Data Science for Beginners: 5 Machine Learning Examples You Use Everyday Without Realizing It
If this article helped you learn something, let me know in the comments below!
As I went back to university to finish my degree, the first question all my friends and family asked was “So what are you majoring in?”
“I’m majoring in Machine Learning for Business Applications.”
Watching the immediate look on peoples’ faces is like watching a comedy show. Here are some of the common reactions I get:
Its become very clear to me that no one knows what machine learning is. One year ago I really didn’t either.
Trying to figure out what machine learning is is like trying to fish with your bare hands in muddy water. For example, this week I learned about multi-collinearity, support vector machines, and the significance of a p-value in linear regression.
Sure thing Michael!
I won’t bore you with any technical jargon or statistical concepts. That’s only for data science insiders.
However, I’m willing to bet that you use ML multiple times a day.
In fact if you have a Gmail address or an Amazon account…you use Machine Learning without even realizing it.
I find that the easiest way to understand ML is by taking a look at where we use Machine Learning in our own daily habits.
5 Machine Learning Models We Use Everyday
1. Portrait Mode on the iPhone
Have you ever seen someone posting photos on Instagram, and the entire background of their selfie is blurred out?
On the iPhone, this is called “Portrait Mode.”
Most people don’t know that our cell phone cameras have machine learning built in to the software. It’s a type of computer vision technique where the pixels surrounding the main focus area are blurred out.
The example above shows just how powerful that machine learning can be in creating an emotional reaction. The picture on the left is certainly nice, but after using portrait mode the entire photo is dramatically enhanced.
The dual camera system that was introduced with the iPhone 7 Plus began to blur the lines of what was possible with the images captured with a cell phone. (slrlounge.com)
These machine learning algorithms do several things to improve and enhance the smartphone’s camera output:
- Object detection to locate and single out the object(s) (or human) in the image
- Filling in the missing parts in a picture
- Using a certain type of neural network using GANs to enhance the image or even extend its boundaries by imagining what the image would look like, etc.
(analyticsvidhya.com)
Using ML models in our phones, suddenly everyone can be a professional photographer…no photoshop needed!
2. Spam Filter in Gmail
When’s the last time that you actually checked your spam inbox?
Unless someone actually says, “Hey check your spam folder, that thing I sent you might be in there…” most of us don’t bother to check. Of the over 300 billion emails sent everyday, at least HALF are spam!
Here’s a screenshot of my Gmail spam box today:
So then…how does my Gmail account know what’s considered spam and what isn’t?
That’s right! ML shows up here too.
The chart below is from an early research report written in 2011. It shows some of the more popular ML models used in testing spam detection, and their accuracy to detect such emails.
The models used in this paper (from top to bottom on the list) were:
- Naïve Bayes
- Support Vector Machines
- K-Nearest Neighbors
- Neural Networks
- Artificial Immune System
- Rough Sets
As you can see different models produce different results, but achieving a 99% accuracy on any test is considered excellent. You can also help the learning algorithm by identifying and filtering some of the spam yourself. When you see something sneak past the guard and end up in your inbox, make sure to mark it as spam or junk.
The filtering accuracy is enhanced in the long run by adding your newly classified “spam” email to its pattern recognition.
3. Buying Shoes on Zappos
During the pandemic, I’ve completely worn out my comfy clothes. In fact, I can’t even remember the last time I bought any clothing in an actual store.
But if you know anything about ordering clothes online, getting correct the correct size is like throwing darts on a wall.
I’m constantly stuck in this loop:
- Order shirt online; size = Large
- Wait three days for Amazon to deliver
- Try on the shirt only to realize the size is more like a “Sch-medium”
- Stuff the damn shirt back into the box
- Five days later take the trip to UPS and send back to Amazon
- Repeat Step 1
Shoes are also one of the most frustrating things to order online…especially when you have jenky looking pedals like I me. My cousin once said I had “Hobbit Feet” lol!
Online retailer Zappos uses machine learning to provide a bespoke sizing results for customers. They also use the data and ML models to predict customer behavior.
