Communication networks usually contain thousands of components producing millions of alarms every day. To ensure sustainable high quality network services, network surveillance experts need to inspect these alarms and determine the underlying faults. High number of network alarms cause alarm pollution which may cause not only extra time spent by the network surveillance experts but also delays in handling of important faults. One way of overcoming this alarm pollution problem is to filter and reduce the number of alarms before the network faults can be located. Alarm correlation techniques are used to automatically detect and group related alarms which point to the same root cause faults, and therefore they reduce the number of alarms. In this paper, we present new statistical approaches which automatically produce alarm correlation rules by investigating network alarm history. We apply the, Apriori algorithm which is one of the well-known Market Basket Analysis techniques together with the Sliding Time Window technique on alarm history in order to automatically determine correlated alarm type patterns.