Machine Learning EDX Chapter 19 – Flashcards
Unlock all answers in this set
Unlock answersquestion
a commonly used type of machine learning solution that matches users to items
answer
Recommendation models, or recommenders,
question
a more common approach for Recommenders
answer
is to use a filter-based recommender that uses matrix factorization, not regression ,classification and clustering
question
matrix factorization.
answer
This is a technique in which known ratings given by users to items are used to determine likely ratings that are not present in the matrix.
question
Going across the row in matrix factorization
answer
User based recommendations
question
Going down the columns in matrix factorization
answer
Item based recommendations
question
Creating hierarchy of values and matrix factorization
answer
feature recommendations
question
Collaborative filtering is also called
answer
Matrix factorization
question
Cold start
answer
Not enoough data user similar users
question
Metrics forevaluation of recommendation systems
answer
Cumulative gain and Discounted CG and Normalized DG and For Numberical ratings (RMSE), (MAE)
question
Cumulative gain formula
answer
Sum of relevance
question
Discounted CG
answer
Discounted by position
question
Manhattan Distance
answer
The distance between two points measured along axes at right angles. x 1 minus x 2 + y 1 minus y 2
question
Euclidean Distance
answer
Most commonly used measure of the similarity between two objects. Essentially, it is a measure of the length of a straight line drawn between two objects when repre- sented graphically.
question
RSME
answer
Root Squared Mean Error
question
MAE
answer
Mean absolute error
question
Normalized DCG
answer
DCG/IDCG
question
IDCG
answer
Idealized Discounted Cumulative Gain
question
Use ? for prepare data for a recommender
answer
In Azure ML, you should use the Recommender Split mode of the Split module
question
After splitting the recommender data, you can use a
answer
Train Matchbox Recommender module to train a recommender
question
After training the recommender, you can use the
answer
Score Matchbox Recommender to generate predictions
question
Types of Score Recommender prediction
answer
Item Recommendations Related Items Rating Prediction Related Users
question
Item Recommendation
answer
Predicts recommended items based on a given user.
question
Related Items:
answer
Predicts recommended items based on a given item.
question
Rating Prediction
answer
Predicts ratings for given users and items.
question
Related Users:
answer
Predicts users based on a given user.
question
to evaluate recommender performance
answer
Evaluate Recommender module
question
Ways to evaluate recommender with module
answer
Normalized Discounted Cumulative Gain (NDCG) for item recommendation, related items, and related users. Mean Absolute Error (MAE) and Root Mean Squared Error (RSME) for rating prediction
question
NDCG
answer
Normalized Discounted Cumulative Gain
question
NDCG is used for
answer
item recommendation, related items, and related users
question
Rating prediction uses
answer
MAE and RSME
