Upload data format

The ChaosKit web service is configured to load data in XML format, and to load both sequential and time-tagged data. The web service permits the user to define how to locate the individual data items using XPath statements, so a wide range of XML sources may be used.

For this evaluation we have limited the data formats to just one in order to avoid the many potential data formatting questions that might otherwise arise.

The required format looks like this:

<samples>
  <sample>
    <value>0.1</value>
  </sample>
  <sample>
    <value>0.36</value>
  </sample>
  ……
</samples>

Uploading any other formats will result in the upload failing

Time series options

There are several options that can be set when a time series is created and the data loaded.

The series name is an identifier that you choose to reference this series via the selection list. It defaults to the file name.

 

Sample time is the time in seconds that you want to separate each sample in the temporal database inside ChaosKit.  It will also determine the interval between each prediction.

If you select differencing, the difference between adjacent samples, rather than the absolute value of each will be used in processing and prediction. This is often useful in financial time series, where the absolute level of the series is constantly changing.

 

Options only available by using the web service directly

The web service has several facilities that are not available through the evaluation interface.

•          You can load time-tagged values, (in the world of finance this is known as ‘tick data’) with the web service. This is useful whenever data arrives at irregular intervals.
•         You can set trading times with the web service.  Financial time series stop and start at regular intervals. They also break for holidays.  The web service has a mechanism to define these times, weekdays and holidays.
•     Where the time series stored in chaoskit is irregular in time, ChaosKit will try to make regular samples in order to perform embedding and prediction. Where these samples fall between data values in the temporal database, Chaoskit can either (the default) use the previous value, or interpolate between values.
•         You can set dead periods with the web service.   Dead periods are gaps in a time tagged series with no data.  In normal processing, with interpolation set, the temporal database inside ChaosKit will try to interpolate across the gap, thus filling the pipeline with essentially made up data.  Dead periods permit the user to define a maximum gap that the interpolation will not attempt to bridge. Instead it will stop embedding generation at the start of the gap, and restart it at the other end once the pipeline is full.