Cancer Incidence Prediction Using a Hybrid Model of Wavelet Transform and LSTM Networks

Abstract

Cancer is a major public health concern. Being able to predict the number of future cancer incidences is vital to allocate appropriate healthcare resources and research funding. Real data, such as cancer incidences, exhibit non-linear characteristics along with a high degree of fluctuations, which makes the modelling process difficult. This study explores the potential of time series modelling, especially the long short-term memory (LSTM) recurrent neural networks, to predict the number of cancer incidences. A novel hybrid model of the wavelet transform and LSTM is proposed with the goal of increasing forecasting accuracy. The evaluation of the proposed models for the three most common types of cancer in the Kingdom of Saudi Arabia shows that the proposed hybrid model has better accuracy than the original LSTM model.