Health-related Analysis and Intelligent Early Warning Model Construction Based on Big Data of Pet Behavior

With the increasing number of pet-owning households, China's canine and feline populations have exceeded 120 million as of 2024. However, pet health management faces significant challenges, as most diseases are diagnosed at advanced stages. This study focuses on the dynamic correlations between pet behavioral characteristics, emotional states, and health risks, aiming to develop an intelligent health early-warning model tailored for household settings. Based on Alibaba Cloud's open-source "AnimalBehavior" big data dataset, the research integrates 1,200 canine and feline behavioral videos, multi-dimensional emotional scores, and 500 clinical diagnostic labels, employing an interdisciplinary approach to achieve in-depth behavioral data analysis. The study follows a technical pathway of "data collection-preprocessing-feature engineering-modeling validation": First, behavioral frequency features are extracted using OpenCV video decoding and AlphaPose keypoint detection algorithms, followed by feature matrix construction through principal component analysis and one-hot encoding, with multi-model training utilizing algorithms such as random forest and logistic regression. Further insights reveal the predictive value of key behavioral patterns: The correlation strength between canine paw-licking frequency ≥ 7 times/day and pain score>0.6 reachesr=0.78, while feline hiding frequency ≥ 5 times/week increases anxiety risk by 3-fold. Through dynamic threshold setting across three dimensions of "breed-behavior-emotion," the model achieves precise adaptation to individual pet characteristics.