Forecasting stock index returns using ARIMA-SVM, ARIMA-ANN, and ARIMA-random forest hybrid models
Citations
251 citations
110 citations
101 citations
Cites background from "Forecasting stock index returns usi..."
...Representative applications can be found in many scientific fields including these of engineering [23,24], environmental and geophysical sciences [25–27], financial studies [28,29], and medicine [30], with...
[...]
48 citations
27 citations
References
[...]
79,257 citations
"Forecasting stock index returns usi..." refers methods in this paper
...RF model was developed by Breiman (2001), exhibits outstanding performance with regard to prediction error on a suite of benchmark datasets....
[...]
...The RFs as proposed by Breiman (2001) are easy to use; the only two parameters that have to be determined are the number of trees to be used and the number of variables (m) to be randomly selected from the available set of variables....
[...]
...The RF algorithm (for both classification and regression) as proposed by Breiman (2001) was as follows: a random vector θk was generated, independent of the past random vectors θ1, ..., θk–1 but with the same distribution; and a tree was grown using the training set and θk, resulting in a…...
[...]
3,680 citations
3,155 citations
"Forecasting stock index returns usi..." refers background or methods or result in this paper
...There have been studies which combine fuzzy logic with ANN (Yu and Huarng, 2008), ARIMA and SVM (Pai and Lin, 2005), ARIMA and ANN (Zhang, 2003), Kohonen self-organising map with ARIMA (Voort et al., 1996), ARIMA and ANN (Su et al., 1997), radial basis function and ARIMA (Wedding and Cios, 1996), regression and ANN (Luxhoj et al., 1996), and combining several feed forward neural network (Pelikan et al., 1992; Ginzburg and Horn, 1994) to improve the time series forecasting accuracy....
[...]
...There have been studies which combine fuzzy logic with ANN (Yu and Huarng, 2008), ARIMA and SVM (Pai and Lin, 2005), ARIMA and ANN (Zhang, 2003), Kohonen self-organising map with ARIMA (Voort et al., 1996), ARIMA and ANN (Su et al., 1997), radial basis function and ARIMA (Wedding and Cios, 1996),…...
[...]
...A hybrid model comprising a linear and a non-linear component as represented below was employed in the experiments (Zhang, 2003). t t tY L N= + (13) where Lt denotes the linear component and Nt denotes the non-linear component....
[...]
...Zhang (2003) found that hybrid ARIMA-ANN model outperforms the individual ANN and ARIMA model while Pai and Lin (2005) showed that hybrid ARIMA-SVM model outperforms the independent ARIMA and SVM model....
[...]
...The performance of ARIMA-ANN hybrid model was compared with independent ARIMA and ANN models....
[...]
2,269 citations
"Forecasting stock index returns usi..." refers background or methods in this paper
...The idea behind the model combination is to use the unique features of each model to accurately analyse different patterns in the data and improve the forecasting accuracy (Makridakis, 1989; Clemen, 1989; Palm and Zellner, 1992; Makridakis et al., 1993)....
[...]
...Clemen (1989) and Granger and Ramanathan (1984) concluded that failure of the combination models may be attributed to the performance of benchmark models....
[...]
1,535 citations
"Forecasting stock index returns usi..." refers background or methods or result in this paper
...Literature shows that SVM outperforms models like ANN in forecasting financial time series (Trafalis and Ince, 2000; Kim, 2003; Huang et al., 2005, 2005; Chen and Wang, 2007; Ding et al., 2008; Kumar and Thenmozhi, 2009)....
[...]
...Kim (2003) found in a preliminary test that the polynomial kernel function takes longer time to train the SVM model and provides worse results than the Gaussian radial basis function....
[...]
...Literature also shows that there is evidence of SVM outperforming ANN and other benchmark models for the USA, Korea, Japan, Asian, Indian and Turkey stock markets (Trafalis and Ince, 2000; Kim, 2003; Huang et al., 2005; Chen et al., 2006; Kumar and Thenmozhi, 2009; Kara et al., 2011)....
[...]
...SVM and RF have been used to solve estimation problems and they exhibit excellent performance, because, they avoid the question of modelling the underlying distribution and focus on making accurate prediction of certain variables, given other variables (Kim, 2003; Kumar and Thenmozhi, 2009)....
[...]
...Moreover, proper selection of parameters such as input variables, number of epochs and learning rate, is difficult in ANN (Kim, 2003)....
[...]