When you think about recommendation systems, Netflix might come to mind first based on its ubiquity and power, writes bootcamp grad Molly Liebeskind in a blog post about her final project. But even while recommenders can be complex, Molly identifies two simple approaches (content-based filtering and collaborative filtering) that are good starting points to understanding how they work and to building one of your own.
Having lived in different parts of the world, Aisulu Omar realized the importance of addressing and navigating news media bubbles. She set out to create a news recommendation platform that would provide a balanced perspective for all.
For his capstone project, Nicholas Sherwin decided to continue working with natural language, film characterization and IBM Watson personality insights. By leveraging personality insights derived from a user's twitter profile, in conjunction with other elements, to craft personalized content based on a user’s personality profile.
As a California resident, Isaac Kim is familiar with the danger that earthquakes present. Having grown up near the San Andreas fault, he was taught at a young age to prepare for the inevitable "big one." Naturally, this inspired him to explore how modern computing power and machine learning methods can be employed to try and tackle earthquake prediction.
Did you know? 40 million used vehicles were sold in the US last year; Representing about 70% of the total vehicles sold. Given that a good portion of those sales already use online resources along various stages of purchasing, Bhanu Yerra thought a car image classification system could address several business issues for leading automotive sellers.
Using the Federal Contractor Misconduct Database, curated by the nonprofit Project on Government Oversight (POGO), Luke Persola set out to build a piece of machine learning software that could predict the outcome of a given case of contractor misconduct based on the other information in the database.
As a self-proclaimed beer enthusiast, Medford Xie routinely found himself looking for new brews to try – but he dreaded the possibility of disappointment once actually experiencing the first sips. This often led to purchase-paralysis. For his final project at Metis, he set out " to utilize machine learning and readily available data to create a beer recommendation engine that can curate a customized list of recommendations in milliseconds."
In this blog post, Natasha Borders walks you through how she built a streamer recommender for Twitch (live streaming platform for gamers), including details on the various tools used to make the resulting app, which is available now on Heroku.
After trying out a couple existing recipe recommendation apps, Jhonsen Djajamuliadi thought to himself, “Wouldn’t it be nice to use my phone to take photos of stuff in my refrigerator, and then get (personalized) recipes from them?” He decided to go for it, creating a photo-based recipe recommendation app for his final bootcamp project.
In an attempt to understand the market appeal of magazines, specifically women’s magazines, Nora G. May used data analytical tools to abstract magazines' marketing techniques. She extracted the text from magazine covers, performed NLP topic modeling, and used image processing techniques to understand graphic trends and representation.
For his second project at Metis, Dotun Opasina sought to predict NBA players’ salaries per season based on their statistics using Linear Regression. This project could be used by both individual players and managers to evaluate the impact a particular player is making on a team and to know whether to increase the players’ salary or trade the player.
Aaron Wilson Data Scientist, Strata Decision Technology
"When we hear about an airplane accident, there are a few things we always want to know. Where did it happen? Is everyone okay? And, of course: why did it crash?" writes Aaron Wilson. Find out how he used data science to find answers.
While Bitcoin showcases a vast body of anecdotal evidence that argues its utility, or potential future utility, due to the pseudonymous nature of the blockchain as well as other factors, it's still hard to come across aggregate data that shows this behavior occurring on a consistent and measurable scale. To attempt to address this problem, Ahlborg took a close look at trading volumes on the Peer-to-Peer Bitcoin trading website Localbitcoins.com.
At a recent DataKind SF event, Ash was rather intrigued by the challenges faced in investigating wage theft and other labor violations not just throughout the nation, but also specific to California and the Bay Area regions.
The opioid epidemic is one of the major public health catastrophes for this generation of Americans; similar to what tobacco/smoking or HIV/AIDS were to earlier generations, the opioid epidemic appears to be this era’s defining public health crisis. Lee set out to build a model to predict opioid-related mortality on a county by county basis with location-based insights and interventions in mind as a larger goal.
Imagine you’re hosting a birthday party. Everyone’s having a great time, music’s playing, and the party is noisy. Suddenly, it’s time for birthday cake! It’s too loud to use Alexa, and rather than hunting for your phone or a remote control, what if you could simply raise an open hand while in mid-conversation, your smart home device would recognize that gesture, and turn off the music? And with the same gesture, you could dim the lights just as the candles are lit?
Although uniqueness and personalization are great selling points, keeping up with the fashion trend is still the major theme that runs throughout the retail fashion business. The goal of this project is to find affordable alternatives to a designer outfit by using convolutional neural networks and other deep learning techniques.
