Corn is one of the leading crops in the world as it has about 3500 uses and is the only crop that surpassed the billion-ton mark produced per year. In Brazil, this crop is fundamental for national agriculture being cultivated in all regions of the country in more than two million agricultural establishments [1]. In recent years, this crop has undergone several technological and economic transformations that have made this product very important for Brazilian exports. However, as with all crops, the cost of corn production is dependent on inputs, such as cultivated area and diesel for agricultural machinery, and is affected by climatic factors, such as precipitation and temperature. The fluctuation in the price of inputs and climatic variations can, therefore, impact production, supply, and, consequently, the price of corn, with the unpredictability of the price being one of the main obstacles to the development of this commodity in Brazil [1].
The formation of prices for a commodity is a process by which markets try to balance prices depending on the information each merchant holds at any given time.
The Center for Advanced Studies in Applied Economics (CEPEA) is part of the Depart- ment of Economics, Administration and Sociology at the Luiz de Queiroz College of Agriculture (ESALQ), from the University of São Paulo (USP). Through an API, daily data on the price of the corn buschel is extracted for the period between August 2004 and June 2021. The reference region for purchase prices is Campinas, in the state of S ̃ao Paulo. This first dataset has 4195 observations and three initial attributes, namely: date, the buschel spot price in RS$ and buschel spot price in US$. This first database was complemented with more attributes from other databases, described below.
Historical climatological data from 15 meteorological stations in the southern region of the country are extracted from the National Institute of Meteorology (INMET) portal, also through an API. The raw dataset has 74885 observations of the measurement date, mean temperature (in degrees Celsius), rain (corresponding to precipitation, in mm) and location attributes, ie, the me- teorological station of origin. As our study is focused on the city of Assis, we retrieved data for meteorological stations in nearby cities and interpolated the weather data by the inverse of the distance.
quandl is a Python library that provides structured financial data from the Nasdaq’s Quandl platform. Through this package, the monthly historical series of diesel prices, in dollars, from 1946 to 2021 is obtained. The dataset has 905 observations. From quandl, the daily historical series of the exchange rate of the real against the dollar is also obtained. There are a total of 9152 observations, the oldest being from November 1984 and the most recent, from June 2021.
[1]E. Contini et al. ”Milho - Caracterização e Desafios Tecnológicos”. Série desafios do agronegócio brasileiro. Embrapa. 2019. Avaliable at: https://ainfo.cnptia.embrapa.br/digital/bitstream/item/195075/1/Milho-caracterizacao.pdf. page.11
[2] S. N. Ludovico, ”Previsão de indicadores diários de preços no mercado futuro de com- modities agrícolas usando aprendizagem de máquina”, M.S. thesis, ICE, UNIFAL, Alfe- nas, Brasil, 2020. [Online]. Avaliable: http://bdtd.unifal-mg.edu.br:8080/handle/tede/1762 page.11
[3] R. C. F. Amorim. ”Desempenho do Metodo do inverso da distância ponderada na interpolação de dados horarios de temperatura do Ar”, 2011. Avaliable at: http://www.sbagro.org/files/biblioteca/3442.pdf page.22
[4] V. B. Santos, ”Estimação e previsão de produtividade de Soja por redes neurais no Matopiba”. Jabuticabal, 2020. Unesp. page.22
[5] W. F. Nascimento, ”Efeitos da Temperatura Sobre a Soja e Milho no Estado do Mato Grosso do Sul”. Dourados, 2016. Universidade Federal Grande Dourados. 2016. page.22
Senior data engineer at Serasa Experian
Senior Data Scientist at Cargill
