Performance Comparasion of Adaboost and PSO Algorithms for Cervical Cancer Classification Using KNN Algorithm

Muhamad Romdhan, Ubaidilah and Sutedi, Sutedi (2024) Performance Comparasion of Adaboost and PSO Algorithms for Cervical Cancer Classification Using KNN Algorithm. Performance Comparasion of Adaboost and PSO Algorithms for Cervical Cancer Classification Using KNN Algorithm, 10 (2). pp. 1-10. ISSN 2599-3321

Text 31711-95346-3-PB.pdf Download (502kB)

Abstract

Abstract. Cervical most cancers affects women's reproductive organs and is the second maximum generally identified infection amongst girls international. According to the World Health Organization (WHO), over 600,000 women are diagnosed with cervical cancer annually, resulting in more than 300,000 deaths from the disease. Lack of foresight and early cervical cancer identification results in many deaths. To find out if a patient has cancer cells in her cervix, four screening techniques can be used: Hinselmann, Schiller, Cytology, and Biopsy. The KNN method will be used in this study to assess patient health history data, and the Adaboost and PSO algorithms will then be used to optimize the results. The two optimization methods will be compared to find the most accurate model in identifying patterns in cervical cancer patients and predicting patient screening results, whether positive or negative regarding cervical cancer. This research uses the RapidMiner tool. The final results show that the KNN algorithm is able to carry out multilabel classification analysis effectively, and the classification results optimized with PSO produce an increase in the level of accuracy. Purpose: This research aims to evaluate the performance of K-Nearest Neighbor (KNN) algorithm in multilabel classification of cervical cancer and optimise it using Adaboost and Particle Swarm Optimization (PSO) algorithms. This research is important because it can provide an alternative diagnostic method that is more accurate in detecting cervical cancer using medical record data. Methods/Study design/approach: This study uses the Cervical Cancer Risk Classification dataset from the Kaggle Dataset. Preprocessing was done before applying the dataset into the KNN algorithm model. The performance of the KNN algorithm was evaluated by cross-validation method using 10 folds, and the results were measured using Confusion Matrix. Furthermore, the KNN algorithm was optimised using Adaboost and PSO to evaluate the improvement of its accuracy performance. Result/Findings: The test results showed that the KNN algorithm achieved the best accuracy with k=5, with 95.81%, 91.26%, 94.64%, and 93.01% accuracy for Hinselmann, Schiller, Citology, and Biopsy targets, respectively. Optimisation with Adaboost did not provide a significant improvement in accuracy, while PSO improved the accuracy on the Hinselmann target from 95.81% to 95.92%. The average training time for this experiment was about two minutes. The findings show that the KNN algorithm is effective in performing multilabel classification for cervical cancer diagnosis. Novelty/Originality/Value: This research shows that optimising the KNN algorithm with PSO can improve accuracy, although not significantly. This shows the potential for further development to improve the accuracy of cervical cancer diagnostics. Testing the model with the latest observation data and optimising the parameters may provide a better model and be useful as a tool in early diagnosis of cervical cancer. Keywords: Cervical Cancer, Classification, KNN, Adaboost, PSO

Item Type:	Article
Subjects:	eBooks eTheses
Divisions:	Pasca Sarjana > Magister Teknik Informatika
Depositing User:	Romdan Romdan Muhamad Ubaidilah Romdan
Date Deposited:	08 Nov 2024 07:42
Last Modified:	08 Nov 2024 07:42
URI:	http://repo.darmajaya.ac.id/id/eprint/18821

Actions (login required)

View Item