Enhanced Prediction of Piperine Content in Black Pepper Using Padmam@G Electrode and Cyclic Voltammetry Signal: A Hybrid Sparse Autoencoder-Regression Approach

Abstract

Current methods for detecting piperine primarily rely on chromatography, which can be costly, time-consuming, and complex—limiting its feasibility for routine, large-scale use. To address these challenges, we developed a graphite electrode embedded with molecularly imprinted polymers (MIPs) tailored for piperine detection in black pepper. This sensor leverages a Poly (N,N-dimethylacrylamide) (PDMAM) and ethylene glycol dimethacrylate (EGDMA) matrix with piperine as the template, delivering high sensitivity and selectivity. Using cyclic voltammetry (CV), we validated the sensor's performance on black pepper samples from four brands, demonstrating practical effectiveness. To enhance prediction accuracy, we employed a convolutional sparse autoencoder (CSAE) model to extract essential features from the electrode’s CV response data. These features were then used with four regression models—K-nearest neighbor regressor (KNNR), gradient boost regressor (GBR), decision tree regressor (DTR), and random forest regressor (RFR)—to predict piperine content. Among these, the CSAE-KNNR model achieved the best results, with a mean absolute percentage error of about 0.5% and an R2 of 0.9999 compared to reverse-phase high-performance liquid chromatography (RP-HPLC) data. Our findings suggest that this sensor, paired with the CSAE-KNNR model, provides a precise and cost-effective alternative to chromatography for piperine detection in black pepper, offering a scalable and reliable tool for food quality analysis. This approach could also support similar applications in the food industry for efficient bioactive compound analysis.

Keywords: Molecular imprinted polymer, piperine, poly(N, N-dimethylacrylamide), differential pulse voltammetry, convolutional sparse autoencoder, KNN regressor