Cross-validated DEA (CVDEA)

Installation

To facilitate installation on a personal computer, we recommend installing git (see: https://git-scm.com/downloads) and the Anaconda distribution (see: https://www.anaconda.com/products/individual). The steps to follow are based on these two installations.

Step1. Open the Anaconda Prompt console, place it in the desired directory for installation and enter the instruction:

https://github.com/MiriamEsteve/CVDEA.git

Step 2. Place us in the folder created by CVDEA, using "cd CVDEA", and execute the instruction:

python setup.py install

By following these two simple steps we will have the CVDEA package installed on our computer. The examples included below are also available in the CVDEA/test folder that has been created in the installation.

Import libraries

All the libraries in the repository are imported since they will be used in all the examples presented.

import numpy as np
import random
import cvdea

Generate MONO OUTPUT simulated data

CVDEA repository includes a simulated data generator module. It is used as an example of the use of the repository. For do that, the seed of the generator and the size of the dataset are stablished.

N = 50
sd = random.randint(0, 1000)
matrix = cvdea.Data(sd, N, 1).data

This dataset include the theorical output. In this sense, it is necessary save it in a differet variable.

matrixTheoric = matrix.copy()
matrix = matrix.iloc[:, :-1]

Create the CVDEA model

The creation of the CVDEA model consist on specify the inputs and outputs columns name (x and y), the depth configuration, i.e. the number of hyperplanes in each combination of supporting hyperplanes (D) and the number of different combinations of supporting hyperplanes (P). Once this is done, the model is created and fitted to build it.

Step 1. The name of the columns of the inputs and outputs in the dataset are indicated. If these ones don't exist in the dataset, the CVDEA model returns an error.

y = [matrix.columns[-1]]
x = list(matrix.drop(y, axis=1).columns)

Step 2. the depth configuration, i.e. the number of hyperplanes in each combination of supporting hyperplanes (D) and the number of different combinations of supporting hyperplanes (P) are specified.

D = 10
P = 10

Step 3. The creation and fit of the CVDEA model are done.

model = cvdea.CVDEA(matrix, x, y, D, P)
model.fit_CVDEA()

Predictions

The prediction of the CVDEA model can be with one dataset or with a single register of the dataset.

matrix_pred = model.prediction(matrix)