A Study on Classification and Performance Evaluation in Logistic Regression Analysis
A Study on Classification and Performance Evaluation in Logistic Regression Analysis, Yalçın Tahtalı,Lütfi Bayyurt,Samet Hasan Abacı, Editör, Ozgur Press , Gaziantep, ss.121-138, 2026
- Yayın Türü: Kitapta Bölüm / Araştırma Kitabı
- Basım Tarihi: 2026
- Doi Numarası: 10.58830/ozgur.pub1404.c5659
- Yayınevi: Ozgur Press
- Basıldığı Şehir: Gaziantep
- Sayfa Sayıları: ss.121-138
- Editörler: Yalçın Tahtalı,Lütfi Bayyurt,Samet Hasan Abacı, Editör
- Ankara Üniversitesi Adresli: Evet
Özet
Logistic
regression is a widely used statistical method for binary classification
problems. This study evaluates variable selection, multicollinearity
management, parameter estimation, and classification performance of the
logistic regression model for binary classification. The dataset consists of
5880 synthetic observations (4000 normal, 1880 anomaly) generated based on the
CytoDiffusion model, initially containing 24 explanatory variables (11
morphological, 4 color based, 9 patient clinical). The data were split into 70%
training and 30% test sets using stratified sampling. For variable selection, a
bidirectional stepwise selection algorithm minimizing the AIC criterion was
used, followed by the Wald test to assess the significance of individual
coefficients. Multicollinearity was examined using the Variance Inflation
Factor (VIF). The stepwise selection automatically resolved the multicollinearity
issue by removing one of the highly correlated variables from the model. The
final model consisted of 12 statistically significant variables (9
morphological, 3 color based), while all patient clinical variables were
excluded from the model. Model performance was evaluated on the test set using
sensitivity, specificity, accuracy, F1 score, and AUC-ROC. The optimal decision
threshold was determined by the Youden J index and found to be close to the
default threshold of 0.50 (0.517). The generalizability of the model was tested
using 10-fold cross-validation. The classification performance metrics
calculated on the test set showed consistency with the 10-fold cross-validation
results. In the odds ratio analysis, the signs and magnitudes of the
coefficient estimates provide interpretable information regarding the direction
and strength of the effect of each explanatory variable on the response
variable. The findings demonstrate that the developed logistic regression model
is a statistically valid and interpretable tool for the given classification
problem.