![](bg6.png)
CNN, etc., is evaluated using an independent dataset,
a task that presents challenges due to limited available
data (Yadav and Kumar, n.d.; Gao, Xu, and Wang
2003). Assessing classifier performance is intricate
when comparing different learning methods, as it
involves evaluating the error rate, which determines
the classifier's success in correctly categorizing
instances (Agarwal et al. 2020). This evaluation is
achieved by considering the mistakes made by the
classifier in each instance. To effectively gauge
classifier performance, independent test data not used
in the model is employed. If additional data is
required, it can be partitioned into training and testing
sets.
Increasing the volume of training data leads to
higher classification accuracy and enhances the utility
of testing data. Nevertheless, a challenge emerges
when the available data is insufficient. To address
this, manual separation of training and test data is
essential (Samek et al. 2019) (Ramkumar, G. et al.
2021). Insufficient data can also introduce issues. To
mitigate this, the holdout approach is commonly
employed, allocating one-third of the data for testing
and the remaining two-thirds for analysis. Cross-
validation is another effective strategy, necessitating
a decision on the number of data folds or partitions to
utilize. In this research, a 10-fold cross-validation
method was adopted, splitting the data into ten
segments with equal representation across classes
(Gunjan and Zurada 2020). This approach involves
dividing the data into ten equal parts and iteratively
using 10% for testing and 90% for training. After
each iteration, one tenth is designated for testing. This
process allows for estimating the overall error over
ten iterations (Khanna et al. 2021).
A notable limitation of the twin study research
method lies in the potential influence of significant
gene-environment correlations or interactions. Such
factors can introduce inaccuracies when attempting to
segregate liability into distinct genetic and
environmental components. In the realm of
technology, a parallel concept to twins emerges
through the utilization of Indian twin technology.
This concept, prevalent within the industrial sector,
involves creating digital replicas of objects or
processes. To achieve this, sensors are strategically
positioned to collect real-time data from the physical
process, which is then fed into AI systems for
processing. Subsequently, these digital twins offer a
platform to comprehensively examine and simulate
the operational mechanics of the object or process,
facilitating in-depth insights into product behavior
and performance simulations.
8 CONCLUSION
The research study focused on the evaluation of two
image processing algorithms, Random Forest and
CNN, for the purpose of identification using face and
ID recognition. The results revealed that Random
Forest exhibited a higher accuracy of 52.3965% in
contrast to CNN's accuracy of 64.3050%. These
findings signify that Random Forest outperforms
CNN in the realm of ID recognition-based
identification.
REFERENCES
Agarwal, Basant, Valentina Emilia Balas, Lakhmi C. Jain,
Ramesh Chandra Poonia, and Manisha Sharma. (2020).
Deep Learning Techniques for Biomedical and Health
Informatics. Academic Press.
B402., Escuela Politécnica Superior Uam, Escuela
Politécnica Superior Uam, B402., and Hadi Abooei
Mehrizi. (2021). “Identifies Polyps in Real Time with
Accuracy 96.67% in Screening Colonoscopy Using
Convolutional Neural Networks (CNN).” IBJ Plus.
Chandana.S, Harini, Chandana S. Harini, and Kumar R.
Senthil. (2022). “A Deep Learning Model to Identify
Twins and Look Alike Identification Using
Convolutional Neural Network (CNN) and to Compare
the Accuracy with SVM Approach.” ECS Transactions.
https://doi.org/10.1149/10701.14109ecst.
DeVerse, Shawn, and Stefan Maus. (2016). “Improving
Directional Survey Accuracy through Real-Time
Operating Centers.” Day 2 Tue, August 23, 2016.
https://doi.org/10.2118/180652-ms.
Gao, Jingbo, Minqiang Xu, and Rixin Wang. (2003).
“Study About Real-Time Finite Element Method Using
CNN.” Computer Technology and Applications.
https://doi.org/10.1115/pvp2003-1908.
Gharbi, Salem Al, Salem Al Gharbi, Qinzhuo Liao,
Salaheldin Elkatatny, and Abdulazeez Abdulraheem.
(2018). “IIncreasing ANN Accuracy, by Improving the
Training Dataset Criteria. Case Study: Identify the
Formation Density from The Drilling Surface
Parameters in Real-Time.” All Days.
https://doi.org/10.2118/192363-ms.
Gunjan, Vinit Kumar, and Jacek M. Zurada. (2020).
Proceedings of International Conference on Recent
Trends in Machine Learning, IoT, Smart Cities and
Applications: ICMISC 2020. Springer Nature.
Hegde, Chandan R., and H. T. Manjunatha. (2022). “Real
Time Indian Traffic Sign Detection Using Image
Processing and CNN.” International Journal of
Advanced Research in Science, Communication and
Technology. https://doi.org/10.48175/ijarsct-5363.
Khanna, Ashish, Deepak Gupta, Zdzisław Pólkowski,
Siddhartha Bhattacharyya, and Oscar Castillo. (2021).
Data Analytics and Management: Proceedings of
ICDAM. Springer Nature.
AI4IoT 2023 - First International Conference on Artificial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry
and Consumer Electronics
492