Real-World Uses of Machine Learning, Explained

UPI Study › Courses › Introduction to Artificial Intelligence › Free lessons › Lesson 3.5 Free with a free account

Real-World Uses of Machine Learning

Introduction to Artificial Intelligence · Chapter 3: Programming Foundations and SecuritySection 3.5 · lesson 22 of 6911 min · free with a free account

← Recognizing Threats in Computer SecurityNext lesson: Handling Uncertainty and Measurement in Scientific Work →

Earn 3 college credits for Introduction to Artificial IntelligenceRecommended for credit by NCCRS · Computer Science 373 · self-paced · one payment, lifetime access
Enroll in Introduction to Artificial Intelligence · $250 →Download the syllabus (PDF)

In this lesson

Five machine learning tasks and five kinds of data behind everyday products.

  • Distinguish classification, regression, clustering, recommendation, and anomaly detection
  • Identify tabular, text, image, audio, and time series data and their challenges
  • Describe a real application by naming its task and its kind of data
  • Avoid common mix-ups such as treating 0 or 1 labels as regression

Free study notes

Get this lesson's notes by email

The key points and 4 practice questions with answers, in your inbox. One email, free.

One email. Unsubscribe in one click.

Transcript

Everything said in the video, section by section (1,481 words). Read along, or search it.

Opening Examples

Machine Learning Already on Your Phone

Open a phone, and machine learning is already at work. A message is quietly sorted out of the inbox. A map app predicts when you'll arrive. A music app queues a song you've never heard, and you end up loving it. And a bank asks whether you really just made that purchase. Four products that look unrelated. Yet underneath, they share a pattern: each one learns from past examples, to produce an answer about something new.

Framework

Describe Applications by Task and Data

Machine learning, learning patterns from data, is used in thousands of products. Memorizing them one by one is hopeless, but there's a shortcut. Almost every application can be described by two simple things. The task: the kind of answer the system must produce. And the kind of data it works with. This lesson tours five common tasks, five kinds of data, and how they combine. How the techniques actually learn comes later.

Overview

Five Tasks as Five Questions

Each task is easiest to understand as a question the system answers about an item. Classification asks: which category does this belong to? The answer is a label, from a fixed list. Regression asks: how much, or how many? The answer is a number. Clustering asks: which items naturally belong together? The answer is groups found in the data. Recommendation asks: what would this person want next? It gives a ranked list. And anomaly detection asks: what looks unusual? It gives a flag, or an unusualness score.

Task One

Classification Picks a Category

Classification assigns each item to one category, from a fixed list chosen in advance. For an email, the list might be just two: spam, and not spam. The system learns from many past examples that people have already labeled, like messages users marked as junk, or chose to keep. Shown a new email, it picks the label that fits best. Win a free prize now, click here: spam. The team meeting moved to 3 p.m.: not spam. Your invoice for last month is attached: not spam. Some problems have two categories, others have many, like which of a thousand kinds of object is in a photo. And the output is always a category, even when it adds a confidence, like 93 percent likely spam.

Task Two

Regression Predicts a Number

Regression predicts a number on a continuous scale, instead of a category. The classic example: given a house's size, bedrooms, age, and location, estimate what it will sell for. Related jobs include how long a delivery will take, tomorrow's temperature, or next month's sales. Here are made-up houses. Two bedrooms, 950 square feet, 30 years old: about 172 thousand dollars. Three bedrooms, 1,400 square feet, 12 years: 285 thousand. Four bedrooms, newer and bigger: 410 thousand. The key difference is how success is judged. Predicting 281 thousand for a house that sells at 285 thousand is a close miss. A wrong category is simply wrong. So regression is scored by how far off the number is.

Task Three

Clustering Finds Natural Groups

Clustering sorts items into groups, so items in the same group resemble each other more than items in other groups, with no categories decided beforehand. Picture a shop that records how often each customer visits, and how much they spend each month. Here are five made-up customers. A and B visit once or twice and spend about 20 or 25 dollars. C and D come eight or nine times, and spend 55 to 60. And E comes three times, but spends 240. Clustering might find three groups: occasional low spenders, frequent regulars, and rare big spenders. People then examine the groups, name them, and decide what to do. Nobody told the system what the right groups were.

Quick Check

Predicting Readmission Within 30 Days

Quick check. A hospital wants to predict whether each discharged patient will be readmitted within 30 days. Past records are marked 1 for readmitted, and 0 for not. Which task is this? Pause, and decide. Classification. The answer is one of two categories, readmitted or not, chosen in advance, and learned from labeled records. Regression is tempting, because the labels are written as numbers. But a 1 or a 0 standing for yes or no is still a category, not a quantity. And the groups aren't being discovered, so it isn't clustering.

