Prediction model

Prediction model

This application consists of 5 prediction models

model	Error(MSE)
Number of car prediction model	~ 14.05
Number of motorcycle prediction model	~ 2.184
Number of heavy vehicle prediction model	~ 0.497
Total number vehicle prediction model	~ 17.55
Traffic flow prediction model	~ 34.62

Selected algorithms

Every model uses Gated Recurrent Unit algorithms to construct and train a prediction model. Since traffic flow and the number of cars are highly dependent on the time of day, GRU algorithms that predict values from the pattern of the data in time-series are fit for the purpose of this application

Parameter

Every model takes 2 values as a parameter

Parameter	Description
time	Time in the day every 30 minutes from 0:00 - 23.30 calculated by hour + (minute / 60)
Weather	An Integer value that maps to the weather condition range from 0-3

Model construction

Models are constructed and trained by PyTorch using IPYNB, which can be found in notebook.zip in the repository root directory