Markus Schmaus

#introduction

I'm a #mathematician by training and a #datascientist by trade. Living in #munich, #germany. I am a husband to my beautiful wife and father of two amazing children.

I studied math (and some physics) at #LMUMünchen.

I used to work for #mckinsey and #airbnb

I'm interested in #boardgames, #philosophyofscience, #epistemology, #bayesianism, #hierarchicalmodels, #forecasting, #ManyWorlds, #machinelearning, #artificialintelligence, #emergence, #categorytheory

Dec 17, 2022 at 4:04pmWeb
Show threadn-gonsDec 17, 2022
@markus_schmaus Welcome!
Show threadVolker K. 🏳️‍🌈Dec 17, 2022
@markus_schmaus welcome & hope to read from you about your various passions of which some I share 🐱
Show threadLeo de NeviDec 17, 2022
@markus_schmaus welcome 👋
Show threadNiklasRiewaldDec 19, 2022
@markus_schmaus Hi Markus, nice to meet you. I think we share a lot in common. I live in Augsburg, studied computer science and some math and also have two kids. And I'm working in energy forecasting. What kind of forecasting are you especially interested in?
Show threadMarkus SchmausDec 21, 2022
@NiklasRiewald I was responsible for Airbnb's global business forecast for 7 years. Key challenge, but also important source of accuracy, was to make sure all the different subcomponents fitted together into a consistent whole. After building up a team that can carry my work forward, I have now left and I am looking into a couple of personal projects.
Show threadNiklasRiewaldDec 23, 2022
@markus_schmaus Hierachical forecasting is such a cool and underappreciated topic. It's also very important for energy forecasting where forecasts for energy plants are aggregated to the portfolio level that's traded on the exchanges. So people are mostly interested in this portfolio level forecast. Hierachical forecasting in this context seems to be often seen as a necessary evil to get a good tradeoff between forecast accuracy and alignment.
Show threadNiklasRiewaldDec 23, 2022
@markus_schmaus But I feel that if done correctly hierachical forecasting can be utilized to improve forecast quality because the alignment requirement is acting as a sort of regularization. Does this match your experience? What are your goto methods for hierachical forecasting?
Show threadMarkus SchmausDec 23, 2022
@NiklasRiewald Yeah, even if I am only interested in the top line forecast it can be very valuable to look into the disaggregated data. On this levels signals may be visible that are washed out on the aggregated level. By capturing those signals I can also improve top line accuracy and by getting a better understanding of the process that generates the data, I get a better sense of possible failure modes of the model and the reliability of the forecast.