Edible oil wholesale price forecasts via the neural network

For a wide spectrum of agricultural market participants, building price forecasts of various agricultural commodities has always been a vital project. In this work, we approach this problem for the weekly wholesale price index of edible oil in the Chinese market during a ten-year period of January 1, 2010-January 3, 2020 through the exploration of the non-linear auto-regressive neural network as the forecast model. Specifically, we investigate forecast performance stemming from different settings of models, which include considerations of training algorithms, hidden neurons, delays, and how the data are segmented. With the analysis, a relatively simple model is constructed and it produces performance that is rather accurate and stable. Particularly, performance in terms of relative root mean square errors is 2.80%, 3.01%, and 1.80% for training, validation, and testing, respectively. Forecast results here could be utilized as part of technical analysis and/or combined with other fundamental forecasts as part of policy analysis.