Forecasting formulation model for amount of fault of the CPE segment on broadband network PT. Telkom using ARIMA method

Abstract

Currently, there is no forecasting activities undertaken within the PT. Telkom to predict Fault will occur in the broadband network, which prevents fault broadband networks are still focused on the process of correction, on the other hand, the accuracy of a prediction is the value that determines the quality of the model or algorithm is run that prediction. The results determine the accuracy of prediction and prevention process, this involves time and costs [1J. There are several methods for predicting fault will occur in broadband networks, among others; GARCH, ARM A, ARIMA [1], Kalman Filter and Hidden Markov [2J. These methods are methods autoregresi and nonlinear time series [1], the methods used in this research is the method ARIMA (Autoregressive Integrated Moving Average). This method was chosen because the CMSE (Cumulative Mean Square Error) value of this method is the most excellent [1] [2J. In the Broadband Network PT Telkom, 61.7% occurred in the segment disorders CPE (Customer Premise Equipment). Find a formulation to predict the amount of fault that would occur per type of disturbance in this segment with the aid of ARIMA method It will help to prepare a good resource skills, knowledge and cost in prevention activities, improving the quality of CPE devices and improve service quality PT Telkom. This research was conducted with the order or lag monthly data, run two scenarios, first; 24 observation data lag with 12 forecast lag, second; 30 observation data lag and 6 lag forecasts. This research resulted in the value of AR = 4, d = 1 and MA = 5. It appears that more observation data and the shorter lag forecast, then the results produced ARIMA will be more accurate. This can be seen in comparison error deviation and/or CMSE for each scenario.