Time Series Prediction on Population Dynamics

Abstract

Predicting the time series is a challenging topic mainly on the era of big data. In this research, data taken from population dynamics of one dimension of logistic map with various parameters that leading the system into chaos. Various machine learning methods is employed for predicting the time series data such as Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU) and 1 Dimension of Convolution Neural Network (1D CNN). Several data sizes were considered: 1000, 10000, 50000, 100000 and 1 million points of time series data. As evaluation metric, Root Means Square Error (RMSE) is used to assess the accuracy of each method. The result indicating that the LSTM has the smallest RMSE value among all the three machine learning methods.