Due to its versatility, ease of use, and performance, machine learning
is being applied to almost all fields of science, including electrical
engineering. A particular sub-field of electrical engineering which has
seen enormous growth due to the use of machine learning is battery
degradation - the analysis of how batteries decrease in effectiveness
over time. However, while machine learning tools appear appropriate,
as a battery exhibits seemingly non-linear degradation and a complex
dependence on external factors, these tools are being applied without
much thought, to data which is often not very representative of real
operation. In this talk, we will take a look at some of the pitfalls
in the data and machine learning tools applied in this field, as well
as some simpler or more coherent alternatives.