La inflación en Colombia : una aproximación desde las redes neuronales

Misas, A Martha; López-Enciso, Enrique Antonio; Querubín-Borrero, Pablo

doi:10.32468/espe.41-4203

Cited by 18 publications

(23 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…A Performance of Colombian inflation for 2-month to 9-month ahead forecasts Rodríguez and Siado (2003) Neural Networks N eural.N etwork Neural Networks model Misas et al (2002) Neural Networks N eural.N etwork.C Weighted average between an NN for food − by components inflation and an NN for non-food inflation LSTR LST R Logistic smooth transition regression model Jalil and Melo (1999) FLS F LS Flexible Least Squares approach Melo and Misas (2004)…”

Section: Resultsmentioning

confidence: 99%

Bayesian combination for inflation forecasts : the effects of a prior based on central banks' estimates

Melo‐Velandia¹,

Loaiza²,

Villamizar‐Villegas³

2014

View full text Add to dashboard Cite

Typically, central banks use a variety of individual models (or a combination of models) when forecasting inflation rates. Most of these require excessive amounts of data, time, and computational power; all of which are scarce when monetary authorities meet to decide over policy interventions. In this paper we use a rolling Bayesian combination technique that considers inflation estimates by the staff of the Central Bank of Colombia during 2002-2011 as prior information. Our results show that: 1) the accuracy of individual models is improved by using a Bayesian shrinkage methodology, and 2) priors consisting of staff's estimates outperform all other priors that comprise equal or zero-vector weights. Consequently, our model provides readily available forecasts that exceed all individual models in terms of forecasting accuracy at every evaluated horizon.

show abstract

Section: Resultsmentioning

confidence: 99%

Bayesian combination for inflation forecasts : the effects of a prior based on central banks' estimates

Melo‐Velandia¹,

Loaiza²,

Villamizar‐Villegas³

2014

View full text Add to dashboard Cite

show abstract

“…Wang (2007) 8 ,and Misas, López & Querubín (2002) 9 obtained favorable results for ANN that establish their superiority over linear models.However,Faraway & Chatfield (1998) criticized the adjustment of ANN to some of the univariate series used in the article "Time series Forecasting analysis and control" of Box, Jenkins & Reinsel, supporting the use of ARIMA models. SimilarlyCallen, Kwan, Yip, & Yuan (1996) reported that the linear models outperformed the ANN, when forecasting financial returns of the shares of the New York Stock Exchange.…”

mentioning

confidence: 81%

“…Likewise, a large number of hidden units increase the probability that the parameters converge to a local optimum. Therefore this document follows Misas, López & Querubín (2002) approach, where the number of nodes is selected by iterating between different values, not exceeding the double of the maximum number of outgoing nodes.…”

Section: Construction Of the Neural Networkmentioning

confidence: 99%

Forecasting latin-american yield curves: an artificial neural network approach

Vela¹

2013

View full text Add to dashboard Cite

This document explores the predictive power of the yield curves in Latin America (Colombia, Mexico, Peru and Chile) taking into account the factors set by the specifications of Nelson & Siegel and Svensson. Several forecasting methodologies are contrasted: an autoregressive model, a vector autoregressive model, artificial neural networks on each individual factor, and artificial neural networks on all factors that explain the yield curve. The out-of-sample performance of the fitting models improves with the neural networks in the one-month-ahead forecast along all studied yield curves. Moreover, the three factor model developed by Nelson & Siegel proves to be the best choice for out-of-sample forecasting. Finally, the success of the cross variable interaction strongly depends on the selected yield curve.

show abstract

“…variables exógenas y G la función logística, es decir: Dado que esta red presenta una única superficie oculta y que la información va en una única dirección entrada salida (input-output) se conoce como red neuronal "feedforward" multicapa de una única superficie oculta o "single hidden layer feedforward network", (Misas et al, 2002).…”

Section: Arquitecturaunclassified