5 eltcn algorithm
Evolving Long-term Cognitive Networks (ELTCN) algorithm creates a fully connected FCM between the features and classes, similar to MP algorithm, with the main difference - none of the features is dropped by the learning algorithm, where in MP algorithm the feature which is the least important for making the decision, is dropped.
Algorithm:
For more information about the algorithm, please check out Pattern classification with Evolving Long-term Cognitive Networks Gonzalo Npoles Agnieszka Jastrzebska Yamisleydi Salgueiro
from fcmpy.ml.classification.eltcn import run
Results
Results are stored in the result variable. Model loss is automatically plotted below. Same as possible weights range value. The algorithm was run on the iris dataset. We can see that plots are shown ''in pairs'' - algorithm runs k-fold (default = 5) validation and outputting results for each fold below. In the Results we have an access to: all individual weights as well as average features and class weights.
path = 'fcmpy/data'
# The only necessary parameter to run the algorithm is path, which is the directory containing all .arff files
results = run(path)
# ELTCN algorithm automatically plots the results for each fold (default 5). It plots all the weights values and loss function of of learning process.
