The Analysis of Machine Learning Algorithm Performance on Long Term Data using Short Time Scale Analysis

This thesis analyses the performance of various classification algorithms for epileptic seizure prediction algorithm development to find which algorithm performs better than others on long term electroencephalogram dataset. It shows that statistically there is no statistically significant difference between the performance of various classification algorithms for seizure prediction model development. It also demonstrates that there is variability between the performance of a given classification algorithm for various patients. Hence, seizure prediction algorithm is patients specific.