This post features two projects from recent graduates of our data science bootcamp. Take a look at what's possible to create in just 12 weeks.
Earthquake Prediction with Machine Learning
Isaac Kim, Metis SF Bootcamp Student
Metis bootcamp graduate Isaac Kim is a California resident living near the San Andreas Fault, making him no stranger to earthquake-related concerns, including worries of when "the big one" might hit. For his final bootcamp project, Isaac wanted to examine the possibilities of using Machine Learning to more accurately predict when earthquakes will occur. If that's possible, people could take preventative action, potentially saving lives and money in infrastructure repairs, he notes. Up to now, progress on being able to make these types of predictions hasn't evolved as much as expected.
"With the advent of machine learning in the 80s, seismologists were optimistic that we may soon solve this problem once and for all. However, over the past few decades, seismologists have learned that this is a far more complex problem than anticipated, and some dubious predictions made using shoddy science have had dire consequences (google the town of L’Aquila, or read Nate Silver’s excellent book The Signal and the Noise)," wrote Isaac in a blog post about his project, Earthquake Prediction with Machine Learning.
Isaac goes on to explain that instead of aiming to predict earthquakes days or even hours before they strike, knowing just a few seconds in advance can make a huge difference. In fact, many modern power systems already have failsafes in place that try to mitigate earthquake damage based on just seconds of warning.
In describing how this works, he writes: "Earthquake waves are typically composed of different types: p-waves (compressional waves) travel faster and do less damage, whereas s-waves (translational waves) will land a few seconds later but will do the bulk of the damage. In short, these systems attempt to detect p-waves and shut important equipment down before the s-waves arrive."
For this project, he wanted to dive into using modern machine learning methods to improve on such systems. The Los Alamos National Laboratory made its research data public and hosted a competition on Kaggle so that data scientists everywhere could take a crack at it. While the competition had already closed by the time Isaace found out about it, he decided to give it a go regardless.
In a post on Towards Data Science, he goes into detail about his process and outcomes. Read it here.
Visualizing the Personality Profile of Any Film Character Using Python & IBM Watson
Nicholas Sherwin, Metis SF Bootcamp Student
In the opening paragraph of a recent post on Towards Data Science, Metis bootcamp graduate Nicholas Sherwin writes: "The great social psychologist James Pennebaker once said, 'By looking more carefully at the ways people convey their thoughts in language we can begin to get a sense of their personalities, emotions, and connections with others.' He, alongside many other psychologists, linguists, and natural language processing (NLP) practitioners have made great strides in extrapolating detailed (and eerily accurate) personality information from written text using advanced techniques such as bidirectional LSTMs and NLU (Natural Language Understanding). More recently, the pioneering team behind IBM Watson developed a product called Personality Insights geared toward classifying personality for the business use case."
On the Personality Insights homepage, the tagline is: “Predict personality characteristics, needs and values through written text. Understand your customers’ habits and preferences on an individual level, and at scale.” While the business implications of such a product are clear, Sherwin wanted to fuze his interest in film and linguistics to try out the platform using fictional characterizations with the goal of classifying the personality traits of film protagonists through their spoken dialogue. He took the opportunity to do so for his 4th (of 5) projects during his time in the bootcamp.
In this post, he goes into detail about how he took a massive set of data from the University of California Santa Cruz and brought his vision to life.
See more examples of Metis student projects here.