# Datapreparation for time series

Hi I'd like to predict stock prices with time series?

My input- data would have for example GDP Data on a quaterly basis

Capacity utilisation on a monthly basis

weekly jobless claims

daily oilprices

daily stock/index prices

different opening days for different exchanches

How would I have to prepare such a dataset to use this software?

Many thanks, Vektor

My input- data would have for example GDP Data on a quaterly basis

Capacity utilisation on a monthly basis

weekly jobless claims

daily oilprices

daily stock/index prices

different opening days for different exchanches

How would I have to prepare such a dataset to use this software?

Many thanks, Vektor

Tagged:

0

## Answers

2,531Unicorntime series prediction is a complex task. I cannot go in much detail here. For getting a more extensive introduction, I would recommend to take part on one of our courses or webinars on that topic. Here's what you have to do in short, if you are going to teach it yourself:

Spoken in general terms, you have a multivariate time series with your data. The problem is, that each dimension has different time indices. You will have to unify it in some way, so that the values are at the same time point. It's strongly recommend to have equidistant points in time. After this, you might transform the series in some way to improve the classification performance, mainly deriving relative changes instead of absolute values and something like this.

Then you are going to window the series, deriving examples which are then a snapshot in time. For example each window contains the values of the past 10 days and has a target to predict the stock price for the next day. Then you could use the standard statistical learning methods like linear regression, svm and so on, to perform a regression.

You see it's quite complex and if you can't take part in one of the course, I strongly recommend to make familiar with the operators of the value series extension, as well as with all samples coming with RapidMiner.

Greetings,

Sebastian