This article details evaluating the performance of your time series model after you have uploaded your dataset and created a prediction report. This pertains to the Overview tab on your report, that gives the overall details of the performance of the model built and the distribution of the test data. Please note that we use the 80-20 split of the data for train and test, i.e., 80% data for training and the last 20% data for testing

As you can see here, the model predicted Sales and has a MAPE (Mean Absolute Percentage Error) of 7.66%. The lower the value of MAPE the better the performance of the model. You also have the option to move to the Predictions tab/ Export tab using the two buttons for Start Making Predictions and Automate Predictions respectively

Next we move on to the distribution section where we can see the prediction results of the model on the test data. On the x-axis we see the Dates and on the y-axis we see the Sales forecast. Again, as you hover over the graph you will be able to check the (x,y) values of a particular data point

There are multiple functionalities associated with the distribution graph though. On the top right, you will find the +, - icons to zoom in/out of the graph. You can also choose a specific area on the graph and it will automatically be expanded. Finally you can download the graph as an SVG/PNG/CSV

As soon as you zoom in you will be able to see all the corresponding date values on the x-axis that were not visible before

When you select a particular area for expansion, you have a better understanding of the flow of the distribution from one data point to another as seen on the screenshot here

The Advanced Graphs tab gives a sneak peek into the testing results generated by the model

The Actual vs. Predicted graph shows the line plot for the actual sales (blue) and predicted sales (red) on the test set. The more the overlap, the better the forecast. On the x-axis we have the Dates and on the y-axis we have the sales. The graph is downloadable using the download icon on the top right

The error plot histogram also gives us the same information but in a different way. On the x-axis we have the %error (between actual and predicted value) and on the y-axis we have the count. This means that each bar here represents the number of datapoints with the corresponding %error. For example, there are 2 data points corresponding to an error of 2%