Incremental time series algorithms for IoT analytics: an example from autoregression

The Internet of Things (IoT) is emerging as an important application area for time series statistical analysis and data mining of time series. As the volume of sensor data is high, time series analysis of sensor data is a problem of processing large datasets. Moreover, the IoT platforms have to simultaneously process multiple jobs on the same infrastructure. Processing such large datasets requires large amount of memory. To alleviate this problem, we propose use of incremental algorithms. Incremental algorithms can be used for both batch and streaming applications. In this paper, we show an incremental algorithm for an example time series analysis algorithm viz. autoregression. We describe a memory efficient autoregression algorithm and show the memory footprint reduction achieved by using this incremental algorithm.