How to utilize prob-related methods of ECM classifier

[Ramin1368]

Hi

I am utilizing the ECM classifier as my unsupervised classifier for my problem but I keep getting error while calling them that I do not understand why:
ecm.fit(df_feature_vectors)
log_m_probablity = ecm.log_m_probs

which gives the following error: ValueError: Expected input with 6 features, got 5 instead
while my feature_vector has only 5 features. and also upon using ecm.prob, got the following error:
ValueError: Expected input with 11 features, got 5 instead

Interestingly, every time, I run this, it expects 5 more features like expected 16 features, 21 features, .....

what is the solution in order to use these methods such as log_m_probs, prob, log_u_probs, etc.???
Also one more question regarding this is that as I was employing the prodict method: links_pred = ecm.predict(df_feature_vectors) where df_feature_vectors = comparer.compute(All_Index_Pairs, df), it threw error such that the labels had to be either one or zero and I had to use binarizer to make the labels either one or zero in order to avoid the error. why can't the labels be between zero and one?

[Dragut]

I am wondering how could ECM classifier which is Fellegi-Sunter model based could give a Bayesian probability result in the first place. Only Bayesian models could give probability matching result as said Professor Peter Christensen.