Hi Amit, It depends on the use case, Euclidean distance is prefered if you have a continuous and normally distributed data and Manhattan if your input variables are not similar in type i.e categorical features.

→ View original post on X — @sumanth_077