1) The document discusses developing a recommender system for a bartender to suggest drink recommendations to customers based on their preferences and purchase history.
2) Several recommendation techniques are examined, including popular, collaborative filtering using purchase co-occurrence, and matrix factorization to make predictions even for new users or items.
3) Evaluating recommender systems requires balancing precision and recall metrics as the number of recommendations increases to avoid only recommending popular items while still covering preferred items.
An overview of some deep learning methods for recommender systems along with an intro to the relevant deep learning methods such as convolutional neural networks (CNN's), recurrent neural networks (RNN's), autoencoders, restricted boltzmann machines (RBM's) and more.
An overview of some deep learning methods for recommender systems along with an intro to the relevant deep learning methods such as convolutional neural networks (CNN's), recurrent neural networks (RNN's), autoencoders, restricted boltzmann machines (RBM's) and more.
A Steered-Response Power (SRP) based Framework for Sound Source Localization using Microphone Arrays in Reverberant Rooms for Enhancement of Speech Intelligibility
FPGA Based Acoustic Source Localization ProjectShristi Pradhan
Our Bachelor's final year project for active localization of a sound source using FPGA, microphones and stepper motor. Simulation was also accomplished on Simulink.
Toward Better Interactions in Recommender Systems: Cycling and Serpentining A...Qian Zhao
An experience design perspective on recommenders: There is a tradeoff between serving come-and-go users vs. encouraging deeper interaction/engagement!
Better understanding of the trade-off between efficiency vs. engagement can help design a better recommender user experience!
Cycling and serpentining top-N recommendation lists have benefits (higher engagement) but also costs (negative perception)!
More work combining algorithms and user experience is needed!
Past, present, and future of Recommender Systems: an industry perspectiveXavier Amatriain
Keynote for the ACM Intelligent User Interface conference in 2016 in Sonoma, CA. I start with the past by talking about the Recommender Problem, and the Netflix Prize. Then I go into the Present and the Future by talking about approaches that go beyond rating prediction and ranking and by finishing with some of the most important lessons learned over the years. Throughout my talk I put special emphasis on the relation between algorithms and the User Interface.
Recommenders Systems tutorial slides from the European Summer School of Information Retrieval (ESSIR).
Covers basic ideas on Collaborative Filtering, Content-based methods, Matrix Factorization, Restricted Boltzmann Machines, Ranking, Diversity.
The slides include material from Xavier Amatriain, Saul Vargas and Linas Baltrunas.
3. Secondary Data, Online Information Databases, and Measurement.docxtamicawaysmith
3. Secondary Data, Online Information Databases, and Measurement Scaling
1
Primary Scales of Measurement
7
3
8
Scale
Nominal Numbers
Assigned
to Runners
Ordinal Rank Order
of Winners
Interval Performance
Rating on a
0 to 10 Scale
Ratio Time to Finish, in
Seconds
Third
place
Second
place
First
place
Finish
Finish
8.2
9.1
9.6
15.2
14.1
13.4
Primary Scales of Measurement
Nominal Scale: The numbers serve only as labels or tags for identifying and classifying objects.
Ordinal Scale: A ranking scale
Interval Scale: Numerically equal distances on the scale represent equal values in the characteristic being measured.
Ratio Scale: Possesses all the properties of the nominal, ordinal, and interval scales. It has an absolute zero point.
Illustration of Scales of Measurement
Nominal Ordinal Ratio
Scale Scale Scale
Preference $ spent last No. Store Rankings 3 months
1. Parisian
2. Macy’s
3. Kmart
4. Kohl’s
5. J.C. Penney
6. Neiman Marcus
7. Marshalls
8. Saks Fifth Avenue
9. Sears
10.Wal-Mart
Interval
Scale
Preference Ratings
1-7
A Classification of Scaling Techniques
Comparative Scaling Techniques
Paired Comparison Scaling
A respondent is presented with two objects and asked to select one according to some criterion.
The data obtained are ordinal in nature.
Paired comparison scaling is the most widely-used comparative scaling technique.
With n brands, [n(n - 1) /2] paired comparisons are required.
Under the assumption of transitivity, it is possible to convert paired comparison data to a rank order.
Obtaining Shampoo Preferences
Using Paired Comparisons
Instructions: We are going to present you with ten pairs of shampoo brands. For each pair, please indicate which one of the two brands of shampoo you would prefer for personal use.
