ectools

ectools is a collection of the Python tools used to do the calculations in these Economic Complexity papers:

• Angelini, O., Cristelli, M., Zaccaria, A., & Pietronero, L. (2017). The complex dynamics of products and its asymptotic properties. PLoS ONE, 12(5).

• Angelini, O., & Di Matteo, T. (2018). Complexity of products: the effect of data regularisation.

I plan to release most or ideally all of the code publicly, and will incrementally add it to this repository. For the moment the code is simple enough that the docstrings should cover it. Please contact me for any queries regarding its use or any bugs that you may find.

At the moment, the repository contains:

• fc: Implementation of the Fitness-Complexity algorithm, a tied ranking algorithm, and the computations of RCA (Revealed Comparative Advantage following Ballassa) and logPRODY.
• SPSb: Implementation of the SPSb and Velocity SPSb algorithms (based on the convergence with NWKR proven in one of the papers above), and many utilities for backtesting, metaparameter selection, model stacking and error evaluation.
• NWKR: An implementation of the Nadaraya-Watson Kernel Regression in the style of scikit-learn. Compared to that found in scikit-learn, it supports np.nan values in the input and calculates the standard deviation of the predictions. Implemented in numba for speed
• parallelization: A wrapper around joblib for its most common use case. Mainly included as a dependency of NWKR.
• utilities: Utility functions included as a dependency.

Motivation

The code is published to help with reproducibility, and in support of the open source philosophy.

Code Example

import numpy as np
from ectools.NWKR import NWKR
X = np.array([[1,1],[1,1]])
y = np.array([1,1])
model = NWKR()
model.fit(X,y)
prediction, prediction_std = model.predict(X)
print(prediction, prediction_std)

Installation

The code has been tested with python 3.6. Just pull the code with

git clone git@github.com:ganileni/ectools.git

and install the requirements:

cd ectools

pip install -r requirements.txt

if the tests work correctly:

cd tests/

nosetests

you're good to go.

You can install the package with

python setup.py install

or only symlink the files in your python installation with

python setup.py develop

Contributors

Just me.

License

vanilla MIT license. Check out LICENSE.