How to forecast with Mforecast by Moritz


Posted on February 26, 2015 at 9:00 PM


In this post I explain how you can use Mforecast to forecast your time series.

We designed the Mforecast Web-App in such way that it is easy to use. In fact it does not take 5 minutes to understand the functionality. Nevertheless we will work you through the process with an example series!

  • 1. Data
  • 2. Forecasting
  • 3. Validation

1. Data

To get started you need a time series in .txt or .csv format.

Here we have a .txt file with seven columns. There is no header, but we know that the third column contains the date information in fractions of the year. The fith column contains the co2 concentration in unknown measures. The remaining columns are of no interests for us. We note that the columns are separeted by space without any quotes. By scrolling down we checked that no points are missing. This file is ready for uploading!

Now you are ready to upload co2.txt to Mforecast:

Step-by-step

  • 1. File contains date (TRUE) (You can add date information later)
  • 2. Select number of associated columns (3, 5)
  • 3. Choose separator (Space)
  • 4. Choose quote (None)
  • 5. Choose header (FALSE)
  • 6. Choose your file!

In case your file does not contain a date column, select your starting date now!


You should now see your time series plotted and displayed in a data table beneath the plot. Please check, whether the data was imported correctly. You can already change the axes names and download/resize the time series plot (see image below).

2. Forecasting

Select the tab 'Application' and press the 'forecast now!' button. It takes about 5-10 minutes to generate the forecasts. After that you can adjust the horizon and zoom into the forecasting area. You may also resize (1) the plot and download it (2) with a few clicks.

Change the 'Horizon' slider to change the number of steps. The 'Zoom' slider lets you zoom into the forecast area of the plot. You may also change the prediction intervalls (80/95%). Please note that the intervalls are only an approximation for now. Pdf and csv reports are now ready for download.

3. Validation

Select the tab 'Validation'. After the forecast is generated you can check the validation.

Note that we provide a data table where you can see the forecast, the real value and two error measures (APE, Deviation).

Pdf and csv validation reports are now ready for download.


When you are done with your forecast, you may upload a new time series

Warning: As soon as you upload a new series or change anything (i.e. start date) your forecasts are no longer accessible! Please download your forecasts first.

Hope this helps!


comments powered by Disqus