Page 157 - 2023-Vol19-Issue2

153 |
Fig. 5. Inception CM.
the diagonal, it also has a high PREC and REC for most of
the classes with a good f1 score. The support column gives
the number of instances of each class in the dataset, which
can be useful to understand the distribution of the classes in
the dataset. It’s worth noting that the model has a lower per-
formance on some classes, such as class 8, 13, and 17, where
the number of false negatives is higher. This indicates that
the model is struggling to correctly classify instances of these of class ”Dalmatian” was incorrectly classified as ”Zebra” as
classes, and further analysis may be needed to understand why
this is the case and how to improve the model’s performance
for these classes as the Figure 5 shown.
The classification report shows that the model has a high
ACC with an PREC and REC of 0.92 and f1 score of 0.92. The
PREC, REC and f1 score for most of the classes are also high
with a good f1 score indicating that the model is performing
well. The support column gives the number of instances of
each class in the dataset, which can be useful to understand
the distribution of the classes in the dataset.
It’s worth noting that the model has a lower performance
on some classes, such as class 14,15,16 and 18, where the
PREC and REC is lower. This indicates that the model is
struggling to correctly classify instances of these classes, and of the class ’Zebra’, 38 instances of the class ’TrafficLight’,
further analysis may be needed to understand why this is the
case and how to improve the model’s performance for these
classes.
3) MobileNet
This is a matrix of perplexity. It is a table that counts the
examples of one class that were consistently predicted to be
instances of another. Each column represents an actual class,
Mohammed, Oraibi & Hussain
Fig. 6. MobileNet CM.
whereas each row represents occurrences in a forecast class
(or vice versa). The diagonal elements represent the number
of correct predictions for each class. The other elements of
the matrix represent the number of incorrect predictions for
each class. For example, 38 images of class ”Zebra” were
correctly classified as ”Zebra” and 1 image of class ”Zebra”
was incorrectly classified as ”Dalmatian”. Similarly, 1 image
shown in Figure 6.
Thus, the model has an ACC of 0.82, which means that it
correctly predicted the class of an image 82% of the time. The
model is more accurate for some classes like Kangaroo, Eiffel
tower, and Television. On the other hand, it is less accurate
for classes like Narendra Modi and IndiaGate.
4) Ensemble Learning
In the given matrix, it can be observed that the model has
performed well with high ACC for most of the classes. The
matrix’s diagonal elements show how many valid classifica-
tions there are for each class. The number of misclassifications
is shown by the off-diagonal components. From the matrix, it
can be seen that the model has correctly classified 40 instances
39 instances of the class ’Vulture’ and so on. It can also be
observed that the model has misclassified 1 instance of the
class ’Zebra’ as ’Maggi’, 1 instance of the class ’TrafficLight’
as ’Vulture’ and so on as the Figure 7 shown. Thus, the model
has performed well with a high ACC of above 80%.
The model has a high ACC of 95%. Each class has a
PREC, REC and F1-S of at least 0.83, indicating that the