DSA-C02 Exam Guide: AWS Certified Data Analytics – Specialty Preparation
                     Snowflake
DSA-C02
ExamName: SnowPro Advanced: Data Scientist Certification Exam
Exam Version: 6.0
Questions & Answers Sample PDF
(Preview content before you buy)
Check the full version using the link below.
https://pass2certify.com/exam/dsa-c02
Unlock Full Features:
Stay Updated: 90 days of free exam updates
Zero Risk: 30-day money-back policy
Instant Access: Download right after purchase
Always Here: 24/7 customer support team
https://pass2certify.com//exam/dsa-c02 Page 1 of 7
Question 1. (Single Select)
A marketing analyst at a retail company is using Snowflake to perform customer segmentation using
unsupervised learning. They have a table 'CUSTOMER TRANSACTIONS with columns: 'CUSTOMER ICY,
'TOTAL SPENT, 'AVG ORDER VALUE 'NUM_TRANSACTIONS, and ‘LAST PURCHASE DATE. They want
to use k-means clustering to identify distinct customer segments based on their spending behavior. The
analyst wants to scale 'TOTAL SPENT' and 'AVG_ORDER VALUE' to the range [0, 1] before clustering.
Which of the following SQL statements, leveraging Snowflake's capabilities and unsupervised learning best
performs this task and stores results in table ‘CUSTOMER SEGMENTS'?
A: Option A
B: Option B
C: Option C
D: Option D
E: Option E
Answer: B
Explanation:
Option B correctly implements k-means clustering using and performs Min-Max scaling within the query
using window functions to calculate the minimum and maximum values for 'TOTAL_SPENT and This scales
the features to a range between 0 and 1 before clustering, preventing features with larger magnitudes from
dominating the clustering process. Option A requires preprocessing which isnt self contained. Options CID
and E are syntactically incorrect and does not implement Min-Max Scaling.
Question 2. (Multi Select)
https://pass2certify.com//exam/dsa-c02 Page 2 of 7
A data scientist is tasked with predicting customer churn for a telecommunications company using
Snowflake. The dataset contains a mix of categorical and numerical features, including customer
demographics, service usage, and billing information. The target variable is 'churned' (binary: 0 or 1).
Which of the following steps are crucial to address potential issues and ensure optimal performance of a
supervised learning model (e.g., Logistic Regression or Gradient Boosted Trees) deployed within
Snowflake using Snowpark and external functions?
A: Applying one-hot encoding to categorical features before training the model, and ensuring the same
encoding is applied during inference via a Snowflake UDE
B: Scaling numerical features using StandardScaler or MinMaxScaler within a Snowpark DataFrame, and
saving the scaler parameters to apply consistently during inference using Snowflake UDF.
C: Ignoring missing values in the dataset, as Snowflake automatically handles them during model training.
D: Training the model locally using all available data and then deploying the serialized model to Snowflake
as an external function without any pre-processing pipeline.
E: Splitting the data into training and validation sets using 'SNOWFLAKML.RANDOM SPLIT to ensure
reliable model evaluation.
Answer: A, B, E
Explanation:
Options A, B, and E are essential for robust supervised learning. One-hot encoding (A) converts categorical
data into a numerical format suitable for many algorithms. Feature scaling (B) is crucial for algorithms
sensitive to feature ranges, like Logistic Regression and those using gradient descent. Ignoring missing
values (C) is generally detrimental. Deploying a model without preprocessing (D) will lead to incorrect
predictions during inference. Splitting the data into training/validation set using
'SNOWFLAKE.ML.RANDOM_SPLIT (E) is essential for fair evaluation of model performance on unseen
data.
Question 3. (Multi Select)
A data science team is using Snowflake to store historical sales data, including and promotion_spend'.
They want to predict future sales based on these features using linear regression. However, they suspect
and have a non-linear relationship. Which of the following strategies would be MOST effective in
https://pass2certify.com//exam/dsa-c02 Page 3 of 7
addressing this non-linearity within Snowflake without exporting data to an external platform?
A: Apply a logarithmic transformation to the ‘unit_price’ column within Snowflake using before using it in the
linear regression model.
B: Create a new feature in Snowflake that is the square of ‘unit_price’ (i.e., ‘unit_price and include both and
its squared term in the linear regression model.
