The investigators have demonstrated that social software could improve warfarin therapy quality by Human management. however,the efficiency of management will decline if patients' number expands huge. So it is needed that a human-like and automatic management tool could replace the human work. The aim of this study is to evaluate whether an AI-based social software management model could replace human to guide warfarin therapy.

1. The investigators have developed an AI-based miniprogram, and embedded in the Wechat social application.
2. The investigators invited patients receiving warfarin therapy to participant this randomized controlled trial .
3. Participants were randomized to the experimental group and control group in a ratio of 1:1.
4. Participants' warfarin therapy were managed by an AI-based social miniprogram and a Human-based social miniprogram respectively.
5. International normalized ratio(INR) was monitored once a month at least, and the time in therapeutic range was calculated to evaluate the warfarin anticoagulation quality.
6. Participants were followed-up for 24 months, and clinical events, including major bleedings and thrombotic events were recorded during follow-up.

Inclusion Criteria:

• Patients with Atrial fibrillation
• Patients with Mechanical valve replacement
• Patients receiving warfarin therapy

Exclusion Criteria:

• Plan to stop warfarin therapy within 1 years
• Bleeding within 3 months
• Refuse to participate in this study
• Other conditions that the physician considers inappropriate for participation

Participants did not know whether they were assigned to AI-SMART group or Human-SMART group.