Stratifying risk of disease in haematuria patients using machine learning techniques to improve diagnostics