Nowadays, the increase in size and complexity of current cellular networks is complicating their operation, maintenance and optimization tasks. The user throughput has increased dramatically. However, such networks have become more susceptible to failure. Unfortunately, Bahir Dar cityâ€™s mobile data is mainly supported by Universal Mobile Telecommunications System (UMTS). It is challenging to locate the root causes of unsatisfactory user throughput manually for such a system, as it requires expertise and a longtime analysis. This thesis proposes low user throughput Root Cause Analysis (RCA) using a Deep Neural Network (DNN) based on a multilayer perceptron. Local interpretable model-agnostic explanation technique has been utilized to enable DNN for RCA of low user throughput by providing feature importance. Furthermore, the impact of features on the model outcome is studied to understand the causal effect on the user throughput using a partial dependency plot.rnThe effectiveness of the model has been validated using a test dataset and compared with reference models. Accordingly, the proposed DNN model has performed better compared to these models. The result of the proposed model performance evaluation is an accuracy of 87%, precision of 88%, recall of 86%, F1-score of 87% and area under the curve of 95%. RCA conducted on 10 selected cells with a poor throughput of 50 samples per cell has shown the root cause is due to poor channel quality reported by users. It is recommended to optimize the network to improve the channel quality.