Error Discrepancy Between scikit-learn Versions: 1.2.1 vs. 1.4.2

[rohan472000]

Hi Reader,

The issue revolves around the handling of missing data (NaN values) within the dataset.

While running a random forest model on one system, I observed that the model execution terminated abruptly due to NaN values present in certain columns. Interestingly, when I replicated the same experiment on another system, the model executed without any errors despite the presence of NaN values in the data.

Upon further investigation, I discovered that the two systems were running different versions of scikit-learn, with one using version 1.2.1 (shown X_train has Nan) and the other using version 1.4.2 (didn't thrown any error). This discrepancy in behavior raises questions about the changes or updates implemented in the newer version of scikit-learn that may have influenced how NaN values are handled within the random forest model.

Please clarify to me what's the actual cause & its solution...Thanks in advance for help.

[adrinjalali]

We have been improving support for np.nan in recent releases. So the behavior you observe is expected. You can check our changelogs for more details.