Calibration and Simulation Best Practices: Multifactor Interest Rate Models for Risk Applications
Calibration and simulation are a critical, but time-consuming process in modern computational finance applications. Through an example Monte Carlo simulation of interest rate models for counterparty credit risk analysis, Kevin highlights best practices for creating and calibrating models, performing simulations, and optimizing code for performance using MATLAB. He shows how single-factor and multifactor models can be calibrated to both current market data and historical data using Kalman filter and state-space modeling and simulates a portfolio of interest rate instruments. He concludes with discussion on how to deploy MATLAB models into enterprise applications for on-demand risk analysis and reporting.
Using MATLAB to Bridge the Gap Between the Portfolio Construction and Trading
This presentation discusses Kissell Research Group’s latest financial research and findings and shows how the firm has been using MATLAB to help portfolio managers and traders bridge the gap between stock selection and portfolio implementation. Robert introduces techniques that use MATLAB to estimate the cost of trading via non-linear regression analysis, construct MI factor scores to assist portfolio managers with their portfolio construction process, and improve the accuracy and efficiency of algorithmic optimizers. Finally, he discusses how Kissell is using MATLAB to construct the next generation of Global Cost Indexes, and how these indices are used to back-test investment ideas and evaluate broker performance, which ultimately leads to higher portfolio returns for the investor.
Topics covered in this presentation include:
- I-Star Cost Index
- Portfolio construction
- Algorithmic optimization
- Back-testing series
- Broker evaluation
(Re)Defining and Managing Diversification
Attilio introduces a precise quantitative metric for diversification. The metric is based on the "Effective Number of Bets.” The bets are "Minimum Torsion Bets,” or a set of uncorrelated factors optimized to closely track the factors used to allocate the portfolio.
He discusses the advantage of the Minimum-Torsion Bets over the traditional approach to diversification based on Marginal Contributions to Risk.
Statistical Analysis of ETF Flows
11:20 a.m.–12:00 p.m.
The rapid growth of exchange traded products (ETP) has generated considerable interest in flows. With over $2 trillion in worldwide assets under management, ETPs are used by sophisticated and retail investors alike to express their views across asset classes and regions. Flows therefore could be very informative about changes in the investor sentiment, yet to the best of our knowledge, there has been no previous systematic statistical analysis of ETP flows across asset classes. In this presentation, Aleksander discusses how his team constructed unique, survivorship-bias free database of daily ETP flow data, and set up the research agenda to understand ETP flow dynamic.
Approximation Methods in Derivatives Pricing
11:20 a.m.–12:00 p.m.
Minqiang reviews various approximation methods that have been proposed to price financial derivatives, including moment matching, PDE perturbation, lower and upper bounds, distributional approximation, and more. These methods are illustrated using various examples, including timer options.
How MATLAB Did the Work of 40 People and Accelerated Investment Strategy Development
Investment strategy development is essentially the manipulation of data to test ideas. In 2005, Aku was hiring analysts to expand the global quantitative arm of his employer. While going through résumés, he had an epiphany: If all data manipulation could be done easily and efficiently using a single tool, there would be no need for large teams. By combining the entire data retrieval, storage, and manipulation process into a single efficient tool, investment strategy development could be both accelerated and simplified dramatically. A few years later, he developed this tool using MATLAB and started his own fund.
Simulink for Asset-Liability Modeling
This presentation details the use of Simulink to solve the problem of asset liability modeling in a pension fund and endowment context. It showcases the use of a non-traditional finance tool to address a traditional finance question.
Global Tactical Asset Allocation and Portfolio Construction with MATLAB
The investment universe has been defined as seven developed economies and three asset classes, which are stocks, government bonds, and foreign currencies. The alpha strategy is a multifactor model based on valuation, growth, price dynamics, and intermarket action. Each relative value view has been updated monthly and stacked together to input to Black-Litterman portfolio optimization. The performance of long-short portfolio has been back-tested and analyzed. Both alpha model and portfolio construction are programmed using MATLAB.
