Thesis PROPUESTA PARA LA ESTIMACIÓN DE LA PRODUCCIÓN DE UVA MEDIANTE MÉTODOS DE OPTIMIZACIÓN ENTERA.
Loading...
Date
2010-01
Journal Title
Journal ISSN
Volume Title
Program
DEPARTAMENTO DE INDUSTRIAS. INGENIERÍA CIVIL INDUSTRIAL
Campus
Campus Vitacura, Santiago
Abstract
En el siguiente documento se presenta la evaluación del método CRIO (Clasificación y Regresión Vía Optimización Entera) en la generación y validación de una regresión entre el índice PCD (Índice de densidad de Células Vegetales) y el peso promedio del racimo de uva por cuartel en el día cero antes de cosecha estimado por muestreo. Esta evaluación se realizó para las siguientes cepas: Cabernet Sauvignon, Carménere, Chardonnay, Merlot y Sauvignon Blanc. Así, con el fin de predecir la producción del peso del racimo de uva se utilizó un estudio de Imágenes Multiespectrales y Muestreos Dirigidos realizado por Ortega, el año 2008. En el estudio se muestrearon 63 cuarteles en 12 campos distintos, la superficie total de producción de los 63 cuarteles fue de 250,55 hectáreas, los kilos de uva muestreados fueron 9.937,15 [Kg] y la superficie muestreada correspondió al 15% de la superficie total. El método CRIO, el cual se utilizó para generar la relación entre el índice PCD y el peso del racimo de uva en el día cero antes de cosecha, presentó como principal virtud la capacidad de eliminar los “Outliers” antes de calcular la regresión. Los resultados obtenidos se dieron de muy buena forma en las cepas: Cabernet Sauvignon, Merlot y Sauvignon Blanc, sin embargo en las cepas: Carménere y Chardonnay no se lograron obtener patrones que derivasen en una relación, una explicación de esto puede deberse a que en algunos cuarteles se realizó una mala distribución de las muestras y pocas sesiones de muestreo. Luego la validación de los modelos seleccionados para las cepas: Cabernet Sauvignon, Merlot y Sauvignon Blanc, fue bastante buena cumpliéndose casi todos los supuestos estadísticos, el único problema se presentó en la normalidad de los errores, lo cual fue un problema común en los tres modelos. Esto puede deberse a una mala distribución de las muestras o a la falta de información. Por último al comparar el rendimiento en kilogramos de uva por hectárea en cada cuartel estimados por el modelo del índice PCD con el rendimiento en kilogramos de uva real por hectárea en cada cuartel, los resultados fueron muy buenos. Se observó que el 93,1% de lo estimado en promedio es lo que realmente se produce, este resultado permite concluir que el índice PCD es un instrumento de predicción confiable. Además se cumplieron casi todos los supuestos estadísticos, sin embargo debido a falta de información, los intervalos de confianza de la regresión se vieron afectados.
The following document presents CRIO assessment method (Classification and Regression Whole Route Optimization) in the generation and validation of a regression between the PCD index (Index of Plant Cell density) and the average weight of bunch of grapes per quarter in zero day before harvest estimated by sampling. This evaluation was undertaken for the following strains: Cabernet Sauvignon, Carmenere, Chardonnay, Merlot and Sauvignon Blanc. Thus, in order to predict the production of grape bunch weight was used multispectral imaging study and purposive sampling performed by Ortega, 2008. The study sampled 63 quarters in 12 different fields, the total production area of 63 quarters was 250.55 hectares, the kilos of grapes were sampled 9937.15 [kg] and the area sampled corresponded to 15% of the total area. CRIO method, which was used to generate the relationship between the PCD index and weight of bunch of grapes on day zero before harvest, introduced as the main virtue ability to remove outliers before calculating the regression. The results were very good shape in the vines: Cabernet Sauvignon, Merlot and Sauvignon Blanc, however in the strains: Carmenere and Chardonnay are not able to get patterns that result in a relationship, an explanation for this may be because in some quarters there was a poor distribution of samples and sampling a few sessions. After validation of the models selected for strains: Cabernet Sauvignon, Merlot and Sauvignon Blanc, it was pretty nice compliment almost every statistical assumptions, the only problem was presented in the normality of the errors, which was a common problem in all three models. This may be due to poor distribution of the samples or lack of information. Finally to compare the performance of kilograms of grapes per hectare in each quarter estimates for the model of PCD index with performance in real grape kilograms per hectare in each quarter, the results were very good. It was observed that 93.1% of the estimated average is what actually occurs, this result suggests that the PCD index is a reliable prediction tool. In addition, serving almost all statistical assumptions, however due to lack of information, the confidence intervals of the regression were affected.
The following document presents CRIO assessment method (Classification and Regression Whole Route Optimization) in the generation and validation of a regression between the PCD index (Index of Plant Cell density) and the average weight of bunch of grapes per quarter in zero day before harvest estimated by sampling. This evaluation was undertaken for the following strains: Cabernet Sauvignon, Carmenere, Chardonnay, Merlot and Sauvignon Blanc. Thus, in order to predict the production of grape bunch weight was used multispectral imaging study and purposive sampling performed by Ortega, 2008. The study sampled 63 quarters in 12 different fields, the total production area of 63 quarters was 250.55 hectares, the kilos of grapes were sampled 9937.15 [kg] and the area sampled corresponded to 15% of the total area. CRIO method, which was used to generate the relationship between the PCD index and weight of bunch of grapes on day zero before harvest, introduced as the main virtue ability to remove outliers before calculating the regression. The results were very good shape in the vines: Cabernet Sauvignon, Merlot and Sauvignon Blanc, however in the strains: Carmenere and Chardonnay are not able to get patterns that result in a relationship, an explanation for this may be because in some quarters there was a poor distribution of samples and sampling a few sessions. After validation of the models selected for strains: Cabernet Sauvignon, Merlot and Sauvignon Blanc, it was pretty nice compliment almost every statistical assumptions, the only problem was presented in the normality of the errors, which was a common problem in all three models. This may be due to poor distribution of the samples or lack of information. Finally to compare the performance of kilograms of grapes per hectare in each quarter estimates for the model of PCD index with performance in real grape kilograms per hectare in each quarter, the results were very good. It was observed that 93.1% of the estimated average is what actually occurs, this result suggests that the PCD index is a reliable prediction tool. In addition, serving almost all statistical assumptions, however due to lack of information, the confidence intervals of the regression were affected.
Description
Keywords
ADMINISTRACIÓN DE LA PRODUCCIÓN, UVAS--CHILE, MUESTREO, EMPRESAS, ESTUDIOS DE FACTIBILIDAD