Forecasting quarterly hog prices: Simple autoregressive models vs. naive predictions

Abstract

In this note, we study the forecasting performance of some simple models applied to the hog markets in the Nordic countries. In terms of accuracy (MSE and MAPE), a simple autoregressive model outperforms the naive expectations benchmark in some samples, as does a very simple VAR‐type model in which lagged piglet prices are added to the lagged hog prices as RHS variables. Forecasting performance is, however, quite sensitive to the chosen lag structure, and there is reason to doubt whether the simple autoregressive model from an economic point of view yields significantly better results than those of the naive model. Focusing on directional forecasts, on the other hand, the simple VAR‐models perform clearly better. Thus, for producers whose main concern it is whether the price moves up or down, these models may be quite useful. © 1997 John Wiley & Sons, Inc.

An extrapolative model for price prediction of crops using hybrid ensemble learning techniques

“…Thus, data mining, machine learning and statistics play a huge role in extracting knowledge and predicting the future price of the commodities precisely [10,11]. Several statistical models are available in the literature that can be used in forecasting time series data, among which the most popular and widely used techniques are autoregressive moving average (ARMA) [12], autoregressive integrated moving average (ARIMA) [13], simple exponential smoothing (SES) [14], naïve forecasting (NF) [15], Holt's method [16] and more. Many machine learning algorithms, such as linear regression (LR) [17], Gaussian process (GP) [18], neural networks [19], support vector machines (SVM) [20], multilayer perceptron (MP), and fuzzy logic [21], are also widely used in forecasting prices.…”