Recording Form:
aA 1 in a particular box means that the brand in that column was preferred over the brand in the corresponding row. A 0 means that the row brand was preferred over the column brand. bThe number of times a brand was preferred is obtained by summing the 1s in each column.
Paired Comparison Selling
The most common method of taste testing is paired comparison. The consumer is asked to sample two different products and select the one with the most appealing taste. The test is done in private and a minimum of 1,000 responses is considered an adequate sample. A blind taste test for a soft drink, where imagery, self-perception and brand reputation are very important factors in the consumer’s purchasing decision, may n ...
A Steered-Response Power (SRP) based Framework for Sound Source Localization using Microphone Arrays in Reverberant Rooms for Enhancement of Speech Intelligibility
FPGA Based Acoustic Source Localization ProjectShristi Pradhan
Our Bachelor's final year project for active localization of a sound source using FPGA, microphones and stepper motor. Simulation was also accomplished on Simulink.
Toward Better Interactions in Recommender Systems: Cycling and Serpentining A...Qian Zhao
An experience design perspective on recommenders: There is a tradeoff between serving come-and-go users vs. encouraging deeper interaction/engagement!
Better understanding of the trade-off between efficiency vs. engagement can help design a better recommender user experience!
Cycling and serpentining top-N recommendation lists have benefits (higher engagement) but also costs (negative perception)!
More work combining algorithms and user experience is needed!
Past, present, and future of Recommender Systems: an industry perspectiveXavier Amatriain
Keynote for the ACM Intelligent User Interface conference in 2016 in Sonoma, CA. I start with the past by talking about the Recommender Problem, and the Netflix Prize. Then I go into the Present and the Future by talking about approaches that go beyond rating prediction and ranking and by finishing with some of the most important lessons learned over the years. Throughout my talk I put special emphasis on the relation between algorithms and the User Interface.
Recommenders Systems tutorial slides from the European Summer School of Information Retrieval (ESSIR).
Covers basic ideas on Collaborative Filtering, Content-based methods, Matrix Factorization, Restricted Boltzmann Machines, Ranking, Diversity.
The slides include material from Xavier Amatriain, Saul Vargas and Linas Baltrunas.
3. Secondary Data, Online Information Databases, and Measurement.docxtamicawaysmith
3. Secondary Data, Online Information Databases, and Measurement Scaling
1
Primary Scales of Measurement
7
3
8
Scale
Nominal Numbers
Assigned
to Runners
Ordinal Rank Order
of Winners
Interval Performance
Rating on a
0 to 10 Scale
Ratio Time to Finish, in
Seconds
Third
place
Second
place
First
place
Finish
Finish
8.2
9.1
9.6
15.2
14.1
13.4
Primary Scales of Measurement
Nominal Scale: The numbers serve only as labels or tags for identifying and classifying objects.
Ordinal Scale: A ranking scale
Interval Scale: Numerically equal distances on the scale represent equal values in the characteristic being measured.
Ratio Scale: Possesses all the properties of the nominal, ordinal, and interval scales. It has an absolute zero point.
Illustration of Scales of Measurement
Nominal Ordinal Ratio
Scale Scale Scale
Preference $ spent last No. Store Rankings 3 months
1. Parisian
2. Macy’s
3. Kmart
4. Kohl’s
5. J.C. Penney
6. Neiman Marcus
7. Marshalls
8. Saks Fifth Avenue
9. Sears
10.Wal-Mart
Interval
Scale
Preference Ratings
1-7
A Classification of Scaling Techniques
Comparative Scaling Techniques
Paired Comparison Scaling
A respondent is presented with two objects and asked to select one according to some criterion.
The data obtained are ordinal in nature.
Paired comparison scaling is the most widely-used comparative scaling technique.
With n brands, [n(n - 1) /2] paired comparisons are required.
Under the assumption of transitivity, it is possible to convert paired comparison data to a rank order.
Obtaining Shampoo Preferences
Using Paired Comparisons
Instructions: We are going to present you with ten pairs of shampoo brands. For each pair, please indicate which one of the two brands of shampoo you would prefer for personal use.
Recording Form:
aA 1 in a particular box means that the brand in that column was preferred over the brand in the corresponding row. A 0 means that the row brand was preferred over the column brand. bThe number of times a brand was preferred is obtained by summing the 1s in each column.
