CAL

## Tweedie GLM Interpretations and Stochastic Loss Development

I recently had the opportunity to participate in UC Berkeley's Actuarial case competition, hosted by the Cal Actuarial League. The topics of this case competition are Property & Casualty and General Analytics. The full case competition debrief is available here (.pdf), and a copy of the starter files is available here (.zip). Students were allotted roughly 2 weeks (Oct 23 - Nov 8) to complete the following tasks. Property & . . .

• Daniel Suryakusuma
Actuarial

## Markov Chains, Driver Risk Classification, Generating Functions, and Extinction Probability

Markov chains are essential in initiating the study of stochastic processes. A typical example is a random walk, where the transition (probabilities) of going from one state to the next depends only on the current state (location), with no relation to the past. That is, we can ignore and throw away the past history as we continue to take steps forward in the time horizon. Let's look at some examples . . .

• Daniel Suryakusuma
Berkeley

## Numerical Analysis, Predictor Corrector Methods, and Iterative Improvement

The study of numerical methods and machine algorithms for systems modeling and computations exhibits widely applicable themes reaching far and beyond delivering outputs to calculator operations. Such themes and concepts are made explicit in Numerical Analysis (Math 128A) at UC Berkeley, which I had the pleasure of taking under the instruction of Professor John Strain. Because lectures by Strain are orthogonal (but complementary) to many ideas in three textbooks used . . .

• Daniel Suryakusuma
Projects

## Insuring Against Going Viral: Life Expectancy on Spotify Top Charts

This project showcases a spectrum of technical skills in a nontraditional actuarial setting. Many methods are known for analyzing and predicting movement in a financial market and quantifying risk. We seek to give similar treatment in the case of Spotify's Top Charts, where popularity "drives up" the ranking (price) of a given index (song). In particular, we draw special attention to Queen's "Bohemian Rhapsody". Put concisely, in all of Spotify's . . .

• Daniel Suryakusuma
Projects

## LiveType LaTeX: A Library and Guide to Rapid Typesetting in a Technical Lecture Setting

Live demos (.gif / video) will be added to this article as time permits, as I'm still developing and improving my definitions and snippets. There will be more added over time (especially for diagrams and plots). Please see my github for the current development!Examples of my notes, created in-class during lectures. Here's the current version of an ongoing collaboration with John-Michael Laurel for Stochastic Processes by Professor Jim Pitman. Math . . .

• Daniel Suryakusuma
Berkeley

## Uniform Approximation, Bernstein Polynomials (Real Analysis)

Motivation for Bernstein Polynomials:A canonical method to approximating a function is by polynomial interpolation at a specified number of points. We study the asymptotic behavior of particular interpolation schemes in Math 128A, Numerical Analysis. However, we note that interpolation can yield wildly inaccurate results near the endpoints of interpolation (Runge Phenomenon), so we may instead want a sequence of polynomials that approximates our desired continuous function uniformly. As it . . .

• Daniel Suryakusuma
Projects

Suppose we want to create a simple Life Insurance pricing model, as presented in Actuarial Mathematics. Given a set of assumptions, we wish to find the expected value of Life Insurance benefits, discounted to present value. Suppose then we have the following assumptions: Life Insurance Policy A: Pays a flat benefit amount of $50,000 in case of death (up to age 65) Life Insurance Policy B: Pays X% of . . . • Daniel Suryakusuma 1 min read Actuarial ## Basic Ratemaking: Pricing of Insurance Products (Werner, Modlin, CAS) My notes to be transposed here (along with written scans of notes). See the full CAS document here. Contents: 1 - Introduction 2 - Rating Manuals 3 - Ratemaking Data 4 - Exposures 5 - Premium 6 - Losses and LAE 7 - Other Expenses & Profit 8 - Overall Indication 9 - Traditional Risk Classification 10 - Multivariate Classification 11 - Special Classification 12 - Credibility 13 - Other . . . • Daniel Suryakusuma 1 min read Actuarial ## Reserving: Estimation of Unpaid Claims (Friedland, CAS) Here I've compiled outlined notes of the CAS Exam 5 text on Reserving by Friedland. It is my hope that these notes be of some use to other CAS candidates or curious students in gaining more insight and understanding across a spectrum of topics within insurance, from the claims process to various basic techniques for estimating unpaid claims. The contents (whose formatting will be fixed) are as follows. Additionally, my . . . • Daniel Suryakusuma 13 min read Berkeley ## Linear Algebra This Spring 2019 semester I've completed the second course in Berkeley's Linear Algebra series (Math 110) with Professor Olga Holtz. Traditionally at Cal, students' first encounter with linear algebra takes form of computational linear algebra (Math 54), with a heavy focus on factorization and applications in solving linear problems. In this course, we take a purist approach (as done by course textbook author Axler) to not employ determinants of matrices . . . • Daniel Suryakusuma 1 min read Exams ## Weighted Average Cost of Capital, (Un)levered Beta "A company must finance its assets with capital. There are two primary sources of capital: debt and equity. That's how you end up with the fundamental accounting equation:$A=L+E$. "$L\$" stands for liabilities, which is just another name for debt. The capital providers collectively own all the assets. The debt holders and equity holders own the assets. They provided the funds to buy the assets. When the company . . .

