Kovács, Márton András (Citi)
Generating Mid Prices with the
Kalman Filter
In the era of Big Data, it
might seem strange that in some areas of finance, the lack of reliable
information is a still a challenge. Some of the products we want to model can
stop being traded for days or weeks. Also, our goal is possibly to predict a
price not directly observable in the market. On top of all
this, consider that there are potentially thousands of products to price
at every moment. We will introduce and explain the mechanics of a linear Kalman
filter. We will discuss its strengths – e.g. its explainability,
its stepwise nature, its ability to easily and omptimally
combine numerous information sources, etc. But we will highlight its
shortcomings as well – its high dimensionality, its linearity, its tendency to lag behind, etc. Furthermore, we will propose methods of
expanding it to counter such shortcomings. Along the way, we will try – and
most probably fail – to tackle some fundamental questions of modelling – what
exactly do we want to model? How do we know if our model is ‘correct’? What
does being ‘correct’ mean? Do we even care?, etc.
The talk is held in English!
Az elõadás nyelve angol!
Date: Nov 13, Tuesday 4:15pm
Place: BME, Building „Q”, Room QBF13