Aplicação da técnica de regressão para análise de dados climáticos e previsão de safra

Abstract

This work’s objective is to present a crop forecast model based on the soy production history and the weather data for the state of Paraná. The research was made with the soy crop data of the years from 2003 to 2012 from 8 cities of the state, these data were collected together with the Secretary of Agriculture and Supply of Paraná(SEAB). The weather data were obtained through the National Institute of Meteorology. The software responsible for crossing data and making the crop forecast is Weka. Willing to generate crop forecast models some algorithms were used from Weka, Linear Refression, Pace Regression e LeastMedSq Regression are algorithms that work for these models. The weather data to be mined were grouped by monthly, bimonthly and quarterly means aiming to check which one of them would have the best result. After the forecast models were generated, they have been validated through samples, willing to find a model having the result closest to the predicted one and the real value.