The Diagnosis of Dengue Virus Using Principal Component Analysis and Random Forest Based on Raman Spectroscopy Aldo Novaznursyah Costrada, Nina Siti Aminah, Dessy Natalia, Herman, Annita Alni, and Mitra Djamal
Bandung Institute of Technology
Abstract
A classification system based on Raman spectra of dengue virus infected blood serum has been developed and proposed in this study. This development can be useful for health workers in diagnosing dengue fever patients infected with dengue virus. In this study, Raman spectral data from dengue virus infected blood serum samples were used for classification using dimensionality reduction technique combined with Random Forest (RF) classifier. Experimental and quantitative analysis is based on blood serum samples that have been diagnosed with dengue virus through NS1 and IgG/IgM testing. Principal component analysis (PCA) was used as a dimension reduction technique combined with RF (PCA-RF) to highlight variations that can distinguish the Raman spectra of dengue and non-dengue virus infected blood serum. The partition of training data and test data in this model is 8:2 of the total data, also the Iterations of 50 times and estimators of 100 decision trees were applied to each model. The proposed technique has shown good potential to be used in the differentiation between dengue and non-dengue virus infected sera. PCA-RF technique has the best result with 95.89% of accuracy and 0.176 of RMSE, while RF has 89.05% of accuracy and 0.315 of RMSE.
Keywords: Dengue Virus, Raman Spectroscopy, Principal Component Analysis, Random Forest
