why 'learning a new network' cannot capture relationships?

[zhengycole]

I am using a CSV file with 12 cases and 70 features as input for "learning new network." However, it doesn't seem to capture any relationships between the entities.

I suspect this issue may be due to the limited size of my dataset. To address this, I duplicated the dataset five times, resulting in a new dataset with 72 cases and 70 features. With this expanded dataset, I was able to generate a network.

Could someone explain why I couldn't get useful results with my original dataset? Additionally, is my approach to expanding the dataset reasonable?

thanks a lot!

[shooltz[BayesFusion]]

Can you post your data file here? Alternatively, send me a download link using a forum private message.

[zhengycole]

This is the dataset.
The first row contains the feature identifiers. For each subsequent row, "1" indicates the presence of the feature in the given case.
Thank you for helping me.

[marek [BayesFusion]]

The Bayesian Search-based algorithms (Bayesian Search and GTT) learn no arcs, PC and Naive Bayes-based algorithms (NB, TAN and ANB) show plenty of dependencies. With 12 data records, whatever you will learn in terms of structure will be completely unreliable. Naive Bayes may do best if one of your variables is the class variable. Multiplying the data set is unlikely to help, as you will be dealing with overfitting. My advice is to collect more data. Such a small data set will be useful perhaps in refining your parameters once you get the structure and the initial values of parameters from an expert. Sorry to deliver a negative message.

Marek

[zhengycole]

Thank you very much! Your answer really helps me.
I will expand my dataset and seek advice from experts.
Again, I sincerely appreciate your response!