MLS-C01 Question 147
Select 3A data science team at your organization has built a binary classification model to predict whether a customer will purchase a product or not. The team has tried several algorithms, including logistic regression, decision trees, and XGBoost. The XGBoost model consistently outperforms the others. Now, the team wants to further improve model performance. Which of the following strategies should the team consider?
- A
Perform hyperparameter tuning on the XGBoost model using SageMaker Automatic Model Tuning.
- B
Use transfer learning by fine-tuning a pre-trained convolutional neural network (CNN) on the dataset.
- C
Combine XGBoost with other models like logistic regression and decision trees in an ensemble approach.
- D
Cluster the dataset using k-means to create additional features for the model.
- E
Switch to using a large language model (LLM) to solve the classification problem.
Show answer and explanation
Correct answers: A, C, D
Explanation
To improve the performance of an already strong XGBoost model, techniques like hyperparameter tuning, ensemble methods, and feature engineering (e.g., creating additional features using k-means clustering) are effective strategies. Transfer learning using CNNs and switching to LLMs are not appropriate for this binary classification task, as they are tailored for image data and NLP, respectively.
- A. Correct.
Hyperparameter tuning is a standard and effective way to improve the performance of a machine learning model, including XGBoost. SageMaker Automatic Model Tuning can automate this process, making it a valid option.
- B. Incorrect.
Transfer learning with CNNs is primarily used for image data. Since this is a binary classification task, and no mention of image data is provided, this option is not applicable.
- C. Correct.
Combining XGBoost with other models like logistic regression and decision trees in an ensemble approach can improve model performance by leveraging the strengths of different algorithms.
- D. Correct.
Using k-means clustering to create additional features can introduce new insights into the data, potentially improving model performance. This is a valid strategy to explore.
- E. Incorrect.
Large language models (LLMs) are designed for natural language processing tasks and are not suitable for this binary classification task, which does not involve textual data.