The Impact of Data Re-Sampling on Learning Performance of Class Imbalanced Bankruptcy Prediction Models

Sisodia, Dilip Singh; Verma, Upasana

doi:10.15676/ijeei.2018.10.3.2

Cited by 10 publications

(10 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Terdapat sejumlah teknik untuk menangani kumpulan kelas yang tidak seimbang, baik pada tingkat data maupun algoritma. Pada tingkat data, teknik yang diadopsi secara luas adalah resampling seperti oversampling [14,15] dan undersampling [16]. Teknik-teknik ini memodifikasi probabilitas sebelumnya dari kelas mayoritas dan minoritas dalam kumpulan data pelatihan untuk mendapatkan jumlah kasus yang lebih seimbang di setiap kelas.…”

Section: Pendahuluanunclassified

Dampak SMOTE terhadap Kinerja Random Forest Classifier berdasarkan Data Tidak seimbang

Erlin

Desnelita

Nasution

et al. 2022

matrik

View full text Add to dashboard Cite

Dalam aplikasi machine learning sangat umum ditemukan kumpulan data dalam berbagai tingkat ketidakseimbangan mulai dari ketidakseimbangan kecil, sedang sampai ekstrim. Sebagian besar model machine learning yang dilatih pada data tidak seimbang akan memiliki bias dengan memberikan tingkat akurasi yang tinggi pada kelas mayoritas dan sebaliknya rendah pada kelas minoritas. Tujuan penelitian ini adalah untuk mengevaluasi dampak dari SMOTE (Synthetic Minority Oversampling Technique) pada pengklasifikasi Random Forest untuk memprediksi penyakit jantung. Data berjumlah 299 berasal dari UCI Machine learning Repository digunakan untuk membangun model prediksi berdasarkan 12 variabel independen dan 1 variabel dependen. Kelas minoritas dalam dataset pelatihan di oversampling menggunakan teknik SMOTE (Synthetic Minority Oversampling Technique). Model dievaluasi tidak hanya menggunakan ukuran kinerja Accuracy dan Precision saja, namun juga menggunakan alternatif ukuran kinerja lainnya seperti Sensitivity, F1-score, Specificity, G-Mean dan Youdens Index yang lebih baik digunakan untuk data yang tidak seimbang. Hasil penelitian menunjukkan bahwa teknik SMOTE (Synthetic Minority Oversampling Technique) mampu mengurangi overfitting sekaligus meningkatkan kinerja model Random Forest pada semua indikator. Peningkatan skor Accuracy sebesar 3.45%, Precision 4.8%, Sensitivity 7.1%, F1-score 4.8%, Specificity 2.1%, G-Mean 4.4%, dan Youdens Index 6.3%. Penelitian ini membuktikan bahwa dalam menentukan pengklasifikasi dengan algoritma machine learning seperti Random Forest, kemiringan kelas dalam data perlu diperhitungkan dan diseimbangkan untuk hasil kinerja yang lebih baik.

show abstract

Section: Pendahuluanunclassified

Dampak SMOTE terhadap Kinerja Random Forest Classifier berdasarkan Data Tidak seimbang

Erlin

Desnelita

Nasution

et al. 2022

matrik

View full text Add to dashboard Cite

show abstract

“…The Synthetic Minority Oversampling Technique (SMOTE) [10] is an oversampling technique that adds new synthetic data to minority classes to balance them with the majority class sample. The parameters used are the percentage of minority classes that are exceeded, the total number of minority class data, and data parameters that state the value of the nearest neighbor of the minority class to the majority class.…”

Section: B Synthetic Minority Oversampling Technique (Smote)mentioning

confidence: 99%

“…Undersampling is also a technique for the process of resampling with imbalanced data. A portion of the majority class sample is removed to balance it with the minority sample [10]. Mathematically, the undersampling method can be explained through the below equation [9] :…”

Section: Undersamplingmentioning

confidence: 99%

“…The resulting total distance between the two hyper planes is 9 ‖.‖ . To maximize margins, ‖w‖ is minimized by min 8 9 ‖.‖ 9 (10) If training data is not linearly separated, then a slack < variable can be added which is used as a misclassification of the noisy example. Adding slack variables changes the formula to the following: min 8 9 ‖*‖ 9 + = ∑ < (11) with the provision of !…”

Section: F Support Vector Machine (Svm)mentioning

confidence: 99%

See 1 more Smart Citation

Hybrid Preprocessing Method for Support Vector Machine for Classification of Imbalanced Cerebral Infarction Datasets

Rustam

Utami

Hidayat

et al. 2019

Int. J. Adv. Sci. Eng. Inf. Technol.

