COF-C03 Question 134
Single answerSnowflake MLA retail company stores historical sales data in Snowflake and wants to build a demand-forecasting solution without moving data out of the platform. The analytics team wants a low-code approach that can train and generate forecasts directly in Snowflake using SQL. Which Snowflake capability best fits this requirement?
- A
Use Snowflake Cortex AISQL functions to generate future numeric forecasts from historical sales tables
- B
Use Snowflake ML Functions, such as Forecast, to train and generate predictions directly in Snowflake
- C
Export the data to an external Python notebook because Snowflake does not support in-platform ML forecasting
- D
Create a materialized view on the sales table, because materialized views automatically produce time-series forecasts
Show answer and explanation
Correct answer: B
Explanation
The best answer is to use Snowflake ML Functions, specifically Forecast for a demand-forecasting scenario. Snowflake ML is designed to let teams build and run machine learning workflows close to the data, reducing data movement and simplifying operationalization. For SnowPro Core, candidates should recognize the distinction between Snowflake ML for predictive modeling tasks and Cortex AISQL for generative AI and language-focused inference. Materialized views are a performance feature, not an ML capability. This aligns with Snowflake documentation describing Snowflake ML as providing in-platform ML capabilities, including forecasting, while Cortex AISQL addresses LLM-powered SQL functions.
- A. Incorrect.
Incorrect. Cortex AISQL focuses on generative AI and language-related capabilities such as summarization, sentiment, extraction, translation, and similar LLM-based tasks. It is not the primary Snowflake feature for structured time-series forecasting of historical sales data in a low-code ML workflow.
- B. Correct.
Correct. Snowflake ML provides built-in ML Functions, including Forecast, that support in-platform model training and prediction for forecasting use cases. This matches the requirement to keep data in Snowflake and use a low-code SQL-based approach rather than exporting data to another environment.
- C. Incorrect.
Incorrect. A key Snowflake value proposition is bringing compute to the data, and Snowflake ML supports in-platform machine learning workflows for common use cases. Exporting data may still be possible in some architectures, but it is not required here and does not best meet the stated need.
- D. Incorrect.
Incorrect. Materialized views improve query performance by precomputing and storing query results. They do not train machine learning models or generate future forecasts. This option reflects a common misunderstanding between performance optimization features and analytics/ML capabilities.