UCLA

Housing prices affect everyone. In this paper we establish a potential method of predictinglong term home sale prices in the United States using an LSTM neural network model. We took publicly available macro economic data, then massaged it into a manageable form using cubic spline interpolation and logarithmic differencing. We then introduce the LSTM model using feature standardization, min-max normalization, and an Adam optimizer for backpropogation. After training the network on preceding data, we found that the network was able to provide predictions that coincided with similar market movements.