Skip to content

JonnoFTW/traffic-prediction

master
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

LSTM Based Traffic Prediction

This project seeks to use LSTM for traffic prediction using the Keras frontend for Theano. Hyperparameter optimisation is used to find the best set of parameters for the network.

Usage

Run:

pip install -r requirements.txt

Then edit main.py so that it uses your own parameters for the network. It will attempt to store hyperparameter results in mongodb. You can use show_results.py to view them. Please keep in mind this for experimentation only and not appropriate for production.

Run using:

python main.pymain.py <steps eg. 1> </path/to/data.csv> <train_test_split eg. 0.75> <num_epochs eg. 10> <model_trials eg. 20>

The CSV should have a the following format:

timestamp,16,17,18,19,20,21
2011-12-31 23:55:00,4,6,8,13,3,0
2012-01-01 00:00:00,5,7,8,10,3,2
2012-01-01 00:05:00,3,3,7,15,1,2
2012-01-01 00:10:00,9,3,3,7,1,1
2012-01-01 00:15:00,3,4,12,15,2,0
2012-01-01 00:20:00,7,5,11,19,2,2

Basically, the first column must be a timestamp and the rest integer values.

Scripts are also provided to speed up processing via Sun Grid Engine. Though they are unreliable because of the length of time it takes for theano to compile the models. You will also need CUDA and a compatible card if you want the speedups they provide.

About

Predict traffic flow with LSTM. For experimental purposes only, unsupported!

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published