Paired Comparison Selling
The most common method of taste testing is paired comparison. The consumer is asked to sample two different products and select the one with the most appealing taste. The test is done in private and a minimum of 1,000 responses is considered an adequate sample. A blind taste test for a soft drink, where imagery, self-perception and brand reputation are very important factors in the consumer’s purchasing decision, may n ...
From Metrics to Models: Data Science at MetailMatt McDonnell
Metail is a leader in fashion technology that allows users to try on clothes online. This is accomplished by users creating a realistic personalized MeModel from a small set of common measurements and using this to try on digitized garments. This talk given for the Cambridge Data Insights Meetup in December 2017 covers the evolution of Data Science at Metail from measuring performance of new product features to creating models that are data products in their own right.
Slides of my talk I gave at the big data conference: http://www.globalbigdataconference.com/santa-clara/5th-annual-global-big-data-conference/schedule-85.html
Bing Ads' Eric Couch dives in to beginning and advanced Excel tips and tricks for PPC marketers- including data analysis tips, Excel formulas, and incredibly handy plugins.
Big & Personal: the data and the models behind Netflix recommendations by Xa...BigMine
Since the Netflix $1 million Prize, announced in 2006, our company has been known for having personalization at the core of our product. Even at that point in time, the dataset that we released was considered “large”, and we stirred innovation in the (Big) Data Mining research field. Our current product offering is now focused around instant video streaming, and our data is now many orders of magnitude larger. Not only do we have many more users in many more countries, but we also receive many more streams of data. Besides the ratings, we now also use information such as what our members play, browse, or search.
In this talk I will discuss the different approaches we follow to deal with these large streams of data in order to extract information for personalizing our service. I will describe some of the machine learning models used, as well as the architectures that allow us to combine complex offline batch processes with real-time data streams.
MKT 335 Client Packet for Milestones One to Three and Final IlonaThornburg83
MKT 335 Client Packet for Milestones One to Three and Final Project
Client Name: KitchenAid
Part 1: Client Overview
Your client for the final project is KitchenAid, a United States–based appliance brand known for
leveraging best-in-class technology to create state-of-the-art products that solve novice chefs’ toughest
kitchen challenges.
A key component of KitchenAid’s rollout strategy is their countertop appliance offerings, including
tabletop ovens that eliminate the need for a big oven, blenders, juicers, and food processors that peel,
dice, and chop. KitchenAid’s goal is to make the time consumers spend in the kitchen efficient, fun, and
“simply brilliant.”
A core component of the business and KitchenAid’s foremost marketing priority is its product launches.
Once every four to six weeks, KitchenAid introduces a new product to the market. The products are each
introduced with the same amount of attention, care, and of course, a big digital advertising spend.
As part of your work for this client, your job will be to research the category, better understand the
brand’s consumer, and propose a digital campaign strategy that helps ensure the client’s next launch is
one of its most impactful.
Before continuing, take some time to research KitchenAid and their current campaign strategy, paying
special attention to their website user experience and their social media channels. You may even want
to sign up for their emails. Once you have a solid understanding of the current online presence, move on
to Part 2.
Part 2: Campaign Information
Previous Campaign/Current Campaign Structure
Typically, the KitchenAid brand spends their advertising dollars on holistic campaigns that sell the brand
as a whole, as much as they do the product. This year, however, they want to change that structure,
creating campaigns that cater to specific products.
Their main reason for changing their tried and true advertising strategy is that they would like to be able
to leverage the unique targeting capabilities of digital to target users with products specific to their
needs.
Target Persona
The brand has three distinct target personas they leverage to sell products and innovate in the kitchen
category.
Health Nuts: Health lovers of all ages and stages who know that cooking at home is the best way
to control their food choices. They leverage social media to find new recipes and ideas and
show-off their healthy lifestyles. KitchenAid’s countertop products keep them cooking without
creating additional messes and steps that require long amounts of cleanup.
Terrell McGhee
Highlight
The Tasty Chef: Usually females between the ages of 25 and 45, the tasty chef is frequently
inspired by the elaborate creations they see online. They view these videos as entertainment
but do not mind trying them out from time to time. They love KitchenAid’s countertop products
because they are versatile and give them the option ...