Using Amazon SageMaker, Zappos created models to predict customer apparel sizes, which are cached and exposed at runtime via microservices for use in recommendations. The system enabled single-digit millisecond response times and can handle more than 10 trillion requests per day. (cio.com)
Pretty epic use cases for machine learning. Also really expensive for a company to implement and monitor. A company with a high customer satisfaction rate like Zappos turns a substantial profit from these ML implementations — simply because they understand the pain points of buying shoes online.
Machine Learning in this case is used to make the purchase process that much more enjoyable. The model is continually learning and improving its accuracy as each customer provides more input on the platform.
4. Recommender Systems
So you just finished a hard week at work.
After examining the fridge, and eating all your kids string cheese, you decide that cooking just isn’t in the cards tonight.
So you open up DoorDash or UberEats, and the first thing you see is the “Your Favorites” or “Order Again” section at the top of the app.
“Damn that meal was good,” you mumble to yourself. And click the re-order button.
“Alexa play Bob Marley.”
“Playing Bob Marley, from Amazon Music,” she replies cheerfully, as the music plays throughout the house.
You grab a beer and turn on Netflix. There’s a new movie that just released today! Based upon the last three murder mysteries you binged, this one grips onto you like a piece of Velcro.
You’re in for a long night of Flix and Snax, so you click play and sit back on the couch…only to immediately pick up your phone and open up Amazon. You click around looking for any funny $10 gadgets to buy.
Before you know it, your doorbell rings.
“Foods here!” you exclaim with utter joy.
As you’re sitting there eating a delicious pasta primavera, you grab your phone again…and hop on Instagram. After scrolling for 10 seconds, you suddenly see an advertisement for the same gadget you just looked at on Amazon.
“Wow that’s creepy. How did they know that?”
So you click buy and continue chomping away.
In this example — you used a few different machine learning systems. One of which is pretty standard, called the recommender (or recommendation) system. For simplicity of understanding, these systems are typically built from a collection of your personal likes and dislikes, categorized into certain buckets. After collecting enough information, the machine learning model will then match your data up with new items in similar categories in from other users. It’s easy to see how valuable this is for businesses to deploy.
Machine Learning models help a recommendation system “learn” what you might like and dislike. They’re like a master tailor, using the custom materials we love to create the fabric of society. Everything from restaurants, to music, to productivity, to finding friends on social media: recommendation systems are almost everywhere we look.
5. Traffic Alerts
Have you ever thought about how frequently we use GPS to get around?
We trust these apps implicitly.
The next time you leave the house and open up google maps, take a look at what’s going on inside the app.
You type in an address, it calculates the distance from your location to this location, and then it lets you know the distance, time, and potential traffic you would encounter on the way. It will even suggest other routes based on whether or not you want to pay tolls, take any highways, get gas, etc.
If there isn’t some god-like drone monitoring every street corner, then how does the app understand the time it would take to travel such a distance?
Machine Learning models capture the speed of other phones moving in between those pathways, and learn to recognize patterns. It will calculate whether or not the average speed of phones are traveling at the speed limit for that road.
If the average speed is lower than it should be, then it will rate how far bad traffic might be by showing you those dreaded red lines.
When you see a long line of red, you expect that traffic will be brutal…and prepare for a longer trip.
Simple in theory, but complex in practice.
This is a feat of engineering that is akin to the first electric vehicle, and just one of many everyday uses of machine learning that we have grown fond of.
Machine Learning models are behind some of the most exceptional feats in technology.
The next decade will be nothing short of extraordinary. Its as if we are living in some crazy sci-fi documentary.
I’ve even seen ML used in an effort to save our planet in ways like predicting future animal extinctions, detecting deforestation patterns, and even recognizing what disease a person could have just from listening to the sound of their voice.
Fraud-detection, virtual tutors that adjust their teaching styles, predicting contagious outbreaks in third world countries, these are concepts that in a decade or two will become as old news as the DVD. Google is currently working on an AI assistant that will place “human-like” calls to your local nail salon so you don’t even have to pick up the phone.
SO!
Now that you understand what machine learning is and some real-world associations, you can begin to recognize it throughout your day.
Hopefully this helps paint a clearer picture for you!
