1 of 12

Forecasting New Jersey

Train Delays for Rail Track Upgrades

​

Project by Stephanie Cheng and Shreya Bansal

ON TRACK

2 of 12

NJ Transit Delays

Data Range:

May-August 2018

​

Newark Penn Station experienced over 2,000 hours of delays cumulatively over 4 months

​

3 of 12

Crowded trains

Source: Kevin R. Wexler-NorthJersey.com

4 of 12

Climate stress on train tracks

Source: Railroadrails.com

5 of 12

Proposal: Prevention is better than cure

NJT Capital Use program (2010)

Source: Transit.dot.gov

Assumption: trains are delayed mostly due to climatic stress and forecasting when delays will occur will allow NJ Transit to optimize repair funding more efficiently and prevent delays.

Source: Bloomberg

6 of 12

7 of 12

8 of 12

9 of 12

Model: Variables

  • Data range: First 5 weeks in July and August 2018
  • A space-time panel and weather panel to predict delayed minutes
  • Variables:
    • Dependent: Delay minutes

​

    • Independent:

​

CLIMATE

Temperature

Precipitation

Wind Speed

Visibility

TIME

Day of the week

Hour

Time of Day

​

TIME LAG

hour lags

day lag

holiday lag

​

SPACE

Station Name

Line

​

​

10 of 12

Models: Regression

​

Regression 3 <- lm (delay_minutes ~ station + hour + day of the week + Temperature + Precipitation + Visibility + Wind_Speed + lagHour + lag2Hours +lag3Hours +lag12Hours + lag1day + holidayLag + holiday

Model Testing:

K-Fold Cross Validation

  • RMSE (Root Mean Squared Error)
  • Rsquared
  • MAE (Mean Absolute Error)

11 of 12

Model Accuracy

Further Explorations: Heat Island Effect, Tree Cover

12 of 12

Benefits:

  • Proactive approach to track maintenance
  • Cost reducing by reduced delay minutes and on-time service
  • Higher resilience towards unpredictable climatic stressors
  • Higher customer satisfaction with on-time service

​

​

​

​

​

Conclusion