A comparison of the optimized LSTM, XGBOOST and ARIMA in Time Series forecasting
Authors:
- 12 July 2021
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-7
Abstract
The term time series refers to historical data comprise of observations that are made in a fixed time step, successively, over a period of time. This work focuses on the training and application of modern Machine Learning approaches, like Deep Neural Network techniques, to model and predict general time series obtained from several open databases. The selection of the data that were used in the experiments was focused on specific economic and social phenomena, intending to predict their evolution over time. The main and final goal remains the comparison of the aforementioned predicting approaches, as well as the optimization of them in order to improve their accuracy.Keywords
This publication has 36 references indexed in Scilit:
- Recurrent neural network and a hybrid model for prediction of stock returnsExpert Systems with Applications, 2015
- A variance inflation factor and backward elimination based robust regression model for forecasting monthly electricity demand using climatic variablesApplied Energy, 2015
- Neural Networks for Time-Series Forecasting2012
- Day-ahead wind speed forecasting using f-ARIMA modelsRenewable Energy, 2009
- A hybrid ARIMA and support vector machines model in stock price forecastingOmega, 2005
- Ensemble Methods in Machine Learning2000
- Long Short-Term MemoryNeural Computation, 1997
- Fractionally integrated generalized autoregressive conditional heteroskedasticityJournal of Econometrics, 1996
- Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom InflationEconometrica, 1982
- Time Series Analysis: Forecasting and ControlJournal of Marketing Research, 1977