Predictive Modeling for Course Demand and Revenue Forecasting on Edupro: Moving from Reactive to Proactive Planning

Imagine you are running a giant, digital lemonade stand, but instead of lemons, you are selling online classes. In the world of data science, this is called "predictive modeling." Think of it like a weather forecast, but instead of predicting rain or sunshine, it predicts how many people will buy your lemonade and how much money you'll make. The big idea here is simple: instead of just looking at yesterday's sales to guess what will happen tomorrow, you use a smart mathematical recipe to look at why people bought lemonade in the first place—was it the price? The flavor? The weather?—and use those clues to plan for the future. This matters because if you guess wrong, you might buy too many lemons and waste money, or not enough and miss out on sales. For a company like EduPro, getting this right means the difference between a thriving business and a lemonade stand that runs out of cups.

---

The Story of EduPro's Digital Lemonade Stand

EduPro is a big online education company, kind of like a massive school that lives on the internet. For a long time, the bosses at EduPro ran their business like a captain steering a ship by looking only at the wake behind them. They would look at which courses were popular last month and say, "Hey, that one was a hit! Let's spend all our marketing money on it!" and "Let's hire that teacher for the next one!" But this "look-back" strategy was causing trouble. They were pouring money into courses that were already so popular they couldn't get any bigger (like trying to fill a bucket that's already full), and they were assigning teachers to classes that didn't match their skills, like putting a master chef in charge of a math class.

To fix this, the researchers at EduPro decided to build a crystal ball. They gathered up their data: 10,000 student transactions, 60 different courses, and a list of teachers. They mashed all this information together into a neat table with just 60 rows (one for each course) to see the big picture.

The "Expertise Match" Secret Sauce

One of the first things they discovered was a bit of a mystery. They found that some courses were making zero money, sitting on the shelf like "dead inventory." They also noticed something weird: just because a teacher had a high rating didn't mean their course made more money. It turned out that the company had been assigning teachers randomly, without checking if they actually knew the subject.

So, the team invented a new trick called the "Expertise_Match." Imagine you have a puzzle. If you try to force a square piece into a round hole, it doesn't fit. The researchers created a simple yes-or-no switch: if a teacher's skills matched the course topic, the switch flipped to "1" (Yes!); if not, it stayed "0" (Nope!). They found that about 35% to 45% of the time, the teachers and courses didn't match. By fixing this mismatch, they hoped to make the courses more appealing to students.

The Great Algorithm Showdown: The Simple vs. The Complex

Now came the big test. The team wanted to build a computer model to predict two things: how many students would sign up (Demand) and how much money the course would make (Revenue). They tried two different approaches.

First, they tried the "Simple Way": Linear Regression. Think of this like drawing a straight line through a scatter of dots on a graph. It's a classic, straightforward method that assumes the relationship between things (like price and sales) is a straight path.

Then, they tried the "Complex Way": Random Forest. This is like a team of hundreds of tiny, super-smart detectives (decision trees) who each look at the data differently and then vote on the answer. Usually, in the world of data science, this complex team is supposed to be the champion because it can find tricky, hidden patterns that a simple line might miss.

But here is the twist: because EduPro only had data for 60 courses (a very small number for a computer to learn from), the complex team got confused. They started memorizing the specific details of the 60 courses instead of learning the general rules. This is called "overfitting." It's like a student who memorizes the answers to a practice test but fails the real exam because they didn't understand the concepts. The Random Forest model only got a score of 94% accurate.

The simple Linear Regression, however, was the surprise winner. Because the data set was small and the relationships were fairly straightforward, the simple straight line worked perfectly. It achieved a 98% accuracy score, meaning it could explain 98% of the revenue changes just by looking at the factors they fed it.

What the Numbers Tell Us

The winning model revealed some surprising truths. For getting students to sign up, the teacher's rating mattered a lot, but the price of the course mattered even more (and in a negative way—higher prices meant fewer students). However, for making money, the price was the hero. If you raise the price, you get fewer students, but the money you make per student goes up. The model helps the bosses find the "sweet spot"—the exact price where they make the most total money without scaring everyone away.

They also built a cool, interactive dashboard (using a tool called Streamlit) that looks like a video game control panel. The bosses can slide a bar to change the price, or swap out a teacher, and the dashboard instantly tells them, "If you do this, you'll get X students and make Y dollars." It also has a special "Elasticity Simulator" that lets them play with different prices to see how the crowd size changes.

The Takeaway

The paper concludes that EduPro needs to stop guessing and start using this data-driven crystal ball. They should use the "Expertise_Match" to make sure the right teachers are teaching the right classes, and they should use the pricing tool to find the perfect balance between how much they charge and how many students they get. While the model is incredibly accurate right now (98%), the author warns that as the company grows and gets more data, they will need to keep updating the model to keep it sharp. For now, though, this simple math has turned EduPro from a reactive ship captain into a proactive navigator, ready to steer toward a more profitable future.