You are tasked with developing a machine learning model to classify different species of flowers based on their physical measurements using the Iris dataset. After training your model, you notice that it performs well on the training set but poorly on new, unseen data. What is the most likely explanation for this issue, and how might you address it?

The model is probably overfitting, meaning it has learned the specific quirks of the training data instead of the true patterns that apply to all flowers. This happens when the model is too complex or trained for too many epochs, so it memorizes noise rather than useful features. To fix it, try simplifying the model, reducing the number of layers or neurons, or adding regularization such as dropout or L2 penalties, and use techniques like cross‑validation to tune hyperparameters. For example, if a network with 200 hidden units works well on the training set but only scores 50 % on a held‑out set, reducing the units to 20 and adding dropout can raise the test accuracy to around 95 %.