Abstract: At present, driven by the rapid development of artificial intelligence technology, the wireless network operation and maintenance field also gradually tries to use algorithm to assist manual work, to increase operation and maintenance efficiency and reduce operation and maintenance cost. This paper takes China Telecom LTE cell of a province on the southeast coast as an example, three key indicators of average activated users, downlink user traffic and average RRC connected users in KPI data were selected to make a trend forecast for the coming week. The advantages and disadvantages of ARIMA and LSTM in machine learning are compared.Finally, 3000 training samples were used to establish the LSTM algorithm and the maximum prediction accuracy of the above three indexes was 92%, 71% and 67.5%, respectively. It can be seen that the prediction effect is the best on the average number of activated users index, which is more than 80%. In the end, the algorithm is extended to forecast the average number of active users in 14016 residential areas in the province to further verify the effect of the algorithm.

Key words: LTE network, trend forecast, LSTM, ARIMA, artificial intelligence