View full text Add to dashboard Cite

Cerebral infarction is one of the causes of ischemic stroke in the brain, and machine learning can be used in the detection of cerebral infarction in the brain. In diagnosing the presence of cerebral infarction in the brain, machine learning is used because it is not enough just to use a CT scan to diagnose. Support vector machine (SVM) is a machine learning method that is known for its high accuracy value. However, SVM can produce less optimal results if the data used is imbalanced. If imbalanced data is used, the resulting model will be biased. Therefore, this study uses a hybrid preprocessing method for SVM on the classification of an imbalanced cerebral infarction dataset obtained from the Department of Radiology at Dr. Cipto Mangunkusumo Hospital. This method is a combination of several sampling methods that deal with the problem of imbalanced data and utilizes undersampling and oversampling techniques in combination with SVM. Oversampling modifying the infarction dataset through the duplication of data with a small number of classes to be balanced with a large number of data classes. While undersampling reducing data with a large number of classes to be balanced with a smaller number of data classes. Undersampling and Oversampling are combined into a hybrid method. This method is a hybrid method of the undersampling and oversampling that can be used in SVM. The results of hybrid method using SVM will be compared with the undersampling and oversampling using SVM, individually. And SVM method without preprocessing the imbalanced dataset. The accuracy of the proposed method reached 94% in our evaluations for SVM using a hybrid preprocessing method.

show abstract

“…A study proposed the variants of SMOTEBoost for imbalanced regression task and evaluated its performance using 30 datasets (Moniz et al, 2018). Studies have also evaluated the effect of different data sampling methods on learning performance of individual and ensemble models using highly skewed bankruptcy and credit card fraud datasets (Sisodia & Verma, 2018;Sisodia et al, 2018a). The above-discussed ensemble methods perform better on some skewed dataset with small class IR but show poor performance on datasets with high IR.…”

Section: Related Workmentioning

confidence: 99%

Distinct Multiple Learner-Based Ensemble SMOTEBagging (ML-ESB) Method for Classification of Binary Class Imbalance Problems

Sisodia¹,

Verma²

2019

IJTech

Self Cite

View full text Add to dashboard Cite

Traditional classification algorithms often fail in learning from highly imbalanced datasets because the training involves most of the samples from majority class compared to the other existing minority class. In this paper, a Multiple Learners-based Ensemble SMOTEBagging (ML-ESB) technique is proposed. The ML-ESB algorithm is a modified SMOTEBagging technique in which the ensemble of multiple instances of the single learner is replaced by multiple distinct classifiers. The proposed ML-ESB is designed for handling only the binary class imbalance problem. In ML-ESB the ensembles of multiple distinct classifiers include Naïve Bays, Support Vector Machine, Logistic Regression and Decision Tree (C4.5) is used. The performance of ML-ESB is evaluated based on six binary imbalanced benchmark datasets using evaluation measures such as specificity, sensitivity, and area under receiver operating curve. The obtained results are compared with those of SMOTEBagging, SMOTEBoost, and cost-sensitive MCS algorithms with different imbalance ratios (IR). The ML-ESB algorithm outperformed other existing methods on four datasets with high dimensions and class IR, whereas it showed moderate performance on the remaining two low dimensions and small IR value datasets.

show abstract

The Impact of Data Re-Sampling on Learning Performance of Class Imbalanced Bankruptcy Prediction Models

Cited by 10 publications

References 34 publications

Dampak SMOTE terhadap Kinerja Random Forest Classifier berdasarkan Data Tidak seimbang

Dampak SMOTE terhadap Kinerja Random Forest Classifier berdasarkan Data Tidak seimbang

Hybrid Preprocessing Method for Support Vector Machine for Classification of Imbalanced Cerebral Infarction Datasets

Distinct Multiple Learner-Based Ensemble SMOTEBagging (ML-ESB) Method for Classification of Binary Class Imbalance Problems

Contact Info

Product

Resources

About