Google Professional Machine Learning Engineer Question 176
Single answerGoogle Cloud PlatformYou are tasked with building a machine learning model on Google Cloud Platform to predict customer churn using a tabular dataset with categorical and numerical features. The dataset is relatively small, and you need to quickly prototype the model to evaluate its performance. Which framework or library would you choose for this scenario, and why?
- A
TensorFlow
- B
PyTorch
- C
scikit-learn
- D
Apache Spark
- E
JAX
Show answer and explanation
Correct answer: C
Explanation
For quick prototyping on small tabular datasets, scikit-learn is the most appropriate choice. It is lightweight, easy to use, and specifically designed for such tasks. The other options either introduce unnecessary complexity or are optimized for different use cases, such as deep learning or large-scale distributed computing.
- A. Incorrect.
TensorFlow is a powerful framework for building deep learning models, but it might be overkill for a small dataset and quick prototyping, especially for tabular data. It is better suited for large-scale deep learning or distributed training.
- B. Incorrect.
PyTorch is highly flexible and popular for deep learning research, but it is not the most efficient choice for quick prototyping or working with small tabular datasets. It requires more setup compared to simpler libraries.
- C. Correct.
scikit-learn is specifically designed for quick prototyping and works exceptionally well with tabular datasets. It provides a wide range of pre-built algorithms and tools for preprocessing, evaluation, and model selection, making it ideal for this scenario.
- D. Incorrect.
Apache Spark is designed for distributed data processing and machine learning on large-scale datasets. It is not optimized for small datasets, and using it in this scenario would introduce unnecessary complexity.
- E. Incorrect.
JAX is a library for high-performance numerical computing and is primarily used for advanced use cases, such as custom gradient computations and large-scale deep learning. It is not suitable for quick prototyping on tabular datasets.