**4.3 Performance metrics**

**Tables 5** and **6** show the forecast performance results in the case of raw data and reduced data resulting from PCA method.

Scatter plots (see **Figure 3**) reveal more information about the model's effectiveness. All points in a good model should be close to the diagonal line and have no practical dependencies.

#### **4.4 Residual analysis**

The difference between the actual and expected values is known as residual. The Residual vs. fitted values plot is the first plot in our residual analysis (see **Figure 4**). It is one of the most used model validation graphs. This figure detects outliers and error

**Figure 3.** *Predicted versus observed values plots.*

**Figure 4.** *Residuals versus observed values plot.*

dependencies. The precision of the forecast for that particular value is shown by the distance from the x-axis (0 line).

Moreover, the Residual density plot, as shown in **Figure 5**, can be very informative. If the majority of the residuals are not grouped at zero, the model outputs will likely be biased.

Finally, the last plot (**Figure 6**) is the residual boxplot. It depicts the distribution of absolute residual values.
