Estimating CO2 Emission Time Series with Support Vector Machines Regression, Artificial Neural Networks, and Classic Time Series Analysis

Çemrek, Fatih; Gunes, Ozge

doi:10.34110/forecasting.1035912

Cited by 1 publication

(2 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[17] compared actual and predicted GHGs emissions by artificial neural networks of Bulgaria and Serbia. [19] applied Support Vector Regression, Artificial Neural Networks, and Box-Jenkins method to model CO2 emissions. [21] studied on CH4 emissions for Tibetan Plateau between years 2006 and 2100.…”

Section: Literature Overviewmentioning

confidence: 99%

“…Time series data of GHGs emissions of countries are able to be obtained in hourly, daily, monthly and annual terms. Forecasting techniques on past univariate or multivariate data generally include methods of artificial neural networks [17], fuzzy logic [18] support vector machine [19], machine learning [20] and classic statistical models as regression [21], autoregressive integrated moving average [22] and grey [23], etc. Statistical methods are more appropriate for future predictions of short univariate time series in regard to other methods preferred more for long univariate and multivariate data.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Estimations of Green House Gases Emissions of Turkey by Statistical Methods

ÖZTÜRK,

EMİR

2024

KONJES

View full text Add to dashboard Cite

The way of life, consumption habits, urbanization rate, type of energy production and increasing energy need with growing economies and population progressively promote the GHGs emissions to Earth’s atmosphere. GHGs consisting of CH4, N2O, CO2, H2O and HFCs cause the climate change, disrupting ecological balance, melting glaciers with global warming in the last decades. Therefore, the issues of future prediction and reduction of GHGs emissions became crucial for policy makers of Turkey and other countries under the international protocols and agreements. This article aims to present the prediction and 8-year future forecasting of CH4, N2O and CO2 emissions of Turkey using past annual data between years 1970 and 2018 with grey, autoregressive integrated moving average and double exponential smoothing models. Based on the results, the best prediction performance is reached by DES model followed by ARIMA and GM for all the emissions. MAPEs calculated from the available data and prediction by DES model from 1970 to 2018 are 0.285, 0.355 and 0.408 for CH4, N2O and CO2 in turn. DES future estimations of CH4, N2O and CO2 at 2026 year are determined as 50700 kiloton of CO2 eq., 38100 thousand metric ton of CO2 eq., and 512000 kilotons.

show abstract

Section: Literature Overviewmentioning

confidence: 99%