• Daniel Suryakusuma
Finance

## Sustainable Investing and the Cross-Section of Stocks Risks and Returns

AbstractThe identification of factors that predict the cross-section of stock returns has been a focus of asset pricing theory for decades. We address this challenging problem for both equity performance and risk, the latter through the maximum drawdown measure. We test a variety of regression-based models used in the field of supervised learning including penalized linear regression, tree-based models, and neural networks. Using empirical data in the US market from . . .

• Daniel Suryakusuma
Finance

## Endogenous Risk, Indirect Contagion, and Systemic Risk

Rama Cont, University of Oxford Abstract: Deleveraging by financial institutions in response to losses may lead to contagion of losses across institutions with common asset holdings. Unlike direct contagion via counterparty exposures, this channel of contagion - which we call indirect contagion - is mediated through market prices and does not require bilateral exposures or relations. We show nevertheless that indirect contagion in the financial system may be modeled as . . .

• Daniel Suryakusuma
Projects

## Research Questions: Topics of Interest

Ranging over different lines of business and applications, these are some of my topics of interest for independent research projects: Analyses on personal health data gathered by wearables; what indicators and variables are missing to construct a better understanding of the user?How can actuarial practices (such as generalized linear models or loss triangles) be extended to other data-science and machine learning motivations such as predictive modeling for sales and . . .

• Daniel Suryakusuma
Actuarial

## Educating Children in Actuarial Science

As a kid growing up, I was only told math could be applied into engineering, else it would be pure math for the purpose of teaching mathematics. To help address this sort of problem, Coaching Actuaries put together an absolutely adorable actuary-themed alphabet book. It's so cute, I'd even recommend grown-ups to have a glance and am including it here for that reason. Full credits go to Coaching Actuaries! For . . .

• Daniel Suryakusuma
Projects

## Tweedie Distribution & the Generalized Linear Model

Tweedie distributions are a special case of exponential dispersion models and are particularly useful in generalized linear models, as in fitting claims data to statistical distributions. We use exponential dispersion models (and particularly the Tweedie distribution) for pure premium approaches for actuarial estimations. There are particular cases where the Tweedie compound Poisson distribution is suitable and appropriate for a given regression. See here for a useful overview on using a . . .

• Daniel Suryakusuma
EPS C82: Oceans

## On Remote Sensing and Bathymetry

This blog entry is the third of four installments in a series of blog posts for a class assignment.   One of the earliest assignments for Oceans (EPS C82) at UC Berkeley involved using Google Earth to visualize depths and heights at certain locations around the world, as well as visualizing the relative locations of the oceans and seas around the world. It’s important sometimes to have a view of . . .

• Daniel Suryakusuma
Health and Benefits

## Health & Benefits: A First Look

Health & Benefits - A First Look Today I attended a CAL informational presentation from Mercer from their San Francisco practice, which has about 100 people in their Health & Benefits team, and about 30 people on the Wealth management team. Presenters included Tim Oakes, Wendy Tan, and Alexa Garzelli on the Health & Benefits side, and Deidre Schelin and Marc Corbeil on the Wealth side. As it stands, Mercer's . . .

• Daniel Suryakusuma