Drewes, MathWorks

MATLAB App and Toolbox Development


As you create larger and more sophisticated applications in MATLAB®, you need to manage the additional complexity that comes with more code and more developers, while continuing to provide uncomplicated programming interfaces for analysts and occasional users. Important considerations include:

  • Grouping related data and functions into classes
  • Separating algorithmic code from user interface code
  • Creating reusable graphical components
  • Sharing your application with other people

In this master class, David discusses some useful design and implementation patterns and highlights several relevant language features of the MATLAB object system. He discusses how and where these techniques have added capability in financial analytics stacks while reducing model implementation risks.

David Sampson, MathWorks

Predictive Analytics, Machine Learning, and Regression


Financial decision-making benefits from predictive modelling techniques, including forecasting, classifying risk, estimating default probabilities, and data mining. However, implementing and comparing modelling techniques to choose the best method can be challenging, particularly as data of varying quality continues to be generated from multiple structured and unstructured sources.

In this session, Paul compares different types of predictive modelling and machine learning techniques in MATLAB® and discusses techniques for evaluating model performance. Highlights include:

  • Supervised and unsupervised machine learning techniques and neural networks
  • Multivariate regression techniques for time series analysis
  • Automated predictor selection, cross-validation, and residual diagnostics analysis
  • Graph theory for visual and statistical representation of networks
Paul Peeling, MathWorks

Building, Scaling, and Implementing Risk Model and Stress Test Frameworks


Many quants and risk managers service the complex and bureaucratic calendars of CCAR, Basel, Solvency II, and the EBA and Bank of England Stress-Tests. In this presentation, Bet discusses how risk teams can build an integrated and efficient risk management and stress testing stack to:

  • Generate scenarios, build, scale, and apply models
  • Communicate regulatory and economic capital calculations to CROs, compliance officers, and other stakeholders