Detecting frequent sequential patterns between weather and cattle behavior using data mining

Climate frequently influences the sustainability of livestock systems. As a result of climate change, heat stress may become a significant challenge for cattle producers. Heat stress occurs during hot weather conditions when animals are unable to maintain homeothermy, which can negatively affect production, reproduction, and animal well-being. In this study, thermal heat index was used to monitor thermal conditions facing cattle on rangelands. Three metrics—movement rate, activity, and distance traveled from water—obtained from GPS tracking were used to represent behavior changes in response to variation in thermal conditions. Each of these behavior metrics was categorized into four behavioral levels (high, medium, slight, and low) using a well-known k-means clustering algorithm. Additionally, daily thermal conditions were categorized into three weather levels (hot, medium, and cool) based on heat index values, also using the k-means clustering. The objective was to identify and detect the relationship between hot weather and cattle behavior, with the hypothesis that consecutive hot days have a clear negative effect on cattle behavior, particularly leading to a reduction in activity and movement. To investigate this, the unsupervised Co-occurrence Map Sequential Pattern Mining (CM-SPAM) algorithm in data mining was applied to analyse tracking data collected in the summers of 2019 and 2021 at Deep Well Ranch, Prescott, Arizona, USA. The CM-SPAM algorithm successfully identified that consecutive hot days (two, three and four days in a row) resulted in a consistent decrease in movement rate on the second, third and fourth days, respectively, suggesting a decrease in cattle activity during the morning and evening grazing bouts. The activity and distance to water metrics were not able to establish a connection between hot weather conditions and behavioral change. The CM-SPAM algorithm successfully identified impacts of consecutive days of hot weather on cattle rather than only daily evaluations. Our study demonstrates the potential to remotely detect changes in cattle behavior during potentially stressful thermal conditions. This type of analysis could enable early interventions to manage heat stress, preventing potential negative effects on the animals’ health and productivity.