Cellular networks usually suffer from failures or performance degradations duernto several reasons, such as external interference, hardware/software bugs on networkrnelements, power outages, or cable disconnections. Therefore, to avoid customerrndissatisfaction and loss of revenue due to failures telecom operators needrnto detect and respond to performance anomalies of cellular networks instantly.rnHowever, the state of art performance anomaly detection framework (CELLPAD),rnwhich uses a correlation between two Key Performance Indicators (KPI)s as arnmeans to detect anomaly is not capable of detecting anomalies happing during thernoff-pick hour, and; could not differentiate the causes of the anomalies. In this thesis,rnwe propose a system model, which is capable of detecting anomalies happingrnat any trafï¬c load and could differentiate the two causes of a correlational changernanomaly. The proposed system model uses a newly added parameter called meanrnReceived Total Wideband Power (RTWP) and ï¬ltering rules. To assess the performancernof the proposed system model, we conducted an experiment using fourmonthrnperformance counter data collected from 20rnselected sites. The result showsrnthat the proposed approach improves the detection of sudden drop anomaly byrn10rn% when compared to the state of the art statistical model, Weighted Moving Averagern(WMA). Besides, we can differentiate the two causes of correlational changernanomaly with an F1rn-score above 75rn%.