QUALITATIVE ANALYSIS OF OFFICIAL MILK CONTROL IN VALENCIA COMMUNITY (SPAIN) BY SELF ORGANIZING MAPS

Abstract

Breeding programs in dairy goats are mainly based on milk production and composition. Murciano-Granadina goats are located principally in the central and southern regions of Spain. This study is focused in Valencia Community (Spain) and the objective is to study the Murciano-Granadina livestock based on the database from Murciano-Granadina Goat Breeders Association of Valencia (AMURVAL).  The aim of this study is to analyze the relationship among different variables related with milk production; milk yield, fat, protein, lactose, SCC, the number of births, lactation number and season. This analysis is carried out by using the Self Organizing Map. This tool allows mapping high-dimensional input spaces into much lower-dimensional spaces, thus making much more straightforward to understand any representation of data. These representations enable to visually extract qualitative relationships among variables (Visual Data Mining). A total of 3221 Murciano-Granadina dairy goats from AMURVAL were chosen. Self Organizing Maps (SOM) were used to analyze data with the system identification toolbox of MATLAB v7. Data were obtained from Official Milk Control during 2006 campaign. SOM considered in this study is formed by 21´14 neurons (294 neurons); the chosen architecture is given by the range of the input variables used. The map shown that more than 70% of the goats has milk yield greater than 300 kg per lactation and goat, indicating good performance of farms. Besides, the SOM obtained indicate a group of neurons that included goats with high SSC (2%). The use of Self Organizing Maps in the descriptive analysis of this kind of data sets has proven to be highly valuable in extracting qualitative conclusions and guiding in improving the performance of farms.