House Prices ANN#

In this project we solve the house price prediction problem from Worked Example: House Prices II with an ANN. Read ANN Basics and Training ANNs before your start.

The ANN#

In Worked Example: House Prices II we saw that not all features or of importance for price prediction. Thus, here we restrict our attention to the important features only.

Task: Train an ANN for house price prediction based on following features:

  • region’s average land prices and income,

  • lot size and living space,

  • build and renovation year

  • number of rooms and bath rooms,

  • building type (duplex, villa, other).

Apply regularization (parameter choice via Scikit-Learn’s hyperparameter optimization methods) and try different ANN sizes (number of layers, number of neurons per layer) as well as different activation functions.

Solution:

# your solution

Some Theory#

Task: Think about implications of very small batch sizes in case of noisy data. Are small batches good or bad here?

Solution:

# your notes