During my summer internship at Anheuser-Busch, I worked with the Category Management and Solutions department to develop a predictive model for SKU unit movement. I utilized machine learning techniques to process 100+ variables including total facings, price, capacity, as well as demographics variables per store ZIP code.
14. DRINK RECOMMENDER
14
Story:
After spending 8 years studying engineering in college, I decided to become a bartender.
I kept notes of what my customers liked and disliked.
I want to create a recommender system to suggest and create new drinks that my
customer will most likely buy.
15. GETTING TO KNOW MY MAKE-BELIEVE
CUSTOMER
ANALYZING DATA
15
?
?
?
?
?
?
?
?
?
?
?
?
?
User
info
itemsUser
16. Miranda:
Likes more dessert-y type drinks. Not into bitter drinks. Can’t tell if Miranda is rich.
Sansa:
She likes Martini and hates everything else? Don’t know much about her.
Cersie:
Only into strong coffee drinks. She likes alcoholic drinks and she’s rich.
? ?
?
??
?
16
17. I CAN SPEND HOURS, DAYS, MONTHS, YEARS AND PERHAPS DECADES
ANALYZING THE DATA MANUALLY.
I NEED TO FIND A WAY TO FACTOR INFORMATION ABOUT MY
CUSTOMERS, DRINKS AND INTERACTION BETWEEN THEM TO
AUTOMATE THE RECOMMENDING PROCESS USING A COMPUTER.
17
INFORMATION OVERLOAD
I should’ve taken that
machine learning course...
18. PERSONALIZATION
18
Connecting user to items
User | Movies
User | Products
User | Music
Is browsing every movie, product, or music even
practical?
“Need new ways to discover content”.
19. PERSONALIZATION
19
● User to Drinks?
● User to User?
● User to medicine?
Connecting user to items
User | Movies
User | Products
User | Music
20. PERSONALIZING DRINKS
20
Bartender's note:
Enio ordered a Martini last weekend in the evening.
He ordered a coffee on Monday at noon.
A good recommender system adapt with time and is capable of
considering multiple sessions.
21. POPULAR RECOMMENDER (LEVEL 0)
21
Count:
Item:
Popular recommender, recommends the
most popular item on the menu, Espresso.
● Completely lacks personalization. My
customers have different taste and they
want drinks that match their interests.
22. CLASSIFICATION MODEL (LEVEL 1)
22
User info
Item info
Past history
Others
Classification
model
Yes, Miranda will probably like Mocha.
No, Miranda will probably dislike Mocha.
Pros : Personalized. Capture context.
Cons: We don’t have access to all these info and if we input wrong information, we
will make wrong predictions.
23. COLLABORATIVE FILTERING (LEVEL 2)
23
People who liked X also liked Y.
If Ramin likes X, he might also like … Y.
Co-occurrence Matrix (A symmetrix item-item matrix)
Item Espresso Martini Mocha
Espresso a 3 2
Martini 3 b 7
Mocha 2 7 c
Number of people
purchase both Mocha
and Espresso
Jose bought Mocha the other day, what should I recommend him to now?
Martini
24. COLLABORATIVE FILTERING (LEVEL 2)
24
Item Espresso Martini Mocha Coffee
Espresso a 3 2 45
Martini 3 b 7 23
Mocha 2 7 c 39
Coffee 45 23 39 d
Popular Item Effect
No matter what Jose has purchased, the recommender system will recommend coffee.
There are ways to normalize the data to avoid the popular item effect.
25. COLLABORATIVE FILTERING (LEVEL 2)
25
Item Espresso Martini Mocha Coffee
Espresso a 3 2 45
Martini 3 b 7 23
Mocha 2 7 c 39
Coffee 45 23 39 d
No History
I’m only looking at Jose most current purchase of Mocha. What if he bought a Martini before
and he didn’t like it?
26. COLLABORATIVE FILTERING (LEVEL 2)
26
Item Espresso Martini Mocha Coffee
Espresso a/N 3/N 2/N 45/N
Martini 3/N b/N 7/N 23/N
Mocha 2/N 7/N c/N 39/N
Coffee 45/N 23/N 39/N d/N
Weighted Average of purchased items
I want to know if I should recommend coffee to Jose.
Score_coffee = ⅓(S_coffee,mocha + S_coffee,martini + S_coffee,espresso)
Sort the scores and pick the one with highest score.
