The aim of this study was to estimate, using data mining, which microclimate and behavioral variables affect the behavior of animals to seek shaded or sunny areas. The experiment was carried out between January and May 2016 in an integrated crop-livestock-forest system. In this system, we defined two different areas: shaded and sunny. Microclimatic variables (At, BGt, RH, and WS) were measured in each area on 4 consecutive days per month. With these variables, we determined the bioclimatic indicators (THI, BGHI, HLI, MRT, RTL, and ETI). In addition, we calculated the absolute difference (Δ) by subtracting the value recorded in shaded areas from the value recorded in sunny areas for all microclimatic variables and bioclimatic indicators, except for WS. The behaviors (grazing, ruminating, and other activities), posture (standing or lying), and use of areas (shaded or sunny) of 38 Zebu cattle were recorded on 2 consecutive days per month. The data mining technique was applied for analysis in a classification task. The model correctly classified 76% of the instances with a Kappa statistic of 0.51 after features selection from the database. The ΔBGt was the most important feature in the model to classify the decision of Zebu cattle to seek another area or remain in a determined area. The model was built with seven classification rules, being one simple rule, composed of the interaction between ΔBGt and rumination; and other more complex rules, composed of the interactions among the ΔBGt, WS, and rumination. The preference of Zebu cattle to seek or remain in shaded or sunny areas was influenced by eight features: rumination, drinking water, WS, ΔBGt, MRT in shade, BGHI in sun, ΔBGHI, and HLI in sun.

本研究的目的是使用数据挖掘来估计哪些小气候和行为变量会影响动物寻找阴凉或阳光充足区域的行为。该试验于 2016 年 1 月至 5 月在作物-畜牧-森林一体化系统中进行。在这个系统中，我们定义了两个不同的区域：阴影区域和阳光区域。每月连续 4 天测量每个区域的小气候变量（At、BGt、RH 和 WS）。通过这些变量，我们确定了生物气候指标（THI、BGHI、HLI、MRT、RTL 和 ETI）。此外，我们通过从阳光地区记录的所有小气候变量和生物气候指标（WS 除外）记录的值中减去阴影区域记录的值来计算绝对差 (Δ)。行为（放牧、反刍和其他活动）、姿势（站立或躺下）、每月连续 2 天记录 38 头瘤牛的区域（阴凉或晴天）的使用情况。将数据挖掘技术应用于分类任务中的分析。从数据库中选择特征后，该模型正确分类了 76% 的实例，Kappa 统计量为 0.51。ΔBGt 是模型中最重要的特征，用于对瘤牛寻找另一个区域或留在确定区域的决定进行分类。该模型建立了七个分类规则，一个简单的规则，由ΔBGt和反刍之间的相互作用组成；以及其他更复杂的规则，由 ΔBGt、WS 和反刍之间的相互作用组成。瘤牛寻找或留在阴凉或阳光地区的偏好受到八个特征的影响：反刍、饮用水、WS、ΔBGt、阴凉处的 MRT、阳光下的 BGHI、ΔBGHI、