Statistical Significant Test.
Budget: $30 – $250 USD
Using Friedman and Nemenyi post- hoc test, I want to assess the performance results of these algorithms statistically. It will be help me know whether the classifiers are significantly different from each other or not.
The null hypothesis (H0) considered in this case is that there is no performance difference among classifiers. While alternative hypothesis (HA) is that there is at least one classifier that performs significantly different than at least one other classifier for each performance metrics.
Secondly, I need to find which classifier pairs performs significantly different. Obtain the p value of all the pairwise comparisons using Nemenyi post-hoc test. Let’s assume the classifiers being tested are ‘LR’ and ‘SGDC’, ‘LR’ and ‘RF’ etc.
The null hypothesis (H0) considered in this case is that there is no performance difference among classifiers. While alternative hypothesis (HA) is that there is at least one classifier that performs significantly different than at least one other classifier for each performance metrics.
Secondly, I need to find which classifier pairs performs significantly different. Obtain the p value of all the pairwise comparisons using Nemenyi post-hoc test. Let’s assume the classifiers being tested are ‘LR’ and ‘SGDC’, ‘LR’ and ‘RF’ etc.