Climatologists have studied the change in air and ocean temperature and have come up a set of risk-factors as the primary causes of climate-change. Reputable data sources have put together datasets for climate-change risk factors. Many of these factors are attributed to population- specifically overall population size, urban population size, educational level, life expectancy, poverty rate, population density, land usage types, energy consumption of renewable and non-renewable sources, greenhouse gas emissions. In a previous publication [1] I looked at using the Long Short Term Memory model [2] to create a time series prediction model based on global warming world temperatures. Within this new work, I take into consideration 30 climate-change features and use a deep neural network which includes stacked and bidirectional LSTM layers among others as shown below. For each of the top 13 countries based on GDP [3], a multivariate, time-series based dataset, with a dimensionality of 30, was created for each of the 13 countries. Each country’s dataset was used with the LSTM based model and a future climate change prediction is made.
References
[1] Jennings, R. & Kaleemunnisa LNU, “Modeling Climate Change Through DNN and LSTM”, 51st Southeast Decision Sciences Institute Conference, February 2022.
[2] Hochreiter, S. & Schmidhuber, J. Neural Computation, Volume 9, Issue 8, pp 1735–1780, November 1997.
[3] World Population Review, https://worldpopulationreview.com/countries/countries-by-gdp