Page 124 - Kaleidoscope Academic Conference Proceedings 2024

2024 ITU Kaleidoscope Academic Conference

per the diagram. By using the opinion classifier, we can get 4.5.3 Recall
the polarities of the input sentence.
Recall measures the proportion of true positives among all
actual positives. Its main goal is to reduce misleading
negative results. Recall becomes essential when the cost of
false negatives is high. For instance, you wish to reduce the
amount of real disease cases that are overlooked in a medical
diagnosis system.

=
+
4.5.4 F1-Score

The F-score calculates the harmonic mean of these two
metrics, assigning equal weight to recall and precision. It
offers a harmony between recall and precision. When you
wish to consider both false positives and false negatives, the
F1-score comes in handy. However, it might not be suitable
for all scenarios, especially when you have imbalanced
Fig.4. Support Vector Machine classification. classes.

∙
1 = 2 ∙
+
4.5 Performance Evaluation

This model can evaluate the model performance by accuracy,
precision, recall and F-score, it is crucial to consider the 5. CONCLUSIONS
trade-offs between these metrics and the problem's context.
Below is an explanation of each: This research study concludes with a thorough framework
for fine-grained sentiment analysis on code-mixed language
4.5.1 Accuracy that is especially designed to comprehend the activities of
social media users. There are several frameworks available
Accuracy is defined as the proportion of correctly detected for Sentiment Analysis but none of them worked on Fine
instances among all instances. Although it's an easy-to- Grained and with Code-Mixed language. Therefore, this
understand statistic, it can be deceptive, particularly in framework (FGSACML) has a futuristic approach and has
datasets that are unbalanced and have a dominant class. For not been done before. The framework tackles the difficulties
instance, even if a model with 95% accuracy that classifies brought about by the dynamic and multilingual character of
everything as class A might not be particularly useful if 95% social media discourse by using a methodical approach that
of your data falls into class A and only 5% into class B. combines linguistic elements, machine learning techniques,
and domain-specific knowledge. The model offers user
+
= attitudes, preferences, and interactions of sentiment analysis
+ + +
on code mixed language. This study advances our knowledge
4.5.2 Precision of code-mixed language sentiment analysis and applies new
feature models for improving user engagement, sentiment
The percentage of true positives, or positively anticipated tracking, and social media monitoring. This framework's
cases, among all positively predicted cases is known as continues development and improvement offers bright
precision. Its main goal is to reduce false positives. When the opportunities for improving our understanding of social
cost of false positives is significant, precision matters. For media behavior in the digital era.
instance, you wish to reduce the quantity of valid emails
flagged as spam in a spam email detection system.

=
+

– 80 –