Page 69 - eBook_Proceedings of the International Conference on Digital Manufacturing V2

Fabrication and Characterization of a Low-Cost Piezoelectric using Rochelle
Salt for Energy Harvesting and Sensor Applications

weight of 0.3 to reduce its influence on the overall loss, thereby
mitigating overfitting to dominant classes.

Model performance was monitored using validation loss, and
the best model checkpoint was saved based on the lowest observed
validation loss during training.

Performance Metrics

A comprehensive evaluation of classification models for cervical
cancer requires a suite of performance metrics that capture each
approach's predictive accuracy and clinical relevance. In this
context, true negatives (TN), false positives (FP), false negatives
(FN), and true positives (TP) serve as the fundamental building
blocks for most model assessment metrics:

● A true negative (TN) refers to an individual without
cervical cancer (or a negative cervical sample) who is
correctly classified as disease-free.
● False Positive (FP): An individual without cervical cancer
who is mistakenly classified as having the disease.
● False Negative (FN): An individual with cervical cancer
who is incorrectly classified as disease-free.
● True Positive (TP): An individual with cervical cancer
who is accurately classified as having the disease.

The following metrics leverage these core definitions to
provide quantitative insights into model performance:

1. Accuracy: One of the most reported metrics, indicating
the proportion of correctly predicted samples (both
positive and negative) out of the total number of samples
evaluated by equation 8;
+
= (8)
+ + +

2. Recall (R): Measures the model's ability to correctly
identify individuals with cervical cancer. It is the ratio of
true positives to the total number of actual positives. In

53