Machine Learning — And Why It Wasn’t As Scary As I Thought
Machine Learning — when most people hear about it, they may initially think of AI robots taking over the world like Skynet from the Matrix or some dystopian world similar to the movie I,Robot. I will admit that prior to beginning my studies as a data scientist, visions of sentient robots also crossed my mind most likely due to my exposure to too much media and scare tactics from the news.
Instead Machine Learning is much more diverse than building battle robots that will steal everyone’s jobs and take lives. In the most simplest of definitions, Machine learning is the study of computer algorithms that improve automatically through experience, according to Wikipedia. For many, the scariest part of that definition may be the mathematical algorithms used.
The more I learn about the topic, the more I see how this is already ingrained into our everyday lives, and this is not something new either. Many industries have been using learning algorithms for decades and most people don’t think twice about it. Most notably, grocery stores use this to print coupons for items that a customer is more likely to buy based off of shopping habits. Usually people are more than excited when the coupon printer at their favorite store gives them a discount for an item they actually want or are planning on purchasing.
Figuring out the best coupons for someone is just a version of a recommendation model, similar to what provides Spotify weekly playlists, YouTube recommendations, and Instagram ads. However diving deeper into this, because of the power of these mathematical algorithms, models are able to even process images and classify words in an essay or any other pieces of writing.
For the third project assigned at Flatiron School, we are tasked with creating models to solve a classification problem of our choice. Data we could use ranges from car crashes, stop and frisk related to Terry v. Ohio, telecom customer retention , and information regarding providing clean water to Tanzania. Just based off the different types of data and the multiple classification problems that can be created, the fact that Machine Learning can be used to solve any of these is just amazing! Also, tying Machine Learning to important social issues such as providing services to impoverished countries and communities, or finding traffic trends to prevent car crashes is not only important work but also shows that Machine Learning is used for much more important things and not something to be fearful of.
For my particular project, I decided to look more into the Terry Traffic Stops dataset that was related to the Terry v. Ohio case. The basic overview of this case is that in 1967 three men were stopped by a policeman described to be in plain clothes. This stop resulted in the officer frisking the men based on his belief that they were suspicious and led him to find weapons on two out of three of the men. It was argued that this stop violated the men’s Fourth Amendment right which protects people from unreasonable searches and seizure by the government. However, in an 8 to 1 decision it was deemed by the court that the officer did not violate the Fourth Amendment.
Despite this case occurring over 50 years ago, issues like this are still important and relevant today as we hear more about racial injustice that occurs to minority communities, but most commonly to the Black and African American community. While it is important to be sensitive to the injustice these communities will sometimes experience, I wanted to see if I can build a classification model to determine whether or not these frisking tactics are useful in making an arrest and what possible other features could be important when making an arrest.
To view my work on this project, it can be viewed here:
