Novel algorithm for diatom classification in Lake Prespa using log-normal distribution

One way to discover the indicating properties of the diatoms and the suitable conditions for their survival is to use ordination techniques. However, these techniques lack of easy interpretability. Also, according to some researches the latest technique used, detrended correspondance analysis (DCA), can hide the real data structure or can even introduce new distortions. In this paper, we propose a novel algorithm that aims to discover the diatom indicating properties directly from the measured data. Our algorithm induces pattern trees (PTs) that are resistant to over-fitting and are robust to data changes. In order to provide a more accurate prediction, we use log-normal distribution.