← Back to Projects
Data AnalyticsApplied Math

Stochastic Traffic Flow Analysis

Poisson process modeling for traffic flow prediction and simulation.

Simulated vs Actual Traffic Flow
  • Created homogeneous and non-homogeneous Poisson process models to analyze and predict traffic flow patterns from a multi-scenario dataset of 20K+ entries across 90 scenarios.
  • Derived maximum likelihood estimates for parameter estimation to fit exponential distributions to interarrival times of traffic events.
  • Applied the Lewis–Shedler thinning algorithm to simulate NHPP realizations, enabling time-varying rate estimation.
    • Estimated Nonlinear Traffic Density
  • Evaluated model performance via Chi-Square and Kolmogorov–Smirnov tests (p-values: 0.089–0.103) and error metrics (Relative RMSE: 2.98–3.45%).
  • Validated Poisson process assumptions using diagnostic plots (histogram fits, autocorrelation checks, cumulative arrival comparisons) to confirm the model adequately captures traffic event dynamics.