A Multi-dimensions Data Traffic Forecasting Model for Rural Areas

Traffic forecasting is the baseline for planning future network. A failure to forecast traffic demand might lead to sloppy network design which needs to be rearranged later. That such rearrangement will require additional cost. The network infrastructure in rural area is not attractive for commercial based broadband network providers. But, it is important to provide internet access for villagers since internet may give positive impacts on social development. Hence, the mobile traffic demand forecast method needs wider perspectives other than technical records of network providers only. This research introduces a multidimensional model to predict data traffic demand in the future, by combining government spatial planning, demography statistics, and network records. Two different areas were studied and analyzed: Panimbang and Leuwidamar districs, which are located in Banten Province, Indonesia. Main parameter to compare single- and multi-dimensions model is the areal traffic demand. For Panimbang, the areal traffic demand for multi-dimensions model has 24 Mbps/km(2) higher, compared to the single-dimension model. For Leuwidamar, the demand for multi-dimension is 10 Mbps/km(2) lower, compared to the single-dimension. In Panimbang, a pessimistic forecasting might not be a big problem since adding cells in a dense area is not costly, however for Leuwidamar, multi-dimensions models could help to design a more efficient network since the single-dimension model is too optimistic and leads to a high capital investment for providers. This multidimensional model suits best for remote and sparsely distributed users(Leuwidamar). However, it might gives no high impact for residential or urban areas (Panimbang).