Lean Data Analysis: The Awesome Data Dexterity of MATLAB Desktop
Today's world is increasingly characterized by internet-availability of massive open datasets, repositories teaming with fresh algorithms, around-the-clock market data feeds, and many forms of scalable, web-integrated NoSQL datastores. Joan discusses a "Lean Data Analysis" approach to managing a Big Data project when the tools available include: a desktop equipped with net connectivity, MATLAB, and a freely-available NoSQL datastore (MongoDB). Anecdotal examples and results are demonstrated first-hand, reflecting one data scientist's journey to train and test an improvised ML model over billions of data points and then benchmark-test the fitted model against competing algorithms.
Factor Research and FactSet’s Integration with MATLAB
Quantitative investing challenges investors to analyze empirical results, draw meaningful conclusion, identify and exploit market inefficiencies, and evaluate these findings to be successful.
Join FactSet and learn how you can take advantage of this integration and build flexible models, create powerful time-series analysis, and validate data effectively. This presentation highlights FactSet’s innovative approach to help analyze asset allocation by creating stock selection models FactSet specialists demonstrate how to take these results and integrate them into MATLAB for further analysis.
Operational Risk Modeling
In order to apply the advanced measure approach (AMA), it is necessary to bring together a number of components. The regulatory guidance from Basel names these components—such as internal data, external data, BEICF, and scenario analysis—but stops short of providing detailed guidance on how to use them together to get to an operational value at risk that is inclusive of all these inputs. A key factor in gaining approval from the regulator to use AMA calculations is getting approval for the calculation methodology. Wolters Kluwer Financial Services (WKFS) recommends combining internal loss data and scenario analysis to arrive at an AMA capital measure as described by the Dutta-Babbel Change of Measure (COM) approach. It is, in fact, a robust way to apply the loss distribution approach (LDA).
The WKFS of implementation of the Dutta-Babbel COM model directly deals with problem that some lines of business (LOBs) or event types especially for newly acquired subsidiaries may have data-scarcity-related challenges. WKFS has preconfigured their system to use to use the Dutta-Babbel COM approach to calculation, including a proprietary .xla add-in that invokes the use of the MATLAB Compiler Runtime. This presentation also looks into MATLAB Compiler™ and MATLAB Builder™ EX (for Microsoft® Excel) as a powerful development tools.
Using MATLAB for Trading Strategy Optimization in Electronic Trading Systems
As algorithmic trading becomes more competitive and sophisticated, there is a need for effective and easily implemented automated methods to select trading strategies. Genetic algorithms (GAs) are machine learning tools that are parsimonious with respect to input information and can accommodate new information as it becomes available, making them attractive for trading rule optimization.
This presentation describes a MATLAB application of GA methodology to stock trading and discusses the combination of GAs with other optimization methods. A hybrid of GAs with particle swarm optimization algorithms has been tested, in some cases giving a better performance than GAs alone. Potential areas of improvement and further development are discussed.
Extending the Power and Scalability of MATLAB Computations within Optimization Solutions
FICO® frequently uses MATLAB to develop new types of analytically powered applications. Working with leaders in financial services, they deliver high value solutions that combine predictive analytics, business rules management, and optimization to improve marketing, origination, and collection decisions.
In this session, Horia shows how you can extend MATLAB capabilities with advanced FICO Xpress Optimization Suite solvers to deliver leading edge financial solutions. Learn how FICO Xpress can be accessed from MATLAB and how the recent, tighter integration makes modeling and solving optimization problems even easier. This presentation previews plans to incorporate MATLAB into the FICO Analytic Cloud.
Welcome Address: MATLAB Turns 30
This year marks the 30th birthday of MATLAB and MathWorks. This talk explores the origin of MATLAB, how it was conceived, and how it became the standard for technical computing across engineering and science. Stuart then discusses how computational finance has influenced the evolution of MATLAB. The session concludes with an overview of the latest release of MATLAB and the conference overview.
