Benson Duong
Data Science major at UC San Diego
Interests in Data Science, Python Programming, Machine Learning.
Sub-Project 2: Predicting if a User will visit a Restaurant
- This 2nd task on users and restaurants determines how likely a given user would dine at a given restaurant, based on the user’s history of previosuly chosen restaurants (if available) and the restaurant’s history of customers. In other words, this uses a methodology in recommendation machine learning called collaborative filtering, which predicts how a customer will rate a product based on how that product was rated by similar customers.
- Unlike the previous task, this one will employ more advanced deep learning and pre-trained ML tools.
Table of Contents
Overview
- The model neural network will take a given user and a given restaurant as an input pair, and output a value between 100% and 0% confidence on if the user will visit and eat at that restaurant.
- Each user is represented by the set of all foods (or rather, food-related words) that appear in the reviews they’ve written for different restauarants.
- Likewise, each restaurant is represented by the set of all foods (or rather, food-related words) that appear in the reviews written about them.
- This graph visually summarizes the model architecture. As one can see, 80% of my time is actually for the data preparation and cleaning, while the prediction model itself is merely a 20% part at the end.
Text-Data-Processing
Word2Vec
- In order to approximately select only the food-related words, I used a completely separate dataset: a recipe dataset where one of the columns was a list of ingredients for each recipe; while it’s impossible to get a universal list of every food-related word that exists, this is sufficient enough to build a vocabulary of just food words.
- A reason behind my choice to use Word2Vec is its ability to connect relationships between words that usually appear close to one another. This is useful because it allows indirect associations between 2 given words.
- Let’s say there is a user who mostly eat seafood, but the word “seafood” never shows in their reviews. Despite being so, they can still be associated with seafood based on the seafood-related words in their reviews (“salmon”, “crab”, etc.), because there are many other sentences in the dataset that do in fact have the word “seafood” and the said seafood-related words showing up close to each other, which is how the association is learned from.
-
So now, a column where each row is the ingredients list for a recipe, is an effective learning place for the Word2Vec to bridges associations between food words based on their co-occurences.
- One of Word2Vec’s famous features is the ability to do “arithmetic” with words. Think of it like the analogy questions of old SAT exams. For the custom food dataset vocabulary that we trained Word2Vec on, we expect things like “barbeque - july 4th + thanksgiving”, which basically asks “If July 4th is to barbeque, then Thanksgiving is to …?).”. The top leading words are related to turkey.
- For exploratory purposes, it’s also nice to check the word that “represents” a restaurant.
- Sometimes, that representative word is very obvious, being a very frequently appearing word in the reviews (like pizza, for this restaurant)
- Other times, it is a subtle word that might not even have appeared in the reviews at all. For this restaurant with only 1 review, it was deemed “mediterranean”, even though the word “mediterranean” is nowhere to be found. Only 1 word may have been all that was needed to tip it to that label - “Gyro” (a Greek dish).
Image-Data-Processing
Image-Recognition
- Some of the reviews comes with images of the food submitted by the users; this could be fed through a “convolutional neural network”, an image recognition model which is a tool that can label the objects seen in a given picture. These labels are then filtered to only get food-related words using the food vocabulary as made by the recipe dataset. These words are attached to the text review as though it was part of the review.
- For example, if an image in a review is a picture of a cupcake on a plate on a table by a window, the VGGNET16 model might output the top predictions as “cupcake”, “window”, “table”, “plate”. Assuming that the food vocabulary I created is large enough to include the word “cupcake”, then that would be the only word that will be used from the image.
- The python code accomodates this procedure, but I switched off for the baseline model since it’s computationally costly.
Segmentation of Restaurants and Users
Using Unsupervised Machine Learning to automate the segmentation of Restaurants and Users into Distinct Sub-groups
-
It is possible to use Unsupervised Machine learning to automate segmentation of Restaurants and Users into distinct sub-groups; it does so by finding patterns in terms of commonalities between food vocabularies of different reviews. Python may not intrinsically understand what the human idea of “seafood” is, it always sees that words like “tuna”, or “crab”, tend to associate more by being seen in reviews of the same restaurants.
-
With further processing of the Word2Vec tool, each restaurant and each user is representable by a point in 2D space. Each point in space on the left scatterplot is a restaurant, while on the right are users.
- Furthermore, it is possible to group up the points using a statistical clustering method. I separated the restaurants into distinct sub-groups based on the textual content of the reviews written about it, and did the same for users.
- While this is not necessary, it would be insightful to see if there’s any actual meaningfulness in the sub-groups that was automatically created by that statistical clustering method, for restaurants and users. To do so, I extracted the “most distinctive” keywords” of each sub-group.
- The overarching direction of my project is to connect user sub-groups to restaurant sub-groups. For example, restaurant-group 4 tends to make desserts (sweets, confectionery, pastries), and user-group 0 tends to eat mostly desserts.
- In other words, I am grouping up restaurants by the food they serve, and grouping up the restaurant-goers by the food they review, and hoping that a one-to-one correspondence (if it even exists) becomes visible between restaurant-types and customer-types in terms of food. This would allow me to recommend the dessert-lovers (as seen above in user-group 0) to the dessert restaurants (as seen above in restaurant-group 4) …and recommend Italian food lovers to Italian restaurants, seafood-lovers to seafood restaurants, and so on.
- To see more examples of these restaurant/user “sub-groups” and their “most distinctive keywords”, scroll to the bottom half of the visualization notebook in the github repo
Evaluation
- Finally, the last part is to create a prediction model that answers whether or not a given user is likely to interact with a given restaurant… with the underlying data being based on the types of food that the given restaurant serves and the types of food the user has eaten at other restaurants.
-
For this interaction problem, it is necessary to manually create “artificial, unseen” pairs; since every user-restaurant pair in the dataset is real and observed, there is no “nonexistent pair” samples to compare the observed pairs against. A sampling procedure was developed that tries to find “negative” (a.k.a. never-before-seen) combinations of users and restaurant. This of course brings into consideration an unspoken assumption that the given dataset is an extremely accurate representation of real life; in other words, I am assuming there are no cases where the artificial pairs that I created actually did exist in real life, but wasn’t recorded in the dataset.
-
The final input that the prediction model will train on is a large set of user-restaurant pairs, equally divided between “real, observed” pairs and “artificial, unseen” pairs.
- After training the prediction model, the metrics are summarized below: it is around 73% accurate in correctly determining whether or not a given user has visited a given restaurant.
- Specifically, it is able to correctly label 73% of actually existing user-restaurant pairs as real, and 68% of invented pairs as fake.
Conclusion
- Overall, this model does not yet have a procedure for never-before-seen users and restaurants.
- This model also must acknowledge several considerations:
- The dataset used is only limited to the customers who willingly and manually wrote a review for a restaurant. This model cannot be extended to the general population of more casual resaurant-goers. Thus, this model’s goal is more so whether a reviewer will review a restaurant or not.
- This dataset lacks the geo-location of the restaurants, which might be essential to this specific prediction task. Realistically, people will be unlikely to eat at a given restaurant if it is far away from their house, and logically can only eat at nearby restaurants, even if the far-away restaurant aligns better with their food preferences than their local ones. This is a glaring inadequacy of this dataset (and in turn, the tasks that we can do with it).
- However, the data analysis done along the way, still had useful results:
- I made a trained a Word2Vec model for food.
- I found a way to divide restaurants and customers into distinct sub-groups simply by the types of food they were eating, as seen in the scatterplot of colored groups. This type of information can still be useful as a base data in any future “spinoff” projects.