Skip to content

Mission 13 - Comparing Predictions

Estimated time for completing this mission: 30 mins

Learning Objective

Creating charts with advanced functionality in App Designer.

BLOX used in this mission:

Logic BLOX

  • MyData/FinalData
  • AI/Regression Model
  • AI/Time Series Model
  • Database/Group By

App BLOX

  • Chart
  • Panel
  • Input
  • Button
  • Label
  • Multiselect

Data

The same data set will be used as for the previous exercise. In case you do not have the data set you can download it here: FinalData.xlsx

Steps

Please refer to the Navigation Guide to perform the steps below

App Designer:

  1. Import the dataset from Data section to Module4 folder
  2. In Module4 create new app named Comparing Predictions
  3. Include all the BLOX which are mentioned in the description above as App BLOX
  4. Make the "panel" BLOX large enough to fit content in it
  5. Drag and drop all the necessary components into the panel (2x chart, 2x multiselect, 4x label, 2x button) and arrange them in proper order (see the image below for help)

startBlox

App Overview

6. In App Designer open logic view and drag and drop all the necessary BLOX

7. First, connect the first button to the "FinalData" BLOX

8. Next, connect the data to "Group By" BLOX, specify the column ("ORDERDATE") and operation ("Sum")

9. Connect the result from "Group By" to 'Historical Data' input of "Train Series Model", specify 'Time Variable' ("ORDERDATE"), 'Time Format' (%Y/%m/%d), 'Target Variable' ("SALES"), 'Look Back Window Size' (50), 'Training Intensity' ('Medium') and 'Model Name'

10. In the 'Train Series Model' left sidebar, open "Feature Variables" input, and connect "Multiselect" element to it

Hint

Multiselect BLOX combines your chosen options to a list with comas. Double click on the BLOX, and add more items with the "+Add" sign. Name all items the same as the column names in the "Group By" function as we are going to use these as an input for both the Regression and Time Series function.

startBlox
Multiselect logo
Feel free to play around with the different feature variables. See how can you produce the best (most accurate) results. (Note that if you chose "ORDERDATE" for the Regression model, in the "Regression Model" BLOX you need to set "x is time series" as yes, but in that case you cannot chose other feature variables)

11. Now connect the output ('Compare Graph') from 'Train Series Model' to the input of 'Chart' element

12. Next repeat the steps (7, 8) for second 'Chart' element

13. Then, connect the output from 'Group By' to 'Regression Model' BLOX's 'Learning Data' input

14. From 'Multiselect' element, connect to 'Features' from the 'Regression Model' BLOX

15. Then specify 'Target' as SALES, 'Exponential Degree' as Fifth Power, 'x is time series' to no and give a name to the model

15. Now, connect the 'Line Chart' output from the 'Regression Model' to the second 'Chart' element

16. Finally run the app, and play around with the two models and the different feature variables

Results and Summary

startBlox

Final Logics

startBlox

Result in Working App View for Time Series

startBlox

Result in Working App View for the Regression

From the results, it is clear that the regression model fit the data better. Different data require different models, that is why it is crucial to have all these tools available when it comes to prediction.

Well done! Now let's move on to the next module!