CAP-359-3
This is a repository for the "Data Mining" discipline at INPE/CAP.
Project idea
This project is being made by:
- Caio Gustavo Rodrigues da Cruz
- Yuri Matheus Dias Pereira
The idea is to use data mining techniques to extract information from satellite telemetry data, to aid with mission control operations.
More specifically, in this project we will apply motif discovery and association rules mining algorithms and techniques to satellite telemetry data, treating them as time series. The aim with the motif discovery is to see patterns on the satellite subsystems that might signal a possible equipment degradation or anomalous behaviour, while the goal of the association rules mining is to see how the different subsystems telemetry readings impact on each others.
Data
The available data consists of several months and years of one of INPE's operational satellites. It consists of 135 telemetries, read at every half a second during each satellite passage over one of the available ground stations.
Algorithms, packages and implementation
For the motif discovery part, the jmotif-R package will be used, as it already implements the relevant SAX algorithms and support algorithms. Additionally, the DTW algorithm will be used to evaluate the found discords in relation to the rest of the telemetries as a case study.
The association rule mining algorithm that will be used is the standard implementation of the Apriori algorithm available in the WEKA software tool.
Results and timeline
A report will be made with the results found and applied algorithms in the end of the discipline period, by the 10th of december, on this repository.