Test/train data split overlap concern

[iAmGiG]

Problem

Both training and testing scripts use the same random split with random_state=17, which could lead to testing on training data if not managed carefully.

Code:

x_train, x_opt, x_test = np.split(df.sample(frac=1, random_state=17), ...)

Appears in:

• train_og.py:26-27
• test.py:37-38

Concern

If test.py is run on the same combined dataset used during training, it will test on data the model has already seen during training.

Recommendation

Use proper train/test split methodology:

• Separate hold-out test set
• Time-based split for network traffic
• Or different random seeds

Priority

MODERATE - Could affect validity of test results