In this session, learn about the new Mixed-Integer Linear Programming (MILP) solver in Optimization Toolbox. Seth demonstrates how to use the solver to find solutions to problems in which variables represent integer quantities, such as the number of bonds to purchase, or a yes or no (binary) decision.
- Formulating and solving an MILP problem
- Tuning the solver to your problem
- Overview of the solver algorithm
Pricing and Analysis of an Insurance Contract
Learn how to get started prototyping a model for a life insurance contract. In this session, you will learn about using MATLAB tools and built-in functions to quickly see the impact of different assumptions on price, returns, and risk at the product development and enterprise risk management level.
- Stochastic differential equations
- Plotting and optimization routines
- Replicating portfolio methodology
These sessions provide a deep dive into workflows around a specific topic. Master classes, which are not hands-on, are most useful for people with who have experience with the products.
Event-Driven and Object-Oriented Programming with MATLAB
11:20 a.m.–12:40 p.m.
Are you ready to take your MATLAB programming skills to the next level? Object-oriented programming in MATLAB enables you to simplify the design of complex algorithms and applications by giving you the ability to define relationships between entities, incorporate automatic error checking, greatly reduce memory requirements, and more. Furthermore, event-driven programming enables you to build applications that respond to real-time events, such as market quotes and streaming data. In this master class you will learn how to harness the power of object-oriented and event-driven programming techniques through real-world examples in complex contract pricing and real-time trading. Techniques covered include:
- Creating classes, properties, and methods to represent entities
- Enabling pass-by-reference semantics to improve data efficiency
- Automating error checking and documentation generation
- Customizing behavior of operators
- Writing event handlers for processing streaming data
Time-Series Modeling with MATLAB
Time-series modeling techniques based on computational statistics are often used for financial analysis tasks such as forecasting, pricing complex instruments, and gaining economic insights. However, implementing and comparing modeling techniques to choose the best approach can be challenging.
In this session, you will learn about econometric modeling techniques available in MATLAB, and how to perform data preprocessing and cleanup, select models and evaluate their performance, compare results, and apply the best techniques for your problem.
- Multivariate linear regression techniques in time-series analysis
- Automated predictor selection and cross-validation
- Residual analysis – diagnostics for autocorrelation and heteroscedasticity
- Dynamic model construction using autoregressive and moving average lag variables
Optimization in MATLAB
In this session you will learn about the different tools available for optimization in MATLAB. Seth demonstrates how you can use Optimization Toolbox™ and Global Optimization Toolbox to solve a wide variety of optimization problems. You will learn best practices for setting up and solving optimization problems, as well as how to speed up optimizations with parallel computing.
- Solving linear, nonlinear, and mixed-integer optimization problems
- Tuning solver options to increase performance
- Using Symbolic Math Toolbox™ to automatically calculate gradients
- Speeding up an optimization with parallel computing
These free, hands-on classes are a sample of MathWorks training offerings and allow attendees to use the products with the guidance of an experienced MathWorks instructor.
11:20 a.m.–12:40 p.m.
This hands-on session provides an introduction to the MATLAB technical computing environment for financial professionals. Themes of data import, analysis, visualization, and modeling are explored throughout the session. Topics include:
- Importing data from spreadsheets
- Representing financial data in MATLAB
- Visualizing data
- Accessing subsets of data
Probability Distributions and Regression in MATLAB
Gain hands-on experience using the functionality in Statistics and Machine Learning Toolbox™ to fit probability distributions to a data set and perform predictive modeling by fitting a linear model to a data set in this session. Topics include:
- Fitting probability distributions to data
- Comparing multiple fits
- Fitting linear models to data
Writing Robust MATLAB Code
Experience writing robust applications by using built-in MATLAB functions, programming constructs, and employing standard techniques for handling error conditions. Topics include:
- Creating flexible function interfaces
- Checking for warning and error conditions