Uso de Algoritmo de Inteligencia Artificial para Desarrollar una Metodología para Medir la Textura de los Suelos

This research is intended a methodology to determine soil texture using the spectrum of diffuse reflectance through the use of the artificial intelligence algorithm. RadiometerFieldSPec 4 was used with the purpose of registering the diffuse reflectance spectrum, the statistical modeler and the McNemar test and then selecting the algorithm with the best performance. It generated 5 algorithms (decision trees) that have a less number of nodes, from these were formed 10 pairs of trees to choose the tree with the lowest percentage of error. The tree A1 presented a better performance with a general precision of 87.50 %, the tree A1 was obtained from segmenting the spectrum of diffuse reflectance in a width of 10 nm, with this, the variables of importance were identified. The observables O 25, O 27, O 24, O 26 and O 29 are in a range of 590 - 640 nm by the presence silt while O 65, O 145 and O 149 are between 990 - 1840 nm for the content of sand and O 177 and O 180 are between 2110 - 2150 nm by the presence clay. Through the confusion matrix for the test stage of the tree A1 was achieved a precision of 89.00 %, a sensitivity of 92.00 %, specificity of 71.00 % and an incorrect classification rate of 11.00 %. The methodology used to measure the texture of the soil with the use of the artificial intelligence algorithm the error is 13.60 % while with the traditional technique of 33.33 %, achieving thus obtain a less uncertainty and a percentage of error.

• Hovedforfatter: Erazo López, Estéfany Alejandra (author)
• Format: bachelorThesis
• Sprog: spa
• Udgivet: 2019
• Fag: ALGORITMO; INTELIGENCIA ARTIFICIAL; REFLECTANCIA DIFUSA
• Online adgang: http://dspace.unach.edu.ec/handle/51000/5710