Session 2: how to transform a systematic inflow of data to insights. This transformations are
wide spread in the financial service sector.
1
FSA Lecture 2
,This slide shows you the problem of a financial services company in the digital area. Every
second new data comes in. That’s why we index data with the index of time. The data
available at time t is indicated as DATAt. That data is something that you observe but it’s not
the data in its self that you find interesting, it’s the signal that you extract from the data that
you want to know. A signal could be: is this person worthy of a new loan and how risky is this
person based on the risk perceived? You can then do riskless pricing in terms of charging a
higher interest rate when your client is more risky. The signal its self you don’t observe it
directly, it’s somewhere in the data due to the presents of noise. You could model data as
being the sum of the signal at time t and the noise at time t. The data science that we want
to do is to extract the signal from that data and that signal should lead to an actional insight
on day t such as offering a loan to a client with a specific value for the interest rate.
2
FSA Lecture 2
,First example: the data that we analyze is price data. We take the view point of a technical
trader, who wants to know if he should invest or not in the risky asset. He observes the
current prices and the past prices. What he wants to know is the future price. The future
price (the price at time t + 1) is the current price at time t + some drift, the expected change
in the price (µ) + noise. He wants to know whether the next price will be above the current
price. If he thinks it’s going to be a higher price then today, then he may say I’m going to
invest in the asset or I’m going stay invested in the asset.
How to do this? Prices are noisy and if you want to extract a signal from the noisy prices, we
have to kill the noise. Killing the noise in statistics is typically done by taking averages.
3
FSA Lecture 2
, Here do we see one type of average, it’s the average price at the current date and the 9
perceiving dates. We have 10 prices. The average of this price is called the simple average.
We can compute this average for each time t we will have so called moving average. Because
we move through time and at every time point we have a different price. This average will be
smoother than the original price because of the fact that we take the average.
The signal from this is that we get some price trend by taking the moving average.
4
FSA Lecture 2
wide spread in the financial service sector.
1
FSA Lecture 2
,This slide shows you the problem of a financial services company in the digital area. Every
second new data comes in. That’s why we index data with the index of time. The data
available at time t is indicated as DATAt. That data is something that you observe but it’s not
the data in its self that you find interesting, it’s the signal that you extract from the data that
you want to know. A signal could be: is this person worthy of a new loan and how risky is this
person based on the risk perceived? You can then do riskless pricing in terms of charging a
higher interest rate when your client is more risky. The signal its self you don’t observe it
directly, it’s somewhere in the data due to the presents of noise. You could model data as
being the sum of the signal at time t and the noise at time t. The data science that we want
to do is to extract the signal from that data and that signal should lead to an actional insight
on day t such as offering a loan to a client with a specific value for the interest rate.
2
FSA Lecture 2
,First example: the data that we analyze is price data. We take the view point of a technical
trader, who wants to know if he should invest or not in the risky asset. He observes the
current prices and the past prices. What he wants to know is the future price. The future
price (the price at time t + 1) is the current price at time t + some drift, the expected change
in the price (µ) + noise. He wants to know whether the next price will be above the current
price. If he thinks it’s going to be a higher price then today, then he may say I’m going to
invest in the asset or I’m going stay invested in the asset.
How to do this? Prices are noisy and if you want to extract a signal from the noisy prices, we
have to kill the noise. Killing the noise in statistics is typically done by taking averages.
3
FSA Lecture 2
, Here do we see one type of average, it’s the average price at the current date and the 9
perceiving dates. We have 10 prices. The average of this price is called the simple average.
We can compute this average for each time t we will have so called moving average. Because
we move through time and at every time point we have a different price. This average will be
smoother than the original price because of the fact that we take the average.
The signal from this is that we get some price trend by taking the moving average.
4
FSA Lecture 2