C: Use Snowflake's built-in feature store capabilities to engineer a custom feature that quantizes into
discrete price tiers (e.g., low, medium, high) and use those tiers as categorical variables in the linear
regression.
D: Fit separate linear regression models for different ranges of 'unit_price’. This involves segmenting the
data based on price bands and training a unique model for each segment directly in Snowflake.
E: Export the data to a Python environment, perform polynomial regression using scikit-learn, and then
import the model's coefficients back into Snowflake for prediction.
Answer: B, D
Explanation:
Options B and D are most effective. Option B addresses non-linearity by introducing polynomial features.
Option D handles non- linearity by creating piecewise linear models. Option A might help, but squaring is
often a more robust approach. Option C changes it to categorization, which might not capture all variation.
Option E is undesirable because the question asks to solve it within Snowflake.
Question 4. (Multi Select)
You are building a linear regression model in Snowflake to predict customer churn based on historical data’.
Your data includes features like 'total _ purchases', , 'average_rating' , and a target variable 'churned' (0 or
1). You've noticed that 'total_purchases' has a very high range compared to the other features. What
preprocessing steps should you take in Snowflake to improve model performance and stability and why?
A: Apply min-max scaling to all features using the formula: '(feature_value - MIN(feature)) / (MAX(feature) -
MIN(feature))' within Snowflake SQL.
B: Apply standardization (Z-score normalization) to all features using the formula: '(feature_value -
AVG(feature)) / STDDEV(featurey within Snowflake SQL.
C: Drop the ‘total_purchaseS feature because its high range will negatively impact the linear regression
model.
https://pass2certify.com//exam/dsa-c02 Page 4 of 7
D: Apply robust scaling using interquartile range to feature ‘total_purchases’ using the formula :
'(feature_value - MEDIAN(feature)) / IQR(featurey within Snowflake SQL. Where IQR is the interquartile
range of total_purchases.
E: Apply one-hot encoding to the 'churned' feature, creating separate columns for ‘churned_ff and
‘churned_1'
Answer: A, B, D
Explanation:
Options A, B and D are valid preprocessing steps. Min-max scaling and standardization can help normalize
the range of features, preventing ‘total_purchases’ from dominating the model. Robust scaling also handles
outliers that may be there for total_purchases. Dropping ‘total_purchases' (option C) might lead to loss of
important information. One-hot encoding on the target is unnecessary for linear regression and is a
misconception about classification tasks.
Question 5. (Multi Select)
You have built a linear regression model in Snowflake using the SNOWFLAKE.ML.REGRESSORS.LINEAR
REGRESSION function to predict house prices based on features like square footage, number of
bedrooms, and location. The model appears to be performing well on the training data, but you suspect it
might be overfitting. Which of the following techniques can you implement directly within Snowflake (without
relying on external tools) to mitigate overfitting and improve the model's generalization performance?
A: Implement L1 regularization (Lasso) by adding a penalty term to the cost function based on the absolute
values of the coefficients directly within the SNOWFLAKE.ML.REGRESSORS.LINEAR REGRESSION
function.
B: Increase the size of the training dataset by generating synthetic data using techniques like SMOTE
directly within Snowflake.
C: Use cross-validation techniques (e.g., k-fold cross-validation) by creating a stored procedure that
partitions the data and trains/evaluates the model on different folds within Snowflake.
D: Reduce the number of features used in the model by performing feature selection using techniques like
recursive feature elimination within Snowflake.
E: Decrease the learning rate of the gradient descent algorithm used by
SNOWFLAKML.REGRESSORS.LINEAR_REGRESSION to allow the model to converge more slowly.
Answer: C, D
https://pass2certify.com//exam/dsa-c02 Page 5 of 7
Explanation:
Options C and D can be implemented. K-fold cross-validation helps to have better evaluation. Recursive
feature elimination would select the best features. Option A, Snowflake's built-in linear regression currently
does not expose L1 regularization. Option B: SMOTE is used to generate synthetic data for minority class
problem. Option E is not a direct option available.
https://pass2certify.com//exam/dsa-c02 Page 6 of 7
Need more info? Check the link below:
https://pass2certify.com/exam/dsa-c02
Thanks for Being a Valued Pass2Certify User!
Guaranteed Success Pass Every Exam with Pass2Certify.
Save $15 instantly with promo code
SAVEFAST
Sales: [email protected]
Support: [email protected]
https://pass2certify.com//exam/dsa-c02 Page 7 of 7