ANALISIS PERFORMA KLASIFIKASI UNTUK DIAGNOSIS PENYAKIT PARKINSON
Parkinson's is a disease that attacks the brain and can cause a gradual loss of motor control due to a lack of dopamine in the brain. The brain is an important organ in the human which controls all the activities performed by the body. In the work, the brain need fluid that is used as the sende...
| Main Authors: | , |
|---|---|
| Format: | Thesis |
| Published: |
[Yogyakarta] : Universitas Gadjah Mada
2014
|
| Subjects: |
| _version_ | 1797036007965589504 |
|---|---|
| author | , sarini vita dewi , Adhistya Erna Permanasari, S.T., M.T., Ph.D. |
| author_facet | , sarini vita dewi , Adhistya Erna Permanasari, S.T., M.T., Ph.D. |
| author_sort | , sarini vita dewi |
| collection | UGM |
| description | Parkinson's is a disease that attacks the brain and can cause a gradual loss
of motor control due to a lack of dopamine in the brain. The brain is an important
organ in the human which controls all the activities performed by the body. In the
work, the brain need fluid that is used as the sender of the signal, the fluid is
called dopamine. Various attempts to treat Parkinson disease has been developed,
method of drug delivery in commonly done in the treatment of this disease, but
long-term given can adversely impact on other organs. Parkinson's gold handling
method is a brain transplant but this method has a high risk and costly.
To minimize the risk, early accurate diagnosis is necessary. The diagnosis
of Parkinson is still relatively rare implemented in machine learning, computerbased
diagnostic method is a promising solution in doing an early diagnosis of a
disease that is by doing classification. It's just that every method certainly having
problems, one of the obstacles identified in medical analysis is a feature that is not
relevant to the classification process. Therefore to reduce the irrelevant features,
feature selection process is used in this study. The presence of feature selection is
expected to improve the performance of data analysis in a more accurate
classification but the implementation of a feature reduction can significantly affect
the classification results. This influence can be a good thing but it can also be an
obstacle in the process of diagnosis, because of the possibility that omitted
features have value that affects the outcome of diagnosis. Because of this
problem, the comparison of the classification results for the full dataset with
datasets that have been in the feature selection is done.
The aim of this study is to obtain the best classification results by using
performance comparison of two classification algorithms, namely SMO and J48
Bagging before and after feature selection is done. Feature selection method used
are CFS, Gain Ratio, RELIEF and Wrapper. From these four results obtained, the
best accuracy performance obtained from a subset feature selection of CFS with
algorithms Bagging J48 in the amount of 91% with a sensitivity value of 0.996,
specificity 0.755, ROC 0.941 and running time of 0.01 second. From the results of
the four evaluation parameters used, shows that the subset with 9 attributes
generated by the CFS method labeled P1, P2, P3, P6, P13, P15, P16, P19, P20 are
the best subset to determine the classification of the dataset of Parkinson disease.
Key words � Parkinson, Brain, Classification, Bagging J48, SMO, Feature
selection, CFS, Wrapper, Gain Ratio, RELIEF. |
| first_indexed | 2024-03-13T23:39:22Z |
| format | Thesis |
| id | oai:generic.eprints.org:133460 |
| institution | Universiti Gadjah Mada |
| last_indexed | 2024-03-13T23:39:22Z |
| publishDate | 2014 |
| publisher | [Yogyakarta] : Universitas Gadjah Mada |
| record_format | dspace |
| spelling | oai:generic.eprints.org:1334602016-03-04T08:19:50Z https://repository.ugm.ac.id/133460/ ANALISIS PERFORMA KLASIFIKASI UNTUK DIAGNOSIS PENYAKIT PARKINSON , sarini vita dewi , Adhistya Erna Permanasari, S.T., M.T., Ph.D. ETD Parkinson's is a disease that attacks the brain and can cause a gradual loss of motor control due to a lack of dopamine in the brain. The brain is an important organ in the human which controls all the activities performed by the body. In the work, the brain need fluid that is used as the sender of the signal, the fluid is called dopamine. Various attempts to treat Parkinson disease has been developed, method of drug delivery in commonly done in the treatment of this disease, but long-term given can adversely impact on other organs. Parkinson's gold handling method is a brain transplant but this method has a high risk and costly. To minimize the risk, early accurate diagnosis is necessary. The diagnosis of Parkinson is still relatively rare implemented in machine learning, computerbased diagnostic method is a promising solution in doing an early diagnosis of a disease that is by doing classification. It's just that every method certainly having problems, one of the obstacles identified in medical analysis is a feature that is not relevant to the classification process. Therefore to reduce the irrelevant features, feature selection process is used in this study. The presence of feature selection is expected to improve the performance of data analysis in a more accurate classification but the implementation of a feature reduction can significantly affect the classification results. This influence can be a good thing but it can also be an obstacle in the process of diagnosis, because of the possibility that omitted features have value that affects the outcome of diagnosis. Because of this problem, the comparison of the classification results for the full dataset with datasets that have been in the feature selection is done. The aim of this study is to obtain the best classification results by using performance comparison of two classification algorithms, namely SMO and J48 Bagging before and after feature selection is done. Feature selection method used are CFS, Gain Ratio, RELIEF and Wrapper. From these four results obtained, the best accuracy performance obtained from a subset feature selection of CFS with algorithms Bagging J48 in the amount of 91% with a sensitivity value of 0.996, specificity 0.755, ROC 0.941 and running time of 0.01 second. From the results of the four evaluation parameters used, shows that the subset with 9 attributes generated by the CFS method labeled P1, P2, P3, P6, P13, P15, P16, P19, P20 are the best subset to determine the classification of the dataset of Parkinson disease. Key words � Parkinson, Brain, Classification, Bagging J48, SMO, Feature selection, CFS, Wrapper, Gain Ratio, RELIEF. [Yogyakarta] : Universitas Gadjah Mada 2014 Thesis NonPeerReviewed , sarini vita dewi and , Adhistya Erna Permanasari, S.T., M.T., Ph.D. (2014) ANALISIS PERFORMA KLASIFIKASI UNTUK DIAGNOSIS PENYAKIT PARKINSON. UNSPECIFIED thesis, UNSPECIFIED. http://etd.ugm.ac.id/index.php?mod=penelitian_detail&sub=PenelitianDetail&act=view&typ=html&buku_id=74131 |
| spellingShingle | ETD , sarini vita dewi , Adhistya Erna Permanasari, S.T., M.T., Ph.D. ANALISIS PERFORMA KLASIFIKASI UNTUK DIAGNOSIS PENYAKIT PARKINSON |
| title | ANALISIS PERFORMA KLASIFIKASI UNTUK DIAGNOSIS PENYAKIT PARKINSON |
| title_full | ANALISIS PERFORMA KLASIFIKASI UNTUK DIAGNOSIS PENYAKIT PARKINSON |
| title_fullStr | ANALISIS PERFORMA KLASIFIKASI UNTUK DIAGNOSIS PENYAKIT PARKINSON |
| title_full_unstemmed | ANALISIS PERFORMA KLASIFIKASI UNTUK DIAGNOSIS PENYAKIT PARKINSON |
| title_short | ANALISIS PERFORMA KLASIFIKASI UNTUK DIAGNOSIS PENYAKIT PARKINSON |
| title_sort | analisis performa klasifikasi untuk diagnosis penyakit parkinson |
| topic | ETD |
| work_keys_str_mv | AT sarinivitadewi analisisperformaklasifikasiuntukdiagnosispenyakitparkinson AT adhistyaernapermanasaristmtphd analisisperformaklasifikasiuntukdiagnosispenyakitparkinson |