The invention discloses a drowning detection alarm instrument based on a cloud server and a machine learning algorithm, wherein the drowning detection alarm instrument relates to the field of computerapplication technology. The drowning detection alarm instrument comprises a wearable device which is worn on a hand or a foot of a user, a server and a client which is held by a guardian. The wearable device acquires action gesture data and heart rate data of a user. The server analyzes the data through integration of multiple factors, thereby determining whether drowning of a swimmer occurs, andtruly realizing a drowning event monitoring function. Through data connection between a server end and a client, alarming information can be accurately transmitted to the guardian or a swimming trainer in real time. The drowning detection alarm instrument has advantages of high accuracy, high efficiency, real-time performance and large monitoring range.