Explains about the different types of Recommender systems and the advantages and disadvantages of both. There are examples in the presentation which showcase how the recommender systems recommend to the user. Further it explains how they are used by E-commerce platforms to influence user choice.
2. Contents:
• Introduction
• What are Recommender Systems ?
• How Recommender Systems work ?
• Approaches used
• Advantages
• Disadvantages
• Ways of recommending
• Evaluation methods of Recommender Systems
• How to make such a system ?
• Recommender System in action.
4. In today’s world we are surrounded by choices.
• Which film to watch??
• Which mobile to buy??
• Which book to read??
All these options confuse us…
5. Example:-
Choosing which mobile to buy???
Google PixelSamsungLG Moto G
5” Screen
3 GB RAM
32 GB Internal Memory
16 MP Camera
5” Screen
3 GB RAM
32 GB Internal Memory
16 MP Camera
5” Screen
3 GB RAM
32 GB Internal Memory
16 MP Camera
6. Another example…
Suppose I want to hear a good song. Then how
should I find that song online ??
Also what does “good” mean??
A “good” song can be a romantic song or a
classical song or anything else…
7. Can Google help me ??
Yes, but only if I know the name of song.
But I just want to hear a “good” song…
8. Can Facebook help me ??
Yes, I can see which songs my friends like and
listen to them.
What if I do not like the songs that my friends
like ??
9. Can Experts help me ??
Yes, expert opinions can be very useful.
But their opinions are what they like and not
what I like.
10. So how am I going to find a good song ??
The solution is
Recommender Systems
How ????
11. What are Recommender Systems ???
Recommender systems are a subclass of
“information filtering system” that seek to
predict the “rating” or “preference” that a user
would give to an item.
12. How Recommender Systems work ???
They work in following two ways:
1. Collaborative filtering
2. Content-based filtering
20. Disadvantages:
• Cold start problem: Here the past data of
items are required for recommendation.
• Requires extensive computation for
recommendation
• Example: Colgate may have many different
types of toothbrush as well as toothpaste. So
which one to recommend to whom ???
23. Advantages:
• Doesn’t depend upon past user data. (Because
it can analyze data of similar users to
recommend)
• Since we have analyzed data, extensive
computing is not required.
33. Conclusion:
• Recommender systems are very useful for
users in decision making.
• But we should be careful that these systems
can also be configured to influence bad
recommendations.
Editor's Notes
Amazon does not use Collaborative. Here it is given just for understanding purpose.
Example: Buying Samsung phone and getting earphones as recommendation.
Collaborative: Recommend top phone (LG) vs Best phone (Samsung)
Cold start: New users and items do not have historical data.
Tries to recommend items similar to user’s past likes.
Properties of items are considered to recommend. Every new user gets same recommendation.
Cold start: New items that do not have any properties will not be considered for recommendation.
Mahout: Java. Project of Apache, implements machine learning algorithm based on information filtering.
Shogun: Both
Weka: Java
The core of Shogun is written in C++ and offers interfaces for MATLAB, Octave, Python, R, Java, Lua, Ruby and C#.
Here you have to demonstrate a sample recommender system in action.