You are tasked with building a model to predict customer churn. You have a table named in Snowflake with the following relevant columns: 'customer_id', 'login_date', , 'orders_placed', , and 'churned' (binary indicator). You want to engineer features that capture customer engagement over time using Snowpark for Python. Which of the following feature engineering steps, applied sequentially, are MOST effective in creating features indicative of churn risk?

• A. 1. Calculate the number of days since the customer's last login, and use nulls instead of negative numbers to indicate inactivity. 2. Calculate the rolling 7-day average of 'orders_placed' using a window function, partitioning by 'customer_id' and ordering by 'login_date'. 3. Calculate the slope of a linear regression of page_views' over time for each customer, indicating the trend in engagement using Snowpark ML. 4. Calculate the percentage of weeks the customer logged in. 5. Create a feature showing standard deviation of page_views per customer over the last 90 days.
• B. 1. Calculate the maximum 'page_views' in a single day for each customer. 2. Calculate the total number of days with no 'login_date' for each customer. 3. Create a feature indicating if a customer has ever placed an order. 4. Use a simple boolean for the 'subscription_type' column.
• C. 1. Calculate the average 'page_views' per day for each customer. 2. Calculate the total number of for each customer. 3. Create a feature indicating whether the customer has a premium subscription ('subscription_type' = 'premium').
• D. 1. Calculate the average 'page_views' per week for each customer over the last 3 months using a window function. 2. Calculate the recency of the last order (days since last order) for each customer. 3. Create a feature indicating the change in average daily page views over the last month compared to the previous month. 4. Create a feature showing standard deviation of page_views per customer over the last 90 days.
• E. 1. Calculate the total 'page_views' and 'orders_placed' for each customer without considering time. 2. Use one-hot encoding for the 'subscription_type' column.

A data scientist is analyzing website click-through rates (CTR) for two different ad campaigns. Campaign A ran for two weeks and had 10,000 impressions with 500 clicks. Campaign B also ran for two weeks with 12,000 impressions and 660 clicks. The data scientist wants to determine if there's a statistically significant difference in CTR between the two campaigns. Assume the population standard deviation is unknown and unequal for the two campaigns. Which statistical test is most appropriate to use, and what Snowflake SQL code would be used to approximate the p-value for this test (assume 'clicks_b' , and are already defined Snowflake variables)?

• A. A one-sample t-test, because we are comparing the sample mean of campaign A to the sample mean of campaign Snowflake code: 'SELECT t_test_lsamp(clicks_a/impressions_a - clicks_b/impressions_b, 0)'
• B. Az-test, because we know the population standard deviation. Snowflake code: 'SELECT normcdf(clicks_a/impressions_a - clicks_b/impressions_b, O, 1)'
• C. A paired t-test, because we are comparing two related samples over time. Snowflake code: 'SELECT t_test_ind(clicks_a/impressions_a, 'VAR EQUAL-TRUE')
• D. An independent samples t-test (Welch's t-test), because we are comparing the means of two independent samples with unequal variances. Snowflake code (approximation using UDF - assuming UDF 'p_value_from_t_stat' exists that calculates p-value from t-statistic and degrees of freedom):
  
• E. An independent samples t-test, because we are comparing the means of two independent samples. Snowflake code: SELECT
  

A data scientist is tasked with predicting customer churn for a telecommunications company using Snowflake. The dataset contains call detail records (CDRs), customer demographic information, and service usage data'. Initial analysis reveals a high degree of multicollinearity between several features, specifically 'total_day_minutes', 'total_eve_minutes', and 'total_night_minutes'. Additionally, the 'state' feature has a large number of distinct values. Which of the following feature engineering techniques would be MOST effective in addressing these issues to improve model performance, considering efficient execution within Snowflake?

• A. Calculate the Variance Inflation Factor (VIF) for each CDR feature and drop the feature with the highest VIE Apply frequency encoding to the 'state' feature.
• B. Apply Principal Component Analysis (PCA) to reduce the dimensionality of the CDR features ('total_day_minutes', 'total_eve_minutes', 'total_night_minutes') and use one-hot encoding for the 'state' feature.
• C. Apply min-max scaling to the CDR features to normalize them and use label encoding for the 'state' feature. Train a decision tree model, as it is robust to multicollinearity.
• D. Use a variance threshold to remove highly correlated CDR features and create a feature representing the geographical region (e.g., 'Northeast', 'Southwest') based on the 'state' feature using a custom UDF.
• E. Create interaction features by multiplying 'total_day_minutes' with 'customer_service_calls' and applying a target encoding to the 'state' feature.

You're developing a fraud detection system in Snowflake. You're using Snowflake Cortex to generate embeddings from transaction descriptions, aiming to cluster similar fraudulent transactions. Which of the following approaches are MOST effective for optimizing the performance and cost of generating embeddings for a large dataset of millions of transaction descriptions using Snowflake Cortex, especially considering the potential cost implications of generating embeddings at scale? Select two options.

• A. Implement caching mechanism based on a hash of transaction description if transaction description does not change then no need to recompute the emebeddings again.
• B. Use a Snowflake Task to incrementally generate embeddings only for new transactions that have been added since the last embedding generation run.
• C. Generate embeddings on the entire dataset every day to capture all potential fraudulent transactions and ensure the model is always up-to-date.
• D. Create a materialized view containing pre-computed embeddings for all transaction descriptions.
• E. Generate embeddings using snowflake-cortex-embed-text function, using the OPENAI embedding model

You are building a time-series forecasting model in Snowflake to predict the hourly energy consumption of a building. You have historical data with timestamps and corresponding energy consumption values. You've noticed significant daily seasonality and a weaker weekly seasonality. Which of the following techniques or approaches would be most appropriate for capturing both seasonality patterns within a supervised learning framework using Snowflake?

• A. Decomposing the time series using STL (Seasonal-Trend decomposition using Loess) and building separate models for the trend and seasonal components, then combining the predictions.
• B. Creating lagged features (e.g., energy consumption from the previous hour, the same hour yesterday, and the same hour last week) and using these features as input to a regression model (e.g., Random Forest or Gradient Boosting).
• C. Applying exponential smoothing directly to the original time series without feature engineering.
• D. Using Fourier terms (sine and cosine waves) with frequencies corresponding to daily and weekly cycles as features in a regression model.
• E. Using a simple moving average to smooth the data before applying a linear regression model.