Técnicas estadísticas multivariantes para identificar y clasificar los factores influyentes en la producción de arroz, Ecuador 2017

Abstract

This research aimed to compare multivariate statistical techniques to identify and classify the influential factors in rice production in Ecuador in 2017. 18 agronomic variables, 15 qualitative and 3 quantitative measures were used in rice fields that participated in the Survey of Surface and Continuous Agricultural Production (ESPAC) 2017, the research was not experimental and had a relational exploratory scope, an Exploratory Data Analysis (AED), Regression Trees (AR), Mixed Data Factorial Analysis (AFDM) and the R software to develop the study. The AED determined that the average rice production is 212.21 pounds per hectare, 86% of the crops produced between 200 and 250 pounds, it was determined that the variables planted area and area harvested have no correlation with the rice production variable. After the application of multivariate methods, the AR model with an RMSE-7.32 value detected as influencing factors in the production of rice on sown surface, harvested area, use of irrigation, use of organic fertilizer and seed class. On the other hand, the AFDM with a model that explained 81% of the variability of the data generated two groups of factors influencing rice production, the first group is composed of the variables fertilizer use, chemical fertilizer use, use of chemical pesticides and the use of phytosanitary products, while the second group is composed of planted area and harvested area. For this reason, it was possible to determine that the influential factors that the two techniques have in common are planted area and harvested area, the most important factors for RA are use of irrigation and harvested area, while for AFDM they are use of fertilizers and surface sown. It is necessary to expand research with the use of other data mining techniques.