This study presents an optimized dual-core photonic crystal fiber (PCF) surface plasmon resonance (SPR) sensor for low refractive index (RI) detection. The sensor integrates a silver (Ag) layer on the fiber structure to enable real-time monitoring of surrounding medium RI changes, with a titanium dioxide (TiO₂) layer protecting the Ag coating. Five key design parameters—pitch, air hole diameter, and Ag thickness—are optimized using the Taguchi L₈(2⁵) orthogonal array, achieving spectral and amplitude sensitivities of 10,000 nm/RIU and 235,882 RIU⁻¹, respectively. Artificial Neural Networks (ANNs), specifically Multi-Layer Perceptron (MLP) and Particle Swarm Optimization (PSO), are used to predict confinement loss (α_loss), while a genetic algorithm (GA) is applied to maximize α_loss. The PSO-ANN model outperforms the MLP-ANN model, achieving an R² value of 0.99. The GA further optimizes the sensor parameters, resulting in a maximum confinement loss of 32.2692 dB/cm under optimized conditions. The sensor demonstrates high sensitivity and accuracy in detecting low RI analytes, with applications in pharmaceuticals and other fields. The study highlights the effectiveness of combining Taguchi optimization, machine learning, and GA for improved PCF-SPR sensor design.This study presents an optimized dual-core photonic crystal fiber (PCF) surface plasmon resonance (SPR) sensor for low refractive index (RI) detection. The sensor integrates a silver (Ag) layer on the fiber structure to enable real-time monitoring of surrounding medium RI changes, with a titanium dioxide (TiO₂) layer protecting the Ag coating. Five key design parameters—pitch, air hole diameter, and Ag thickness—are optimized using the Taguchi L₈(2⁵) orthogonal array, achieving spectral and amplitude sensitivities of 10,000 nm/RIU and 235,882 RIU⁻¹, respectively. Artificial Neural Networks (ANNs), specifically Multi-Layer Perceptron (MLP) and Particle Swarm Optimization (PSO), are used to predict confinement loss (α_loss), while a genetic algorithm (GA) is applied to maximize α_loss. The PSO-ANN model outperforms the MLP-ANN model, achieving an R² value of 0.99. The GA further optimizes the sensor parameters, resulting in a maximum confinement loss of 32.2692 dB/cm under optimized conditions. The sensor demonstrates high sensitivity and accuracy in detecting low RI analytes, with applications in pharmaceuticals and other fields. The study highlights the effectiveness of combining Taguchi optimization, machine learning, and GA for improved PCF-SPR sensor design.