Being a statistics-motivated sports fan, Frederick wanted to solve an atypical basketball problem: How can we optimize a typical basketball player’s career in the NBA? The question itself may seem open-ended, so in order to better scope this endeavor, he measured success by dollars earned.
Motorcycle Lean Assist uses a convolutional neural network to detect the lean angle of a motorcycle through image classification, providing you with rider feedback on your current lean angle so you don’t have to guess.
With cannabis being legal for medicinal use in 31 states and recreational use in 9 states, there are thousands of dispensaries from which one can obtain pricing data to analyze.
Davis thought it was a good time to dive into cannabis pricing to build a model that outputs a price benchmark for dispensaries (a “dankstimate” in the vein of Zillow’s “zestimate”).
Xuan Qi's goal for her project was to "accurately match customers with hotel inventory in this highly competitive market." On a personal note, she writes that "as a mom, when I book a hotel, I would like the hotel to be family friendly, closer to the sightseeing, and relatively quiet. But, my standards would be different booking a romantic weekend for me and my husband. We would like to pick the hotels with great food, closer to bars, and musical events are a plus. "
Vladimir Lazovskiy Data scientist working at the intersection of machine learning, content creation, and media.
In this project, Vladimir tackles the question: what is the relationship between machine learning and optimization? He explores how delivery companies can use the power of machine learning to forecast travel times between two locations and use the genetic algorithm to find the best travel itinerary for each delivery truck.
Alex chose to work with music data because it is a type of audio that can evoke emotion in addition to thought. When she listens to music, she asks herself, "Why does a particular song make me feel happy or sad?" The key of a song helps determine the feeling and is made up of the tonic note and the mode. For this project, she aimed to predict the mode.
In San Francisco, human waste is a growing issue, both for the people who run into it and for the people who have no other option than to relieve themselves on public streets. Mattie built a model that predicted where and when human waste will show up, which could be used to better inform resource allocation for programs like San Francisco’s Pitstop (a program that brings portable bathrooms to areas that have high homeless populations).
Tim's project explores the conversations about climate change that took place on Twitter in March 2017. With 1 million tweets from 560,000 users, Tim identified people belonging to different communities and used tools such as the Twitter API, Spark, NetworkX, and Gephi to derive insight from those conversations.
Vicky created a model dedicated to recommending pitches to the Cubs in games against the Cardinals. (Technically, this model could help any team – or any talented pitcher quite frankly – when throwing pitches against Cardinals players, but this model is specifically dedicated to her beloved Cubs.)
Aaron Wilson Data Scientist, Strata Decision Technology
Every week, the New Yorker magazine runs a caption contest. Aaron Wilson has entered this contest (unsuccessfully) dozens of times. "The problem is that I’m not very funny," he writes. "But computers? Computers are funny as hell. What if I could have one write captions for me?"
Alando Ballantyne Founder & Data Scientist, Sovereign Finance
Image analysis and classification is something that Alando is passionate about (specifically as it pertains to analyzing satellite imagery to generate economic data for emerging economies). In this post, he writes about a few of the more common pixel classification techniques used in remote sensing.
For his final project, Ankur decided to see if he could forecast hourly Uber demand across NYC neighborhoods. In addition to time-lagged features (such as previous week’s demand), he added information specific to each neighborhood to improve predictions.
Recommender systems are an effective key solution to overcome information overload. Oren wrote an article exploring the motivation behind recommendation systems, as well as providing an overview of different characteristics and potentials of various prediction techniques.
Orlando started this project to show the potential ethical conflicts created by our new algorithms. In every conceivable field, algorithms are being used to filter people. In many cases, the algorithms are obscure, unchallenged, and self-perpetuating.
Once again, the holiday season is upon us...Should you find yourself preparing the whole meal or offering to contribute a dish or two and in the mood for homemade culinary adventures, there’s a little web application, called the MenuPlannerHelper (abbreviated as MenuHelper) Heng-Ru May developed a while back that could come in handy.
Politicians have used gerrymandering, the practice of drawing political districts for partisan advantage, to skew elections since the early days of this great country...Joseph's goal was to build a tool that would let anyone optimize a map on whatever they think most important.
Rebekah's vision is to be able to attach a camera to the back of her bike, near the seat which captures video in real time and alerts of any cars that are approaching from behind. The alert would be an audio cue that is played in one of the apps that is already running -- Strava, Spotify, or Audible as examples.
