Tem­poral and Spa­tial Data Min­ing

Recommended prerequisites


At least one Bachelor or Master in Machine Learning should have been attended (e.g. Soft Computing), knowledge gaps can be closed in online courses on Machine Learning. Basic knowledge of stochastic, analysis and linear algebra is assumed. Additional, Python knowledge is beneficial.
 
Syllabus


The lecture addresses basic approaches of pattern recognition in time series (e.g., sensor signals) and spatially distributed data (e.g., in sensor networks).
It convers inter alia the following topics:

  • theoretical foundations (e.g., segmentation of time series, correlation of data)
  • time series representation (e.g., features extraction for describing temporal and spatial data)
  • distance and similarity measures for time series, clustering / classification, motifs, and anomaly/novelty detection using various techniques (e.g., nearest neighbor, neural networks, support vector regression),
  • diverse sample applications (signature verification, collaborative hazard warning for automotive, activity recognition, etc.)

Targeted Proficiency 

The students will be able to successfully:

  • explain various tasks, models, and algorithms of Temporal and Spatial Data Mining,
  • develop new modeling approaches for problems such as time series classification, anomaly detection, or clustering,
  • to independently plan and implement new applications of the learned paradigms
  • critically question, compare, and evaluate existing approaches and applications.