skcriteria.datasets
package¶
The skcriteria.datasets
module includes utilities to load datasets.
- skcriteria.datasets.load_simple_stock_selection()[source]¶
Simple stock selection decision matrix.
This matrix was designed primarily for teaching and evaluating the behavior of an experiment.
Among the data we can find: two maximization criteria (ROE, CAP), one minimization criterion (RI), dominated alternatives (FX), and one alternative with an outlier criterion (ROE, MM = 1).
Although the criteria and alternatives are original from the authors of Scikit-Criteria, the numerical values were extracted at some point from a somewhere which we have forgotten.
Description:
In order to decide to buy a series of stocks, a company studied 5 candidate investments: PE, JN, AA, FX, MM and GN. The finance department decides to consider the following criteria for selection:
ROE (Max): Return % for each monetary unit invested.
CAP (Max): Years of market capitalization.
RI (Min): Risk of the stock.
- skcriteria.datasets.load_van2021evaluation(windows_size=7)[source]¶
Dataset extracted from from historical time series cryptocurrencies.
This dataset is extracted from:
Van Heerden, N., Cabral, J. y Luczywo, N. (2021). Evaluación de la importancia de criterios para la selección de criptomonedas. XXXIV ENDIO - XXXII EPIO Virtual 2021, Argentina.
The nine available alternatives are based on the ranking of the 20 cryptocurrencies with the largest market capitalization calculated on the basis of circulating supply, according to information retrieved from Cryptocurrency Historical Prices” retrieved on July 21st, 2021, from there only the coins with complete data between October 9th, 2018 to July 6th of 2021, excluding stable-coins, since they maintain a stable price and therefore do not carry associated yields; the alternatives that met these requirements turned out to be: Cardano (ADA), Binance coin (BNB), Bitcoin (BTC), Dogecoin (DOGE), Ethereum (ETH), Chainlink (LINK), Litecoin (LTC), Stellar (XLM) and Ripple (XRP).
Two decision matrices were created for two sizes of overlapping moving windows: 7 and 15 days. Six criteria were defined on these windows that seek to represent returns and risks:
xRv
- average Window return (\(\bar{x}RV\)) - Maximize: is the average of the differences between the closing price of the cryptocurrency on the last day and the first day of each window, divided by the price on the first day.sRV
- window return deviation (\(sRV\)) - Minimize: is the standard deviation of window return. The greater the deviation, the returns within the windows have higher variance and are unstable.xVV
- average of the volume of the window (\(\bar{x}VV\)) - Maximize: it is the average of the summations of the transaction amount of the cryptocurrency in dollars in each window, representing a liquidity measure of the asset.sVV
- window volume deviation (\(sVV\)) - Minimize: it is the deviation of the window volumes. The greater the deviation, the volumes within the windows have higher variance and are unstable.xR2
- mean of the correlation coefficient (\(\bar{x}R^2\)) - Maximize: it is the mean of the \(R^2\) of the fit of the linear trends with respect to the data. It is a measure that defines how well it explains that linear trend to the data within the window.xm
- mean of the slope (\(\bar{x}m\)) - Maximize: it is the mean of the slope of the linear trend between the closing prices in dollars and the volumes traded in dollars of the cryptocurrency within each window.
- Parameters
windows_size (7 o 15, default 7) – If the decision matrix based on 7 or 15 day overlapping moving windows is desired.
References
[VanHeerden et al., 2021b] [VanHeerden et al., 2021a] [Rajkumar, 2021]