Exam AWS Certified Machine Learning - Specialty topic 1 question 313 discussion

Correct B. It seems there was a typographical error in the provided options. "validation'll" is not a valid metric name. It appears to be an error or a typo. B. Tune the csv_weight hyperparameter and the scale_pos_weight hyperparameter by using automatic model tuning (AMT). Optimize on {"HyperParameterTuningJobObjective": {"MetricName": "validation:f1", "Type": "Maximize"}}.

the ll in the option B is recall, there must be some bug make the system miss some charactor.

Given the imbalanced nature of the dataset where only 5% of customers return items, the focus should be on maximizing the model's ability to correctly identify the returned items, which corresponds to maximizing the recall or F1 score. Option C and D aim to optimize the F1 score, but option D specifies minimizing the F1 score, which is incorrect. C. Tune all possible hyperparameters by using automatic model tuning (AMT). Optimize on {"HyperParameterTuningJobObjective": {"MetricName": "validation:f1", "Type": "Maximize"}}.

B. Tune the csv_weight hyperparameter and the scale_pos_weight hyperparameter by using automatic model tuning (AMT). Optimize on {"HyperParameterTuningJobObjective": {"MetricName": "validation:recall", "Type": "Maximize"}}. The dataset is imbalanced, with only 5% of customers returning items (or the positive class). The goal is typically to capture as many instances of the minority class (returned items) as possible, even at the expense of some false positives. Option D might be incorrect, as the goal is to maximize the model's ability to capture instances of returned items, not minimize the F1 score.

A: No - tuning all hyperparemeters requires compute - not very cost effective B: No - Log Loss has to be not applicable to imbalanced dataset C: No - tuning all hyperparemeters requires compute - not very cost effective D: Yes - F1 metric combines both precision and recall which is more suitable for unbalanced datasets