<p>Through this web platform you can verify the results obtained from the learning algorithms applying the statistic tests to the experiments, which, among other uses, support the decision making process (the election of the most suitable algorithm, for example). At present, this is the most widely accepted method for validating this kind of experiments.</p>

<p>The tests are based in the verification of hypothesis. For example: a test could verify the null hypothesis or \({H_0}\) that all the averages of the data obtained from an algorithm are equal. These tests, verify whether or not this initial hypothesis is valid. For that, an statistic is obtained (subject to a particular statistic distribution), and the probability of obtaining another statistic at least as extreme as the former (p-value). If this p-value is less than a given (\({\alpha}\) value), the null hypothesis will be rejected, otherwise accepted.</p>

<p>The image shows the two regions in which a probability distribution function is divided (in this example the chi-squared distribution) after setting a significance level (\({\alpha}\) value): the acceptance region and the rejection region. The p-value, determines to which of these regions the statistic belongs to. The critical point (in this example, 1 point), is the value separating the two regions.</p>

<i><b>Final Degree Project by: Adrián Canosa Mouzo. Universidad de Santiago de Compostela - Escuela Técnica Superior de Ingeniería.</b></i>

<p>STAC</p>

<p>Web plataform for the statistical comparison of algorithms</p>

<p>Version 1.0</p>

<br/>

<p>Final Degree Project developed by by: Adrián Canosa Mouzo. Universidad de Santiago de Compostela - Escuela Técnica Superior de Ingeniería.</p>

<p>STAC logo was designed by <ahref="https://github.com/angelpinheiro">Ángel Piñeiro Souto</a></p>