Application of Dynamic Time Warping for Clustering Cellular Data Traffic Pattern

The demand for internet access has been increasing every year. According to data from the Indonesia Central Bureau of Statistics in 2022, there has been a significant rise since 2018, with the percentage of the population accessing the internet growing substantially. Cellular operators, as Internet Service Providers (ISP), need to analyze internet consumption within their networks, one of which is data traffic. Data traffic, typically measured in Bytes, is a measure for network productivity and has a direct impact on a company's business. The higher the network's data traffic consumption, the greater the potential revenue for cellular operators. Failure to analyze the patterns can result in serious consequences for cellular operators, including network capacity planning failures, a decline in customer expe-rience performance, and not optimal revenue. Therefore, cellular operators need to analyze data traffic consumption to understand the potential and behavior of their users. Unfortunately, data traffic consumption has characteristics which fluctuate over time, depending on user behavior. These characteristics present challenges for operators, operators require a method which can learn traffic data patterns in the time domain. In this study, a data traffic pattern analysis will be conducted using time-series clustering methods to analyze data traffic in 61 districts from one of the cellular operators in Indonesia. By utilizing the Dynamic Time Warping (DTW) algorithm, this study aims to identify the characteristics of each cluster based on pattern similarities. The characteristics of each cluster will highlight potential data traffic consumption according to specific time conditions in the selected districts.