Project: Evaluate Amazon Web Services as a platform for Machine Learning
Business Purpose: Evaluate the use of standard platforms to create machine learning/prediction algorithms which can then be used at the edge of IoT networks to reduce bandwidth cost, battery life and latency inIoT systems.
First Stage of Project: Evaluate AWS IoT Core as an MQTT broker and device authentication to receive and assemble data from remote sensors. Use AWS IoT Analytics to handle processing, storing and analysing data.
Project Detail: Electricity Substation monitoring data from Celsius Project used as base data. Goal is to predict temperature cycle given time/day of the week, so that just variation from predicted alert can be reported.
Project Outcome: Discovery Protocol buffer not natively supported. Python scripts were necessary to reassemble received data packets into the correct order, remove duplicates etc. MQTT process worked well. Prediction model created and tested.
Follow up: Additional work steps on same project defined for Xmas vacation internship.
Future project: Edge Implementation of algorithms