We have used boston housing datset for our problem.
check for null values check for duplicated entries check for correlation among features--if two independent features have strong positive corelation then we may drop one of them for our problem if we have corelation close to zero between independent and dependent feature then we may drop that dependent feature
internally our algorithm uses gradient descent which will aim to come near to global minima so we convert all data in same scale