Task Four

Recommendation Ranks Items for a Person

Recommendation produces a ranked list of items a particular person is likely to want: movies, songs, products, or articles. The input is a record of what that person, and many other people, have watched, bought, rated, or skipped. The output is a personal ordering, and the order matters, because people usually see only the first few. Someone who watched three space documentaries and a nature series might see: first, a new ocean documentary. Then a science series. Then a travel show. Where classification asks which category, recommendation asks: which few items, out of thousands, and in what order?

Task Five

Anomaly Detection Flags the Unusual

Anomaly detection finds items that look unusual, compared with what's normal. In fraud detection, a card's normal pattern might be small purchases near home. Fraud is rare and keeps changing, so it's hard to collect labeled examples of every kind. Instead, these systems often learn what normal looks like, and flag departures from it. Five made-up payments: 12 dollars at lunchtime, 40 at dinner, 25 at noon, all in the home town. Then 2,400 dollars at 3 a.m., in another country. That one is flagged for review. The next, 31 dollars at suppertime, looks normal again. The output is a flag or a score that sends an item to a person. And unusual doesn't always mean wrong: a real traveler's purchase can look odd too.

Kinds of Data

Tabular, Text, Image, Audio, Time Series

Tasks describe what you want to do. Data describes what you have. Five kinds cover most applications, and each brings its own challenges. Tabular data is rows and columns, like a spreadsheet of customer records. It's tricky when entries are missing, or columns mislead. Text is words and sentences of varying length, whose meaning depends on order, context, and tone. An image is a grid of pixels, millions of numbers per picture, and lighting and angle change its look. Audio is a sound wave measured thousands of times a second, with background noise and different accents. And a time series is values recorded in time order, like temperatures or step counts, where order matters, with trends, cycles, and sudden spikes. So ask what the data looks like, before anything else.

Pairing Tasks and Data

Two Labels for Any Application

Any task can meet any kind of data, but some pairings are especially common. A spam filter classifies text, and a photo app classifies images: same task, very different difficulty. Anomaly detection on card records spots a strange payment. On a time series, it spots a sudden spike in a sensor. And tabular data supports all five tasks. So describe any new application with two labels. Grouping news articles by topic: clustering, on text. Estimating a used car's price from age and mileage: regression, on tabular data. Deciding whether a photo shows a cat or a dog: classification, on images. And suggesting songs from listening history: recommendation, on tabular listening records.

Quick Check

Forecasting Tomorrow's Electricity Use

Quick check. A power company records its electricity use every hour. It wants to estimate tomorrow's total use, in megawatt-hours. Name the task, and the kind of data. Pause, and decide. Regression, on a time series. The answer is a number, an amount of electricity, and the readings are recorded in time order. Classification would pick a category, like high or low, not estimate an amount. And the data is hourly readings, not words or sentences, so it isn't text.

Looking Ahead

Questions Later Chapters Answer

This tour also raises questions the rest of the course answers. Images, audio, and text arrive as raw numbers, where no single value means much on its own. How can a system learn to recognize them, without anyone describing what to look for? That's neural networks, in chapter five. How can learned patterns be combined with rules, stored knowledge, and step-by-step reasoning, say, to justify why a payment was flagged? That's logic and knowledge systems, in chapter six. And how do the learning methods behind these tasks actually work, along with language and perception? That's chapter seven.

Common Mistakes

Four Mix-Ups to Avoid

Four common mistakes to avoid. Don't confuse classification with clustering. Classification sorts into categories decided in advance. Clustering discovers the groups itself. Don't assume any number means regression. A label written as 0 or 1 is still a category. Regression predicts quantities, like a price or a time. Don't treat a flagged item as proven fraud: a person still decides what it means. And don't think each kind of data needs its own task. The same five tasks apply across all five kinds.

Key Takeaways

Five Tasks Across Five Kinds of Data

To sum up. Most machine learning applications can be described by a task, the kind of answer wanted, and a kind of data, what the system receives. Classification picks a category, regression predicts a number, clustering finds groups, recommendation ranks suggestions, and anomaly detection flags the unusual. The five kinds of data are tabular, text, image, audio, and time series, each with its own challenges, and any task can meet any of them.

About Introduction to Artificial Intelligence at UPI Study

Introduction to Artificial Intelligence (Computer Science 373) is a self-paced online course of 69 video lessons in 9 chapters, from intelligent agents and search to game strategy, neural networks, logic and reasoning, machine learning, language, robotics and AI ethics. It carries a NCCRS credit recommendation for 3 semester credits.

Every lesson transcript is free on pages like this one, and the videos are free with a free UPI Study account. Enrolling adds a practice quiz after each lesson, the graded work and final exam that earn the credit, lifetime access, and an official transcript on completion. UPI Study has helped over 130,000 students with college credits. ACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities.

