Tips & Tricks: Reading FRED API data in RapidMiner
Recently I wanted to load in some XML data from the Federal Reserve Economic Data (FRED) website. I didn’t want to use any extensions, just native RapidMiner operators. The FRED has a great API that lets you pull in all kinds of free economic and financial data (in JSON and XML format) – all you need to do is apply for a free FRED API key.
Accessing this data is really easy in RapidMiner Studio and all it takes is two operators: Open File and Read XML.
In order to get any output, you have to form the access URL for that data. The way the URL is formed will tell the FRED data source to return what you want and in the format you want it. It requires you to enter in your API key and pass some parameters for the asset type, start dates, end dates, etc. You can read more about the FRED URL parameters here.
I was interested in accessing Euro/USD time series data and the only FRED API data available for this currency was the date and closing price (time series). This was sufficient for my needs and the symbol for the Euro/USD currency is DEXUSEU.
The resulting URL that I had to form looked like this:
https://api.stlouisfed.org/fred/series/observations?series_id=DEXUSEU&api_key=myAPIkey
Note: myAPIkey is my personal API key, you’ll have to substitute your own.
When I accessed the URL in my browser, I got the following raw XML results.
So far, so good.
Loading the data into RapidMiner
Now that we have a working URL, the trick is to access it and load it into RapidMiner. The way to do that is to plug the URL into an Open File operator. You set the “Resource Type” parameter to URL and then paste in your URL.
Next you have to use the Read XML operator to be able to parse in the XML data correctly. This operation can be a bit tricky but it’s a lot easier if you use the Read XML Import Configuration Wizard. Through this Wizard you can select the nested XML data and parse it right into RapidMiner, like below.
The Result
Once you’ve gotten this all configured you can hit the Play button. This is the result that is returned.
You can now process the time series downstream by renaming the attributes and using our free Series extension.
There have been some major advancements to the RapidMiner platform since this article was originally published. We’re on a mission to make machine learning more accessible to anyone. For more details, check out our latest release.
