foc-forecaster 0.1

Installation
=======

Ubuntu Linux (or other Debian-based distro)
-----------------------------------------------------------
Open a terminal and change directory to the location of this file.

Run the install_dependencies script:

$ sh install_dependencies.sh

Windows
------------
Best to install the python(x,y) bundle (http://code.google.com/p/pythonxy/).

Afterwards all the packages mentioned in the install_dependencies.sh file can be installed via pip. For example to install xlrd one would issue the following command inside Command Prompt:

$ pip install xlrd

Running
=======
Set the desired preferences in the configuration file:

./common/conf.py

Write down the crisis/normal years in the XLS file:

./odabir_uzoraka.xls

Position yourself inside the irb.foc.forecaster folder (pwd output just to show an example of the correct path):

$ cd irb.foc.forecaster
$ pwd
/media/Data/Drazen/Dropbox/dev/eclipse/w2/irb.foc.forecaster

Run the Python interpreter with the entry script run.py as an argument:

$ python run.py


Lay of the code
===============

forcaster - main module, use it to start the program

common
------
|- conf - configuration file with all the preferences
\- exceptions - all the custom exceptions are defined here

model - contains the data structures
-----
|- country - code and list of indicators
\- indicator - internal representation: list of dates, list of values

sources - represents the data sources available online.
-------
\- wb - extracts data from the World Bank

ai - classes regarding pattern recognition, train and test building etc.
--
|- input - parses XLS files to get crisis and normal period years
|- output - writes the dataset into a text file in a subgroup-discovery-friendly format
|- samples_set - the representation of the train and test datasets that can build samples based on the crisis/normal years input and indicators and countries specified in the conf file; fetches the data live from the World Bank API
|- preprocessor - processes the samples to extract useful features (min, max, slope...)
\- metadata - column labels and data type marks used when writing the dataset

tests - unit tests for individual modules
-----