Dynamic Modeling of Italian Housing Market Prices with Recurrent Neural Networks

Abstract

The aim of this dissertation is to study the evolution of the prices in the Italian housing market. The main contribution is a dynamic approach based on the use of recurrent neural networks to fit the time series of house prices in various Italian cities and to forecast future prices. In particular, we train a collection of recurrent neural network models - including LSTM networks, convLSTM networks, and CNN-LSTM networks - and compare the respective performances in modeling and forecasting house prices.