Data Cleaning
Demand Forecasting cleans outliers from the data to improve the forecast results.
The app uses upper and lower boundaries based on the historical data and given confidence percentage to calculate the central moving average. Then the app proposes cleaned data based on settings selected in the configuration.
Data is automatically cleaned when you load your dataset. View the cleaned data by clicking the button on the Home page, or using the top menu Forecast → Data Cleaning.
Data cleaning graph
You can check the statistical information about individual records in your data.
Each record is a given location, customer, and product combination in your dataset.
The graph represents:
Historical data
Cleaned data
Upper and lower boundaries
Mean (central moving average)
The graph shows data for a selected record in the data cleaning level. See the Legend on the right for a description of colors used.
Data cleaning table
This table shows the selected record in accordance with Data Cleaning graph.
The statistical data is calculated only for data cleaning level selected in configuration, and the proposed value is inherited with aggregation and disaggregation to other levels.
You can edit the Cleaned data by double-clicking a value in the table.
Other data rows cannot be edited here.
When you edit a record, the color in that period changes.
Outlier values are shown in red in all levels for the record.
Records overview
At the bottom of the page on the status bar you can see the list number of the current record and number of approved records.
By clicking on the current record, the Records overview dialog is displayed with the outliers and approves list.
On the Records overview you can filter by select a customer, product or location or select the only approved records.
Don't forget to select a record in the table and click on the 'Show select'.
Navigation in the records list
There are 3 options to navigate on the Data Cleaning page:
Use Next Record and Previous Record buttons to move through the records
Select record manually in View side panel using the selection boxes for Product, Customer, or Location
Select record in the Records Overview dialog using the Show selected button
Filtering in the record list
You can filter the records list by turning on the settings:
If any filters are turned on the Next Record and Previous Record has added the 'filtered' word.
Approve all or selected only
You can approve all records or only the selected one which is displayed on the graph and table.
The 'Approve selected' function is useful when you need to check several records and want to keep track of which you have already checked.
Calculate forecast
This button confirms all proposed cleaned data and calculates forecast results.
After the calculation finishes you can choose to:
stay on the data cleaning page to make further changes
check the forecast results
Viewing results
The following step is to check the forecast results. When you click Approve all cleaned data, when the calculation is finished you can select View forecast results or you can go to the second workflow step: Forecast Review on the right side of the page.
If you have not accepted all the proposed data, you can go to the Forecast Review page from workflow or the menu to calculate the forecast with the raw historical data for the calculation.