Enroll now → Course details Download the syllabus All Computer Science courses Every course with PRO →

Compare: AI vs Current Trends in CS and IT

Why take it here

Why students take Introduction to Artificial Intelligence with UPI Study

How machines search, reason and learnIntroduction to Artificial Intelligence covers how AI systems are built: intelligent agents, search and heuristics, game strategy, logic, neural networks, machine learning, language and perception, and the ethics of AI. It is a common upper-level course in computer science degrees.
Credit at a fraction of the priceThe same 3 credits can cost $750 or more at a US college. Here it is one payment, self-paced, with lifetime access.
Finish your degree soonerTransfer credits reduce the courses you still owe, so you can graduate earlier or lighten a heavy semester.
Strengthen an applicationA completed college-level course on an official transcript shows admissions and scholarship committees you can do the work.
Study on your own scheduleNo start dates, no deadlines. Watch the lessons, practice, and take the final when you are ready.
130,000+students helped with college creditsACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities.

Where the credit goes

Universities that work with NCCRS

Introduction to Artificial Intelligence carries a NCCRS credit recommendation. These universities work with NCCRS; some are UPI Study's signed partners. ACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities.

Thomas Edison State UniversityThomas Edison State UniversitySigned partnerUniversity of the PeopleUniversity of the PeopleSigned partnerCharter Oak State CollegeCharter Oak State CollegeSigned partnerStrayer UniversityStrayer UniversitySigned partnerColumbia Southern UniversityColumbia Southern UniversitySigned partnerIndiana TechIndiana TechSigned partnerPenn StatePenn StateWorks with NCCRSSouthern New Hampshire UniversitySouthern New Hampshire UniversityWorks with NCCRSSUNYSUNYWorks with NCCRS +many more cooperating universitiesFind my college →

Find my college → Not listed? Ask Athena in the corner of this page, or book a 15-minute call.

FAQ

Questions students ask us

About Real-World Uses of Machine Learning

What is the difference between classification and regression in machine learning?
Classification picks a category from a fixed list, while regression predicts a number on a continuous scale. Labeling an email spam or not spam is classification; estimating a house's sale price is regression. A label written as 0 or 1, such as readmitted or not, is still a category. Regression is scored by how far off the number is, while a wrong category is simply wrong.
How is clustering different from classification?
Clustering discovers groups in the data on its own, while classification sorts items into categories chosen in advance from labeled examples. A shop might cluster customers by visit frequency and monthly spending and find occasional low spenders, frequent regulars, and rare big spenders. Nobody told the system those groups existed; people examine and name them afterward.
How does anomaly detection work for credit card fraud?
Anomaly detection often learns what normal activity looks like and flags departures from it, because fraud is rare and keeps changing, so labeled examples are scarce. For a card normally used for small purchases near home, the system would flag a sudden large purchase at 3 a.m. in another country. The output is a flag or score that sends the item to a person, since unusual activity is not always fraud.
What kinds of data does machine learning work with?
Five common kinds are tabular data, text, images, audio, and time series. Tabular data is rows and columns, text is words whose meaning depends on order and context, images are grids of pixels, audio is a sound wave sampled thousands of times a second, and time series are values recorded in time order. Any learning task, such as classification or regression, can work with any of them.

About the Introduction to Artificial Intelligence course

How many college credits is Introduction to Artificial Intelligence?
Introduction to Artificial Intelligence (Computer Science 373) carries a NCCRS credit recommendation for 3 semester credits. ACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities.
How much does Introduction to Artificial Intelligence cost and what is included?
$250, paid once, with lifetime access: all 69 video lessons, a practice quiz after each, the graded work and final exam, and an official transcript when you finish. Students who want several courses choose UPI Study PRO instead.
Which universities accept UPI Study credits?
UPI Study has signed partner agreements with universities such as Thomas Edison State University, University of the People and Charter Oak State College, and works with universities such as Penn State, SNHU and SUNY. ACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities. Use "Find my college" on this page to check yours.
How long does Introduction to Artificial Intelligence take?
It is self-paced with no deadlines. The 69 lessons total about 16 hours of video; most students add reading and practice time and finish in a few weeks to a few months, on their own schedule.
How are the exams done?
Quizzes and assignments are online in the course. The final exam is taken online with proctoring; a person on the UPI Study team checks assignments, usually within two days.
How do my credits reach my college?
When the course shows 100% complete, request the official transcript from your dashboard. It is sent to the college you name, and you can send it to more than one. Check your college first with "Find my college" on this page, or ask Athena.

Ready for the full course? Earn 3 credits in Introduction to Artificial Intelligence.

69 video lessons, a quiz after each one, and the graded work that earns the credit. Recommended for credit by NCCRS.

Enroll in Introduction to Artificial Intelligence →Download the syllabusEvery course with PRO →
Introduction to Artificial Intelligence
3 credits · NCCRSEnroll →