Time series analysis on sensor data

[bshafiei]

Hello, Is it possible to analyze data with low granularity (i.e. min by min) or they should be summarized before using them for time series forecasting using the notebooks provided in Dataiku.

[Alex_Combessie]

Hi,

This is possible using some code along with the visual ML component of Dataiku.

You can use this example: http://gallery.dataiku.com/projects/TIMESERIES/flow/ as a starting point. The code element would mostly deal with slicing windows of the data, since ML models would only forecast one step ahead.

Let me know if this helps,

Alex

[bshafiei]

This is a great guide for Univariate Time Series time. However, I need to perform Multivariate Time Seriese Forecasting. Any suggestions on univariate ts forecasting using i.e. Vector Auto Regression (VAR)?

[Alex_Combessie]

In Python I would advise using https://www.statsmodels.org/dev/vector_ar.html. In R I would advise using https://otexts.org/fpp2/VAR.html.