COMPARATIVE APPLICATION OF RAW-ANN AND PCA-ANN FOR THE SPECTROPHOTOMETRIC DETERMINATION OF CAFFEINE, PROPYPHENAZONE, AND PARACETAMOL IN TABLETS

Abstract

Comparative chemometric approaches based on raw artificial neural networks (RAW-ANN) and principal component analysis-artificial neural networks (PCA-ANN) were developed and applied to the simultaneous spectrophotometric determination of caffeine (CFN), propyphenazone (PRPN), and paracetamol (PRC) in a commercial ternary pharmaceutical formulation. The spectral overlap of the three components within the 220-300 nm region renders conventional spectrophotometric methods inadequate without prior separation. The proposed ANN-based models enabled direct analysis of raw UV spectral data without requiring any separation procedure over concentration ranges of 2.5-12.0 µg/mL (CFN), 3.0-12.0 µg/mL (PRPN), and 3.0-16.0 µg/mL (PRC).

Both training approaches demonstrated satisfactory analytical performance, with mean recoveries ranging between 97% and 105%. In particular, PCA-ANN yielded relative standard deviations below 2.5%, indicating enhanced precision and predictive stability compared to RAW-ANN. Statistical evaluation confirmed the robustness and reliability of the developed models. The methods were successfully applied to the quantitative analysis of pharmaceutical tablets, demonstrating their suitability for routine quality control of complex multicomponent formulations.