With the rapid development of 5G/5G-A network, network architecture is becoming more complex, and the types of network object are also increasing. In daily network operation, a large number of alarms are generated and reported by network objects, and there are complex and diverse associations between different network objects. Network alarm is a serious threat to the normal operation of the network, but massive alarm and complex object relationship increase the difficulty of network alarm analysis. To ensure the normal network operation, alarm association analysis can effectively locate the alarm root cause, which is conducive to intelligent network operation and maintenance. An alarm association analysis algorithm based on network object relationship is proposed, which constructs a multi-dimensional alarm transaction mathematical model according to network object and time characteristics. This paper uses the frequent item mining of object-level alarm transactions to filter the time-ordered alarm transactions, and defines an alarm confidence calculation method based on the relationship between network objects. The algorithm can effectively analyze alarms and improve the accuracy of association analysis by considering the influence factors of network object association, which is conducive to intelligent alarm root cause location.