I understand how a regression model can be used to calculate unclassified output. If I want a more accurate model do I switch to ANN or does ANN work only for classified outputs? Referring to the same example discussed in lectures of price of house based on various features. Can in improved version of the model be created in ANN to predict the price of house ?

Neural Networks can be used for regression.

The only special characteristic of an NN is that it can learn more complicated models than a simple linear regression, due to the presence of the non-linear functions in the hidden layers.

In that case the output layer should have only one unit that can predict any value?

For regression, you’d have one output that provides a real number (floating point) value.

So that means you don’t have any activation function on the output.