TRAFFIC FLOW PREDICTION PERFORMANCE COMPARISON BETWEEN ARIMA AND MONTE CARLO SIMULATION

  • Arif Hasnat Department of Civil Engineering, Rajshahi University of Engineering and Technology, Rajshahi, Bangladesh
  • Faysal Ibna Rahman Department of Civil and Environmental Engineering, University of Yamanashi, Kofu, Japan
Keywords: Time series analysis, ARIMA, Monte Carlo simulation, Forecast, Traffic flow

Abstract

Time series analysis and forecasting has become a major tool in numerous applications. It has the analysis and forecasting capacity of long term, intermediate term and short-term prediction. Monte Carlo simulation is also an another reliable name for forecasting in the simulation world. In this paper Monte Carlo simulation and time series Box-Jenkins ARIMA model (Autoregressive Integrated Moving Average) model are implemented to figure out the missing data points with full range traffic flow forecasting. As ARIMA deviates on non-seasonal data points or abrupt standard deviation, this two are used to find out the capacity of forecasting on those issue. Here using the traffic volume of previous 75% of a day, the least 25% traffic volume is forecasted. Then it is compared with the actual data. The ACF and PACF are plotted and checked the best model of fit for this data. The mean absolute relative error (MARE) and mean absolute percentage error (MAPE) are calculated and it is 8.51% and 2.19% respectively for time series and MAPE was found 6.66% with Monte Carlo simulation. ARIMA gives a nice forecast overall but fails at abnormal changing points whereas Monte Carlo overcomes this problem and suggests all probable possibilities.

Published
2019-06-04
Section
Transport & Logistics