Recommendation Engines perform a variety of tasks - but the most important one is to find products that are most relevant to the user. Content based filtering, collaborative filtering and Association rules are common approaches to do so.
So let's first Understand basics of Recommendation Engines and then we'll later on Build Our Own Recommendation Engine !!!
HIGH QUALITY, PERSONALIZED ARE THE HOLY GRAIL FOR EVERY ONLINE STORE.
UNLIKE OFFLINE STORES, ONLINE STORES HAVE NO SALES PEOPLE.
USERS ON THE OTHER HAND HAVE LIMITED TIME AND PATIENCE, ARE NOT SURE WHAT THEY ARE LOOKING FOR ONLINE STORES HAVE A HUGE NUMBER OF PRODUCTS.
RECOMMENDATIONS HELP USERS NAVIGATE THE MAZE OF ONLINE STORES FIND WHAT THEY ARE LOOKING FOR FIND THINGS THEY MIGHT LIKE, BUT DIDN’T KNOW OF.
RECOMMENDATIONS HELP ONLINE STORES SOLVE THE PROBLEM OF DISCOVERY.
BUT HOW? Lets Explain this.
ONLINE STORES HAVE DATA
1) WHAT USERS BOUGHT
2) WHAT USERS BROWSED
3) WHAT USERS CLICKED
4) WHAT USERS RATED
All these DATA IS FED into Recommendation Engine to derive below:
TOP PICKS FOR YOU!!
IF YOU LIKE THIS, YOU’LL LOVE THAT!
IF YOU BUY THIS, YOU’LL NEED THAT!
Tasks Performed by Recommendation Engine
Filter Relevant Products
Sub Tasks: 1.1) PREDICT WHAT RATING THE USER WOULD GIVE A PRODUCT
1.2) PREDICT WHETHER A USER WOULD BUY A PRODUCT
1.3) RANK PRODUCTS BASED ON THEIR RELEVANCE TO THE USER
HOW TO FIND RELEVANT PRODUCTS
1) “SIMILAR” TO THE ONES THE USER “LIKED” (Content Based Filtering)
2) “LIKED”BY “SIMILAR” USERS (Collaborative Filtering)
3) PURCHASED ALONG WITH THE ONES THE USER “LIKED” (Association Rules)
RECOMMENDATION ENGINES NORMALLY USE ONE OR MORE OF THESE TECHNIQUES.
BEFORE WE GO AHEAD, LET’S JUST CLARIFY -
HOW DO WE KNOW THAT A USER “LIKES” A PRODUCT?
Ans) THESE ARE PRODUCTS THE USER
PURCHASED
CLICKED ON
ADDED TO CART
RATED HIGHLY
(IF THE USER RATES SOMETHING LOW, THAT’S IMPORTANT DATA TOO!)
SOMETIMES, STORES ALSO ASK FOR EXPLICIT INPUT FROM USERS.
So let's first Understand basics of Recommendation Engines and then we'll later on Build Our Own Recommendation Engine !!!
HIGH QUALITY, PERSONALIZED ARE THE HOLY GRAIL FOR EVERY ONLINE STORE.
UNLIKE OFFLINE STORES, ONLINE STORES HAVE NO SALES PEOPLE.
USERS ON THE OTHER HAND HAVE LIMITED TIME AND PATIENCE, ARE NOT SURE WHAT THEY ARE LOOKING FOR ONLINE STORES HAVE A HUGE NUMBER OF PRODUCTS.
RECOMMENDATIONS HELP USERS NAVIGATE THE MAZE OF ONLINE STORES FIND WHAT THEY ARE LOOKING FOR FIND THINGS THEY MIGHT LIKE, BUT DIDN’T KNOW OF.
RECOMMENDATIONS HELP ONLINE STORES SOLVE THE PROBLEM OF DISCOVERY.
BUT HOW? Lets Explain this.
ONLINE STORES HAVE DATA
1) WHAT USERS BOUGHT
2) WHAT USERS BROWSED
3) WHAT USERS CLICKED
4) WHAT USERS RATED
All these DATA IS FED into Recommendation Engine to derive below:
TOP PICKS FOR YOU!!
IF YOU LIKE THIS, YOU’LL LOVE THAT!
IF YOU BUY THIS, YOU’LL NEED THAT!
Tasks Performed by Recommendation Engine
Filter Relevant Products
Sub Tasks: 1.1) PREDICT WHAT RATING THE USER WOULD GIVE A PRODUCT
1.2) PREDICT WHETHER A USER WOULD BUY A PRODUCT
1.3) RANK PRODUCTS BASED ON THEIR RELEVANCE TO THE USER
HOW TO FIND RELEVANT PRODUCTS
1) “SIMILAR” TO THE ONES THE USER “LIKED” (Content Based Filtering)
2) “LIKED”BY “SIMILAR” USERS (Collaborative Filtering)
3) PURCHASED ALONG WITH THE ONES THE USER “LIKED” (Association Rules)
RECOMMENDATION ENGINES NORMALLY USE ONE OR MORE OF THESE TECHNIQUES.
BEFORE WE GO AHEAD, LET’S JUST CLARIFY -
HOW DO WE KNOW THAT A USER “LIKES” A PRODUCT?
Ans) THESE ARE PRODUCTS THE USER
PURCHASED
CLICKED ON
ADDED TO CART
RATED HIGHLY
(IF THE USER RATES SOMETHING LOW, THAT’S IMPORTANT DATA TOO!)
SOMETIMES, STORES ALSO ASK FOR EXPLICIT INPUT FROM USERS.
Comments
Post a Comment