Exam AI-900 topic 1 question 2 discussion

Vote for 'B' A. Use features for training and labels for evaluation. ==> Wrong (Training need both features and labels) B. Randomly split the data into rows for training and rows for evaluation. ==> Correct (few data used for training and on few data we can evaluate performance of model) , (so split in rows) C. Use labels for training and features for evaluation. ==> Wrong (for training we need both features and labels) D. Randomly split the data into columns for training and columns for evaluation. ==> Wrong (Data will be split row -wise)

On Exam 04.25.25

if we have a dataset, we should consider splitting the dataset into training and validation/test. When you are splitting, you should consider selecting the random rows for training and validation, because think of your dataset as 200 rows, you picked the first 160 rows for training and the last 40 rows for testing, what if those 40 rows hold a pattern that is not found in the training data? hence, the model can't learn some patterns from our dataset. So, randomly picking the rows is necessary

Came on 28-Mar-2025 exam

Came on 28-Mar-2025 exam

Randomly split the data into rows for training and rows for evaluation. ==> Correct (few data used for training and on few data we can evaluate performance of model) , (so split in rows)

rows are split, columns are not split.80:20 or 70:30 ratio for training and testing

B is correct

correct

vote foe C

B. Randomly split the data into rows for training and rows for evaluation.

Answer is B

training vs evaluation ratio: 70% vs 30%

B is correct

B makes sense

B is correct

Split the data for training and testing. randomly 75-25 split on row may be a good idea.