N: Normalizing factor
Purchase History
Jose:
Mocha - YES
Martini - NO
27. COLLABORATIVE FILTERING (LEVEL 2)
27
More problems!!!
● Can’t utilize context like time of the day.
● Can’t utilize my customer age to my advantage.
● Can’t utilize information about my drinks and
their ingredients to make better recommendation.
And WHAT IF:
I have a new customer, what should I recommend?
What if I’m making a new drink, who would buy it?
A Cold Start Problem.
29. MATRIX FACTORIZATION (LEVEL 3)
29
● We need a recommender system that factors more than just users past purchase history.
● A system that can factor more personalized info about the user and the product and the time
of the visit into account as well plus all the goodies we get from the collaborative filtering.
To learn from data even when not available. (Very sparse matrices, missing data.)
Keep in mind that each user only tries a
few drinks.
Rating =
Users
ItemsRating available from
user U for item V
Rating unavailable
We don’t know what the
user thinks about item V
30. MATRIX FACTORIZATION (LEVEL 3)
30
Rating =
Users
Items
We need to fill in the white boxes using
ALL of the available info we got.
Dictionary
(Bases)
Activation Matrix
(Encodings)
x
~
A whole lot of fancy math
happens here to factorize the
rating matrix as multiplication
of two other smaller matrices
that uncover those hidden
area by minimizing some cost
function.
31. BLENDING MODELS (LEVEL 4)
31
Point : There is no universal recommender systems that work for everything.
We need to blend different models to be able to attack different applications.
Netflix Challenge
From 100 million movie ratings
rate 3 millions of them to highest
accuracy.
Winning team blended over
100 models to gain 10.35%
improvement in the accuracy.
(and got 1 million dollars for it!)
32. PERFORMANCE METRIC
RECOMMENDER SYSTEMS
32
RMS
● Fraction of items correctly recommended.
BUT:
● We care about what the user liked more than what they don’t.
● Imbalanced information can skew the results.
● With this metric you can get good accuracies by recommending nothing at all!
Recall
(# liked & # shown) / # liked
● How many of the items that the user liked, was actually recommended?
The world we’re looking at only contains the liked items.
Precision
(# liked & # shown) / # shown
● Out of the recommended items, how many items did the user actually
liked? The world is all the recommended items. How much garbage should
I look at until I found what I like (attention span).
33. OPTIMAL RECOMMENDER
33
Recall
(# liked & # shown) / # liked
Maximize Recall => Recommend everything!
Recall = 1 But, Precision = very small
Precision
(# liked & # shown) / # shown
Best Recommender would
recommend only the products the user like.
Precision = 1
Recall = 1
Point: Use both precision and recall among other metrics RMS to
evaluate your recommender system.
34. PRECISION - RECALL CURVE
VARY #ITEMS
34
Optimal Recommender:
● Precision stays at 1, because it’s
only recommending what I like.
● Recall increases, because it’s
uncovering more of the items that
the user like as we are increasing
the threshold on the number of
items.
One liked item
recommended.
1 2 3 4 …./ total number of interests
Recall
Precision
All liked item are
recommended.
35. PRECISION - RECALL CURVE
35
A more realistic recommender:
As we recommend more items to the user,
the area below the precision-recall curve
drops, we start introducing garbage!
1 2 3 4 …./ total number of interests
Recall
Precision
Realistic
recommender
(smoothed out)
36. PRECISION - RECALL CURVE
36
A more realistic recommender:
As we recommend more items to the user,
the area below the precision-recall curve
drops, we start introducing garbage!
Based on the application, you could also
look at some weighted average of these
metrics as well. For now, we can look at the
area below the curve to choose which
recommender system to choose.
1 2 3 4 …./ total number of interests
Recall
Precision
Orange curve is a better
recommender system
than the red one.
37. DRINK RECOMMENDER - AN EVALUATION
37
Matrix Factorization and
Similarity based
Recommender (with
item info+ with/out user
info)
Similarity based
recommender
Matrix factorization
recommender
39. BACK TO OUR AARP SERVICE RECOMMENDER
SYSTEM
39
Data and Training:
● 100 participants.
● 46 questions
● Imbalanced demographic classes
● Missing data less than 1%
● Training with 80% of the data
● Tuning with cross validation
● Testing with 20% of the data
User info -
Salary User info -
Age
Item info-
Outdoors