The winning solution for the Mtoto News Childline Kenya Challenge had an RMSE score of 26.04… but what does that actually mean?
In this challenge, you were given the number of calls received by Childline Kenya per hour for five months. The task was to predict the number of calls per hour for the next seven weeks.
The winner of this challenge, @Lawrence_Moruye, plotted this graph focusing on one day in the training set to see how his model performed.
With this graph in front of you, you can see that on average the prediction (in orange) is 26 points (or calls) off the actual number of calls (in blue).
When reporting on the effectiveness of a forecasting model, it is often beneficial to provide a chart with your model score. This can make it easier to understand the prediction, as well as highlight specific points where the models is wrong or right (for example, it might consistently be less accurate on weekends). This can help you improve your model results!
You can find Lawrence’s solution on GitHub. Using some of Lawrence’s techniques can you plot your predictions for:
Try your newly learned skills on Turtle Rescue Forecast Challenge and share your visualisation results in the discussion forum. Let’s learn from each other!