1: \begin{abstract}

2: We study the multiclass classification problem where the features come from a mixture of time-homogeneous diffusion.

3: Specifically, the classes are discriminated by their drift functions while the diffusion coefficient is common to all classes and unknown.

4: In this framework, we build a plug-in classifier which relies on nonparamateric estimators of the drift and diffusion functions.

5: We first establish the consistency of our classification procedure under mild assumptions and then provide rates of convergence under different set

6: of assumptions. Finally, a numerical study supports our theoretical findings.  

7: \end{abstract}

8: