Use Case: Shorten the time to predict entrance exam pass rates to just 10 minutes
With over 35,000 students, Eishinkan is the largest tutoring school in the Kyushu, Japan. Every year, Eishinkan predicts the average score for the public high school entrance exam in Fukuoka Prefecture, as well as the passing score for each school. They published their predictions the day after the test, and their predictions were usually within about 2 percent of the average score. They used these predictions to create materials students could reference to see if they were likely to pass or not.
However, making these predictions wasn’t easy. After the exams, 3,000 students would self-grade their question sheets from the exam, then 25 teachers would meet for two hours to calculate the average and passing scores.
Starting with the March 2018 exams, Eishinkan used the Model Generator in MAGELLAN BLOCKS to train a regression model. They trained their model with ten years worth of data for things like past students scores, acceptance rates for each school, and the number of test takers.
With the model trained, they can enter a new test score and get prediction results in just ten minutes. In terms of accuracy, the new system is about the same as the calculations made by their veteran teachers. With their BLOCKS system working well, Eishinkan’s teachers can use their time more efficiently and focus on the things that only they can do.
- Model GeneratorRegression
- Flow Designer