Housing price prediction using sequence transformers

Abstract

The objective of this project is to create a forecast of Singapore’s housing prices using a dataset that includes prices of Housing and Development Board (HDB) flats over 5-10 years. The machine learning technique used in this research will be Sequence Transformers which is often used in Natural Language Processing (NLP). The paper applies the multi-layer attention layer, which improves processing time by parallelizing input data. The Transformer model allows for a bigger dataset to be used as compared to Recurrent Neural Network (RNN) tools such as Long-Short Term Memory (LSTM). Therefore, this project aims to test the feasibility of using Sequence Transformers by validating the output with loss functions by comparing training loss to validation loss.