Abstract: Nowadays estimation of disease and estrus conditions for large animals such as cattle mainly relies on farmer’s visual judgment. This practice by manual observation had an obvious fundamental restriction, which a large number of experienced farmers were required. Therefore, it was not cost-effective to be adopted in large scale and intensive breeding. This approach also inclined to have unreliable detection rate due to human errors. In order to recognize estrus or diseases of cattle automatically and accurately, this paper presented a monitoring system based on wireless sensor node installed on a cattle’s neck. Body temperature, respiratory rate, movement acceleration and other parameters of the cattle could be easily captured through a variety of sensors on the node. The collected data were sorted out and classified into different classes according to the behavioral features extracted by using the K-means clustering algorithm. This process enabled an empirical detection model to be developed which could then be applied routinely to predict estrus or diseases. Experimental results showed that the monitoring models could effectively distinguish behaviors such as standing, walking, mounting etc., and estrus or diseases could be observed accurately based on these unique features. This methodology in behavioral modeling could be adapted in monitoring other animals and holds great importance for development of other stockbreeding.