Is there a way to analyze TimeSeries-Data with irregular timeSamplingInterval?

Forum|Forum|7 years ago
October 12, 2018
2 replies
3307 views

When uploading Data Sets to Analytics Builder I need to add a meta data file which defines the different fields in the csv-file containing the data.

For the column containing the Time Stamp I need to define a parameter called "timeSamplingInterval" which defines how much time passes between each measurement.

However in my data set, the time intervals often vary between different measurements and the file upload gets rejected because of this, as shown in the attached png.

Is there some way to deal with this or an established workaround solution?
Should I try to use another module/extension to deal with TimeSeries-Data with irregular intervals?

Best answer by eengelbrecht

Hello,

My team has recently had a similar problem and we found a workaround solution. In the meta data file we set the time stamp to be an informational field (dataType: STRING and opType: INFORMATIONAL). We then created an additional field called timecount which is our time series field (dataType: DOUBLE, opType: TEMPORAL and timeSamplingInterval: 10 000).

The timecount field is essentially just and integer field counting the number of rows using a constant value. in Our case, we used a 10 second interval. 10 seconds is roughly the average time it took for us to gather a data sample.

I have attached snippets of our meta data file and our data.

Please see the following community post as well.

https://community.ptc.com/t5/IoT-Tech-Tips/How-to-work-with-ordinal-and-categorical-data-in-ThingWorx/m-p/542406

Hope this helps.

J

jgreiner

13-Aquamarine

Hello,

Timeseries in ThingWorx Analytics typically requires a regular timeSamplingInterval although a little variability is accepted.

For the timeSamplingInterval you could try using the average interval that the data inputs are coming in at as long as there is not too much variability.

If there is a good amount of variability between each input, you could create a separate column called count and increment it to increase by 1 for each column, and then set the the timeSamplingInterval to 1.

Let me know if you have any other questions or concerns.

Warm Regards,

John

E

eengelbrecht

Answer

1-Visitor

Hello,

My team has recently had a similar problem and we found a workaround solution. In the meta data file we set the time stamp to be an informational field (dataType: STRING and opType: INFORMATIONAL). We then created an additional field called timecount which is our time series field (dataType: DOUBLE, opType: TEMPORAL and timeSamplingInterval: 10 000).

The timecount field is essentially just and integer field counting the number of rows using a constant value. in Our case, we used a 10 second interval. 10 seconds is roughly the average time it took for us to gather a data sample.

I have attached snippets of our meta data file and our data.

Please see the following community post as well.

https://community.ptc.com/t5/IoT-Tech-Tips/How-to-work-with-ordinal-and-categorical-data-in-ThingWorx/m-p/542406

Hope this helps.