afranck64 / benzlim

Benzlim

Head over to the QUICK START page if you are familiar with python configuration and the command line. If not, just read the following instructions. Do not forget to take a look at the FAQ page. Your questions could have been answered before.

Installation

Requirements

• Python 2.7.z (z >= 9) (https://www.python.org/downloads/)
• Git (https://git-scm.com/downloads)
• Python packages
  The required python packages can be installed using pipX.Y where X and Y are you majors python versions (pip2.7 for python2.7.*):
  • pipX.Y install -r requirements.txt (with admin rights, for a system wide installation) or
  • pipX.Y install --user -r requirements.txt (for a user only installation) This will install the following packages (and their dependencies) on your system:
    • coverage
    • numpy >= 1.13.3
    • pandas >= 0.18
    • python-dateutil

Configuration

• create a directory InformatiCup with mkdir InformatiCup
• change to the InformatiCup directory with cd InformatiCup
• create a InformatiCup2018 directory inside InformatiCup with mkdir InformatiCup2018
• change to the InformatiCup2018 directory with cd InformatiCup2018
• clone the official InformatiCup2018 project from the link InformatiCup2018 inside the InformtatiCup2018 directory with git clone https://github.com/InformatiCup/InformatiCup2018
• get back to the InformatiCup directory with cd ..
• copy the benzlim project inside the InformatiCup directory

Run

From the InformatiCup directory, run the command: python benzlim -h to get detailed help.
If your InformatiCup2018 folder is not located in the same directory as benzlim, you need to provide the optional argument --informaticup2018-dir when referring to it.
On the first run, the data will be trained, which can take some time.

Please note that the prediction_file needs to have the following format with the three parts between ; being "Use Gasoline prices until this date", "The time for the prediction" and "Gas station ID" respectively:

YYYY-MM-DD HH:MM:SS+HH;YYYY-MM-DD HH:MM:SS+HH;GAS_STATION_ID

Example prediction file:

2015-02-10 12:18:01+01;2015-02-15 21:18:01+01;24
2016-03-22 10:42:01+01;2016-03-22 10:43:01+01;46
2016-01-27 03:06:01+01;2016-02-26 18:06:01+01;14038
2015-06-11 14:30:02+02;2015-06-12 07:50:02+02;4160

Please note that the route_file needs to have the following format:

TANK_CAPACITY
YYYY-MM-DD HH:MM:SS+HH;GAS_STATION_I

Example route file:

3
2015-08-01 08:00:00+02;10957
2015-08-01 09:55:31+02;11108
2015-08-01 10:27:25+02;11172
2015-08-01 11:02:43+02;11150
2015-08-01 11:48:16+02;11152
2015-08-01 12:20:23+02;11238
2015-08-01 12:52:37+02;11320

Predict

• python benzlim predict $prediction_file

Routing with price prediction

• python benzlim route $route_file

Routing with predicted prices

• python benzlim route -g $predicted_prices $route_file

Train the model

• python benzlim train --force

Run the tests

• python benzlim test

Run the benchmark

• python benzlim benchmark

Run the coverage benchmark

• python benzlim coverage

Technical Documentation (German)

If you are interested in the technical aspects of benzlim and want to know more about how the fuel price predictions are done, the technical documentation would be what you are looking for. The text is in German.