For mobile operators, delivering quality service to their customers is very important as servicernquality significantly affects their business. To achieve a consistent quality service delivery, theyrnneed to continuously monitor and analyze their network performance and timely address anyrnobtained performance drops. Performance drops in mobile networks can be observed in keyrnperformance indicators (KPIs) of spatially distributed cells with different magnitude and atrndifferent time periods. As it is practically difficult to address all performance drops simultaneously,rnit is preferable to make prioritized corrective actions starting from cells with critical drops, alsorncalled anomaly cells. To detect anomaly cells, different automated methodologies have beenrnproposed and analyzed. rnYet, ethio telecom still applies manual and subjective anomaly detection method where measuredrnKPI values are manually compared with fixed thresholds to determine if the measured values arernwithin defined required ranges or not. Cells with KPI values out of the required range are analyzedrnfor identifying performance drop root causes and taking corrective actions. The manual andrnsubjective anomaly detection method is prone to detection errors and is maintenance time,rnmanpower and then cost inefficient. These challenges of manual and subjective detection can bernimproved by applying advanced automatic methods based on machine learning algorithms. rnIn this thesis work, KNN based anomaly detection algorithms such as KNN classification, localrnoutlier factor (LOF) and connectivity outlier factor (COF) anomaly detection models isrnimplemented, and their comparative evaluation are made for Addis Ababa LTE cells. Therncomparison is made based on type of output, complexity and their true positive rate (TPR) for timernseries and cell level detections. Unlike KNN classification, LOF and COF do not need heavy datarnset training and are able to provide anomaly scores instead of two-class labels. Experimentationrnresults show that COF provides slightly better performance than the other models with negligiblernperformance difference. For instance, the performance of COF with respect to TPR for RRC setuprnsuccess rate in the experimentation is 97.91% for cell level detection and 88% for time seriesrndetection.