3.5.1.1. Streams
A data stream is a live data source that delivers instances sequentially. Unlike offline datasets, data instances cannot be scanned on demand in case of streams. Data instances are only available at the order they arrive. For example, think of a live RADIO signal that delivers new data with time, where complete access to entire data is not possible.
As shown in the figure above, at every timestep, new information is available. However, the number of instances delivered at each instance and availability of historical instances depends on the type of stream and the processing task respectively.
In industrial scenarios, with more and more complex systems, the amount of live data delivered by the systems increases rapidly. This high amount of live data can be leveraged to take optimal decisions for processes. This real-time data is a data stream because of its live nature and processing such real-time data is a relevant field of data-mining and machine learning known as Stream Processing and Online Machine Learning respectively.
After loading the stream provider (MLPro’s native stream provider for example), the list of available streams can be loaded as following:
# Import the stream provider class
from mlpro.bf.streams.native import NativeStreamProvider
# Create an object of the stream provider
mlpro = NativeStreamProvider()
# Get a list of streams
mlpro.get_stream_list()
Learn more
Cross Reference