ANA215 Exam Predictor App

Gerrie van der Merwe
2019-12-19

Introduction

The ANA215 Exam Predictor App is an application developed by using R Shiny. The app uses input provided by the user to calculate a prediction for the user's exam mark.

ANA215 is the subject “Introduction to Paleoanthropology” offered at the University of Pretoria (South Africa).

The Input

The following input is required for the app:

  • Indicate whether the learner is a repeater of not
  • Indicate whether any practical were missed
  • Numeric input for how many lectures were missed
  • Click-up test marks
  • Assignment marks
  • Semester test marks

The Output

The following output components are provided by the app:

  • Subject Marks, a table that displays a summary of the subject marks based on the input
  • Exam Prediction (Text), indicates the exam admittence and what the predicted exam mark range would be
  • Exam Prediction (Graph), shows the exam prediction range over a normal distribution fit of what the class previously achieved. On the next slide you will see an example of the Graph output

Exam Prediction Graph

  x <- seq(0, 100, length=1000)
  y <- dnorm(x, mean=44.6, sd=14)
  #plot(x, y, type="l", lwd=1)

  plot(y~x, type="l", main="Prediction Range vs Class Distribution" ,xlab="Exam Mark",
       ylab="Density") + abline(v=(45)) + abline(v=(55)) + abline(v=(35))

plot of chunk graph

integer(0)

The Model

Several model fits were tested and GBM (Gradient Boosting Model) was selected for the app. The final model was within 10% accuracy when comparing the validation sets' marks vs the predictions.

The range used for the app is the prediction +/- the average absolute error of the model. This means that the range may not always include the result.

Other Info

For more info, please see the documentation provided: “https://github.com/gpmerwe/Data-Products-Projects/blob/master/README.md”