Three-Tiered Data Mining for Big Data Patterns of Wireless Sensor Networks in Medical and Healthcare Domains

As smartphones become an emerging interface platform between humans and systems, they also enable wireless sensors to interface with host servers. As sensors monitor application domains and sensor data is frequently polled and transmitted to a host server, the server data will be a big volume, big variety and big velocity, which is the characteristic of big data. Mining patterns from big data is a very important and active research topic since it can be used to forecast and "nowcast" for any dynamisms in application domains. However, typical data mining algorithms are not successful yet due to the characteristics of big data. This paper describes three-tied data mining paradigm. Alongside the streamline of sensor data transmission, at the microcontroller tier, sensor data sets are mined to form patterns, at the smartphone tier, negative and positive patterns are grouped and verified, and finally at the host server tier, human expertise is associated with the patterns. The contribution includes 1) lowering data transmission by mining from the lower tiers, 2) mining time-critical data earlier than it would be done at the host server tier and 3) hence urgent responses can be made timely at the proper tier.