Using ARIMA to Predict the Growth in the Subscriber Data Usage

Telecommunication companies collect a deluge of subscriber data without retrieving substantial information. Exploratory analysis this type will facilitate the prediction varied information that can be geographical, demographic, financial, or any other. Prediction therefore an asset in decision-making process telecommunications companies, but only if retrieved follows plan with strategic actions. The exploratory was implemented research to predict usage trends based on historical time-stamped data. predictive outcome unknown approximated using at hand. We have used 730 points selected from Insights Data Storage (IDS). These were collected hourly statistic traffic table and subjected growth usage. Auto-Regressive Integrated Moving Average (ARIMA) model forecast. In addition, we normal Q-Q, correlogram, standardized residual metrics evaluate model. This showed p-value 0.007. result supports our hypothesis predicting increase growth. ARIMA predicted 3 Mbps maximum 14 Gbps. experimentation, compared Convolutional Neural Network (CNN) achieved best results UGRansome performed better execution speed by factor 43 for more than 80,000 rows. On average, it takes 0.0016 s execute one row, 0.069 CNN same thus making 43× (0.0690.0016) faster provide road map so telecommunication productive improving their Quality Experience (QoE). study provides understanding seasonality stationarity involved usage’s growth, exposing new network concerns facilitating development novel models.