CS 326 REVISION EXAMS 2025/2026 QUESTIONS AND
SOLUTION RATED A+
✔✔Content-Based Filtering - ✔✔Based on a description of the item and a profile of the
user's preferences, recommend items that are similar to those a user liked in the past
✔✔User Matrix - ✔✔A set of data points that represents all of a user's preferences
✔✔Item Matrix - ✔✔A set of data points that represent every item's content. Items can
be represented as a set of features
✔✔User-Item Similarity Matrix - ✔✔A set of data points that represents the similarity
between user preferences and items. These similarities can be weighted by previous
reviews or ratings from the user
✔✔Similarity metrics - ✔✔Similarity is commonly determined as the dot product
between two vectors, but different similarity metrics can be used, like Euclidean,
Manhattan, or Cosine Distance
✔✔What should you use to evaluate if content-based filtering is leading to increased
user engagement? - ✔✔A/B testing
✔✔Pros of content-based filtering - ✔✔Easy to implement, no need for data on other
users, can recommend niche items, can provide explanations for recommendations
✔✔Cons of content-based filtering - ✔✔Limited to the user's past preferences, limited to
the item's features, can overfit to the user's preferences, cold-start problem
✔✔Types of collaborative filtering - ✔✔User-user and item-item
✔✔User-User Collaborative Filtering - ✔✔Based on the idea that users who have
agreed in the past will likely agree in the future
✔✔Item-Item Collaborative Filtering - ✔✔Based on the idea that item reviews are often
grouped together. Different from content-based filtering because it has nothing to do
with the item characteristics. Ex: user 1 bought an orange, apple, and banana, user 2
bought an apple and orange, user 3 bought an apple --> recommend orange to user 3
✔✔User-Item Matrix - ✔✔A matrix that represents the relationship between users and
items. You can find the nearest neighbors from two perspectives: user-user (rows) or
item-item (columns)
✔✔Methods to fill missing values in the user-item matrix - ✔✔Use the scale's mean,
matrix factorization
, ✔✔Matrix Factorization - ✔✔A technique to break down a matrix into the product of
multiple matrices. It is used in collaborative filtering to estimate the missing values in the
user-item matrix
✔✔Alternating Least Squares - ✔✔An optimization algorithm used to minimize the error
between the predicted and actual ratings in the user-item matrix (when filled in with
Matrix Factorization)
✔✔Pros of collaborative filtering - ✔✔Can recommend items that the user has not seen
before, can recommend items that are popular among similar users
✔✔Cons of collaborative filtering - ✔✔Cold-start problem for both users and items,
missing values in the user-item matrix
✔✔Trend - ✔✔The long-term movement of a time series that represents the general
direction in which the data is moving over time
✔✔Seasonality - ✔✔The periodic fluctuations in a time series that occur at regular
intervals. For example, sales data may exhibit seasonality if sales increase during the
holiday season
✔✔Residuals - ✔✔Noise in a time series that cannot be explained by the trend or
seasonality
✔✔Seasonal Decomposition - ✔✔A technique used to separate a time series into its
trend, seasonal, and residual components
✔✔Additive Seasonal Decomposition - ✔✔The seasonal component is added to the
trend and residual components
✔✔Multiplicative Seasonal Decomposition - ✔✔The seasonal component is multiplied
by the trend and residual components
✔✔How to distinguish between additive and multiplicative seasonality? - ✔✔The
amplitude of the seasonal variation for additive seasonality is independent of the level,
while for multiplicative seasonality, it's connected
✔✔When is a time series stationary? - ✔✔If its statistical properties such as mean,
variance, and autocorrelation do not change over time
✔✔Augmented Dickey-Fuller Test (ADF) - ✔✔Statistical test that can be used to test for
stationarity
SOLUTION RATED A+
✔✔Content-Based Filtering - ✔✔Based on a description of the item and a profile of the
user's preferences, recommend items that are similar to those a user liked in the past
✔✔User Matrix - ✔✔A set of data points that represents all of a user's preferences
✔✔Item Matrix - ✔✔A set of data points that represent every item's content. Items can
be represented as a set of features
✔✔User-Item Similarity Matrix - ✔✔A set of data points that represents the similarity
between user preferences and items. These similarities can be weighted by previous
reviews or ratings from the user
✔✔Similarity metrics - ✔✔Similarity is commonly determined as the dot product
between two vectors, but different similarity metrics can be used, like Euclidean,
Manhattan, or Cosine Distance
✔✔What should you use to evaluate if content-based filtering is leading to increased
user engagement? - ✔✔A/B testing
✔✔Pros of content-based filtering - ✔✔Easy to implement, no need for data on other
users, can recommend niche items, can provide explanations for recommendations
✔✔Cons of content-based filtering - ✔✔Limited to the user's past preferences, limited to
the item's features, can overfit to the user's preferences, cold-start problem
✔✔Types of collaborative filtering - ✔✔User-user and item-item
✔✔User-User Collaborative Filtering - ✔✔Based on the idea that users who have
agreed in the past will likely agree in the future
✔✔Item-Item Collaborative Filtering - ✔✔Based on the idea that item reviews are often
grouped together. Different from content-based filtering because it has nothing to do
with the item characteristics. Ex: user 1 bought an orange, apple, and banana, user 2
bought an apple and orange, user 3 bought an apple --> recommend orange to user 3
✔✔User-Item Matrix - ✔✔A matrix that represents the relationship between users and
items. You can find the nearest neighbors from two perspectives: user-user (rows) or
item-item (columns)
✔✔Methods to fill missing values in the user-item matrix - ✔✔Use the scale's mean,
matrix factorization
, ✔✔Matrix Factorization - ✔✔A technique to break down a matrix into the product of
multiple matrices. It is used in collaborative filtering to estimate the missing values in the
user-item matrix
✔✔Alternating Least Squares - ✔✔An optimization algorithm used to minimize the error
between the predicted and actual ratings in the user-item matrix (when filled in with
Matrix Factorization)
✔✔Pros of collaborative filtering - ✔✔Can recommend items that the user has not seen
before, can recommend items that are popular among similar users
✔✔Cons of collaborative filtering - ✔✔Cold-start problem for both users and items,
missing values in the user-item matrix
✔✔Trend - ✔✔The long-term movement of a time series that represents the general
direction in which the data is moving over time
✔✔Seasonality - ✔✔The periodic fluctuations in a time series that occur at regular
intervals. For example, sales data may exhibit seasonality if sales increase during the
holiday season
✔✔Residuals - ✔✔Noise in a time series that cannot be explained by the trend or
seasonality
✔✔Seasonal Decomposition - ✔✔A technique used to separate a time series into its
trend, seasonal, and residual components
✔✔Additive Seasonal Decomposition - ✔✔The seasonal component is added to the
trend and residual components
✔✔Multiplicative Seasonal Decomposition - ✔✔The seasonal component is multiplied
by the trend and residual components
✔✔How to distinguish between additive and multiplicative seasonality? - ✔✔The
amplitude of the seasonal variation for additive seasonality is independent of the level,
while for multiplicative seasonality, it's connected
✔✔When is a time series stationary? - ✔✔If its statistical properties such as mean,
variance, and autocorrelation do not change over time
✔✔Augmented Dickey-Fuller Test (ADF) - ✔✔Statistical test that can be used to test for
stationarity