To combine the functionality of individual song-based playlist generators with a focus on making content based recommendations, Zach created a web app that builds a hip-hop playlist of songs with similar lyrical meaning and mood around a song specified by the user.
Who among us hasn’t fallen victim to the addictive power of a binge-worthy Netflix show? For Jamie's final project at Metis, she chose to explore elements in popular shows that might lead you to start “binge watching” on Netflix.
I love basketball. I love playing it, watching it, or arguing scenarios with friends like who would win one on one, Kobe or Lebron. I had to combine my two passions, basketball and data science, in a machine learning project.
Naoya explores the intersection between data and art by designing a recurrent neural network utilizing Long Short-Term Memory nodes (LSTMs) to learn patterns in the Six Cello Suites by J.S. Bach and generate its own musical fragments.
To combat flaws in other reccomendation systems, Will decided to use natural language processing of beer reviews to find similarity of language used to describe beers. He found the words people use to describe beer give better results than arbitrary scores or styles.
The legality of and public’s view towards marijuana is rapidly changing as more states decriminalize and legalize the drug. As such, how have the words associated with marijuana in news articles changed over time?
You don't have to be an expert to know that password security is a big issue for companies these days. It seems every other week you hear of a well known website getting hacked. Hasan Haq's project uses neural networks to generate "dictionary" word lists to be used in password cracking.
MemoTrek is an application that takes your travel photos as input and makes personalized recommendations for future travel destinations. It provides two types of recommendations: you-may-also-likes for a similar type of experience and something-different for new adventures.
Micheal Lai Strategy Consultant & Data Scientist at IBM
Micheal created a system that can track players in a basketball clip and translate them to a coordinate grid. This kind of motion tracking already exists in the form of SportVU - but you can use the accessibility of YouTube clips to create player tracking.
Visaurant is a reimagination of the way users search through images that they are interested in. One prime use case for Visaurant is in sorting and filtering through food images (hence VIS -ual rest- AURANT).
The goal of Maresca's project was to perform semantic segmentation on satellite images in order to map out farmland around the city of Shanghai. He wanted to highlight a method that can be used to track farmland, urban development, and natural resources around the world in order to make better decisions for the future of our planet.
For her final bootcamp project, Alison Glazer looked into Airbnb's smart pricing tool, which was introduced years ago but faced immediate problems. According to Alison, the biggest issue was that price suggestions were too low and hosts noticed their revenues decrease when using the tool.
For bootcamp graduate Catherine Magsino, karaoke is more than just an occasional hobby.
"I’ve grown up singing karaoke for as long as I remember – at home, at family parties, and at get-togethers with friends. It is not only one of my favorite hobbies, but also a big part of my own culture. So naturally, I decided to focus my final project on this great pastime," she wrote in a blog post about her project.
When bootcamp graduate Sami Ahmed isn't working on data science projects, he's most likely making music and thinking about his interest in audio more generally. He recently found that NASA and the University of Iowa host a "massive library of electromagnetic waves that happen to fall in the audible human frequency range," he wrote in a blog post. Looking to get his hands on this audio, he figured out how to quickly web scrape hundreds of hours of audio from NASA using Beautiful Soup.
Bootcamp graduate Katherine Bell recently dove into the hiring practices of startups in the United States, analyzing whether or not hiring offshore talent makes economic sense. In order to do so, she looked at data within the 2019 Stack Overflow Developer Survey, which included around 40,000 non-US based respondents.
For her final project during the bootcamp, recent graduate Anupama Garla looked into answering the question: Living in Los Angeles, should I build a backyard home for extra income? She began to build a tool, which homeowners could use to determine the potential income and the feasibility of such a project. From Anupama's point of view, this tool would be beneficial to both homeowners and people like her, who are in the market to rent or own in the area. Using publicly available Airbnb and LA Geo datasets, she first built a model that predicted the income of a property.
To merge his interest in live music with data science, recent graduate Gabriel Bond chose to create a content-based live music recommender during the bootcamp.
"The result of this exploration utilizes unsupervised learning techniques, audio feature extraction with the LibROSA Python library, and both the Spotify and Songkick APIs to generate a playlist of songs by artists with upcoming shows in the user’s city based on the user’s favorite artists," he wrote in a blog post about the project.
For his final project at Metis, Michael trained a convolutional neural network auto-encoder to capture the essential visual features of artwork and developed a recommendation app that would compare a user-uploaded image of street art to a corpus of more than 35,000 images of fine art and then return those images of fine art, along with associated mediate, that were most similar.