Feature-Based Voice Analysis for Parkinson’s Prediction Using ML Classifiers
Abstract
The possibility of disease called Parkinson's pre diagnosis allows radical change in course of this sickness by introducing, in a timely manner, adequate therapies. Using machine learning algorithms, the presented research improves initial screening of Parkinson's disease on the examination of specific vocal characteristics. This ML technique applies by using a very large dataset from UC Irvine ML Repository with 197 distinct cases and 22 unique attributes. The accuracy of the KNN classifier turns out to be very accurate with an accuracy of 85%. Other than the KNN classifier, this study will look into other ML algorithms like Support Vector Machine (SVM), Random Forest Classifier, Decision Tree Classifier, and Extra Trees Classifier. Preprocessing steps like SMOTE remove redundant features and further balance the classes to improve the performance of the classifier. LIME analysis around the critical findings sh ows that vocal characteristics such as Spread2, RPDE, and MDVP (Hz) are of utmost importance in predicting Parkinson's disease. These results are extremely important for preliminary diagnosis of disease because it can totally change patient care and afford possibilities for more special and effective treatment options. This might also be through the use of voice analysis tools by patients themselves, possibly at home, feeding data to the phone-based system for the ML algorithms mapping disease course or response to treatment. Such really is the essence of this research—that it truly typifies the latest machine learning methods for forecasting Parkinson's disease and hence opening the way toward early therapeutic interventions for improved patient health outcomes.
How to Cite This Article
Shaik Zubair Ahmed, Mohammed Uzair Khan (2025). Feature-Based Voice Analysis for Parkinson’s Prediction Using ML Classifiers . International Journal of Multidisciplinary Research and Growth Evaluation (IJMRGE), 6(1), 1597-1605. DOI: https://doi.org/10.54660/IJMRGE.2025.6.1.1597-1605
References
- 1. sease: Areview. Neurol India.2018 Mar;66(7\: S26S
- 35. DOI:10.4103/0028-3886.226451.
- 2. Mei J, Desrosiers C, Frasnelli J. Machinelearningforthe Front Aging Neurosci.2021 May;13:
- 633752. DOI:10.3389/fnagi.2021.633752.3. usingmachinelearningalgorithms. Med Hypotheses.2020 May;138:
- 109603. DOI:10.1016/j. mehy.2020.109603.
- 4. Pahuja G, Nagabhushan TN. Acomparativestudyofexistingmachinelearningapprdiseasedetection. IETEJRes.2021;67(1\:
- 414. DOI:10.1080/03772063.2018.1531730.
- 5. Oliveira AM, Coelho L, Carvalho E, Ferreira-Pinto MJ, Vaz R, Aguiar P. Machinelearningforadaptivedeeplosingtheloop. JNeurol.2023 Nov;270(11\:
- 53135326. DOI:10.1007/s00415-023-11873-1.
- 6. Coelho BFO, Massaranduba ABR, Souza CAS, Vianabiomarkersbasedon Hjorthfeaturesimprovedbymachinelearning. Expert Syst Appl.2023 Feb;212:
- 118772. DOI:10.1016/j. eswa.2022.118772.7. diseaseusingmachinelearning. Procedia Comput Sci.2023 Jan;218:
- 249261. DOI: International Journalof Multidisciplinary Researchand Growth Evaluationwww. allmultidisciplinaryjournal. com1605|Page10.1016/j. procs.2023.01.007.
- 8. Little M, Mc Sharry P, Hunter E, Spielman J, Ramig L. Suitabilityofdysphoniameasurementsfor Biomed Eng.2009 Apr;56(4\:
- 1015. DOI:10.1109/TBME.2008.2005954.
- 9. Ali L, Javeed A, Noor A, Rauf HT, Kadry S, Gandomi AH. Parkinsonrefinementthrough L1regularized SVManddeepneuralnetwork. Sci Rep.2024 Jan;14(1\:
- 114. DOI:10.1038/s41598-024-51600-y.
- 10. Neto OP. Harnessingvoiceanalysisandmachinedisease: Acomparativestudyacrossthreedatasets. JVoice.2024 May. DOI:10.1016/j. jvoice.2024.04.020.
- 11. Naranjo L, Perez CJ, Martin J, Campos-Roca Y. Atwo-stagevariableselectionandclassificationapproachforicerecordingreplications. Comput Methods Programs Biomed.2017 Apr;142:
- 147156. DOI:10.1016/j. cmpb.2017.02.019.
- 12. Pahuja G, Nagabhushan TN. Acomparativestudyofdiseasedetection. IETEJRes.2021;67(1\:
- 414. DOI:10.1080/03772063.2018.1531730.
- 13. Ali L, Zhu C, Zhang Z, Liu Y. Automateddetectionofphonationsusinglineardiscriminantanalysisandgeneticallyoptimizedneuralnetwork. IEEEJTransl Eng Health Med.2019;
- 7. DOI:10.1109/JTEHM.2019.2940900.
- 14. Wang W, Lee J, Harrou F, Sun Y. Earlydetectionoflearning. IEEEAccess.2020;8:
- 147635147646. DOI:10.1109/ACCESS.2020.3016062.
- 15. Gunduz H. Deeplearning-classificationusingvocalfeaturesets. IEEEAccess.2019;7:
- 115540115551. DOI:10.1109/ACCESS.2019.2936564.
- 16. Nizamuddin MK, Raziuddin S, Farheen M, Atheeq C, Sultana R. An MPL-CCNmodelforreal-timehealthmonitoringandintervention. Eng Technol Appl Sci Res.2024;14(4\:
- 1555315558. DOI:10.48084/etasr.7684.
- 17. Haque R, Islam MB, Parameshachari BD, Khushbu KG, Rahman S, etal. Bengaliemotionclassificationusinghybriddeepneuralnetwork. In:2023 International Conferenceon Ambient Intelligence, Knowledge Informaticsand Industrial Electronics(AIKIIE\.2023 Nov:
- 17. DOI:10.1109/AIKIIE60097.2023.10389834.
- 18. Liu Y, Wu F, Liu M, Liu B. Abstractsentenceclassificationforscientificpapersbasedontransductive SVM. Comput Inf Sci.2013 Sep;6(4\:
- 125. DOI:10.5539/cis. v6n4p125.
- 19. Shaheed K, Abbas Q, Hussain A, Qureshi I. Optimized Xceptionlearningmodeland Xg Boostclassifierfordetectionofmulticlasschestdiseasefrom X-rayimages. Diagnostics.2023 Aug;13(15\:
- 2583. DOI:10.3390/diagnostics13152583.20.?akir M, Yilmaz M, Oral MA, Kazanci HO, Oral O. Accuracyassessmentof RFerns, NB, SVM, andk NNmachinelearningclassifiersinaquaculture. JKing Saud Univ Sci.2023 Aug;35(6\:
- 102754. DOI:10.1016/j. jksus.2023.102754.