Time series data are everywhere: from economic forecasting to electrocardiograms and infrared sensors measuring nutrient content in food. There are many challenges here for machine learning practitioners: there is a continuum of features, which immediately puts us in a “curse of dimensionality” scenario. For some applications, there is either way too little data (100-150 points in chemometrics) which makes classification and regression hard. For some other problems, there is too much data (sensor measurements) which makes indexing and retrieval hard. Out of the box, machine learning requires significant tweaks to be effective. In this course, we introduce and motivate some of those tweaks.
Familiarity with machine learning in Python
Python, scikit-learn, pandas, matplotlib, scipy/numpy
About the lecturer
Dr. Pablo Maldonado
Data Scientist and Applied Mathematician (Ph.D. Applied Math, Universite Paris VI, France). Previously adjunct professor at Czech Technical University and Senior Consultant at PricewaterhouseCoopers Czech Republic, I now do a mix of training and consulting for different organizations.