Detecting food adulteration, particularly in powdered spices like cinnamon, remains a significant challenge in ensuring quality control and consumer safety. The unique flavor and medicinal properties of cinnamon make it one of the most valued spices in the food industry. However, due to its high commercial value, cinnamon is prone to adulteration throughout the supply chain, necessitating reliable and accurate screening techniques. Fourier transform infrared (FT-IR) spectroscopy is recognized as a non-destructive, rapid, and accurate tool for identifying adulterants in foods and spices. When combined with machine learning algorithms, FT-IR data enable enhanced classification performance, providing a reliable method for distinguishing pure samples from adulterated ones. This approach not only improves detection precision but also establishes a scalable framework for broader applications in food fraud detection, ultimately supporting food safety and quality control standards. In this study, the adulteration of cinnamon powder was assessed using variable selection techniques in conjunction with FT-IR spectral data. Partial least squares (PLS) regression and its variations were developed as calibration models for the quantitative prediction of adulterants. To evaluate the effectiveness of the proposed model, three common adulterants—soybean powder, hazelnut shell powder, and dry bread powder—were selected based on their economic impact and prevalence in industrial adulteration practices. The results indicated that the newly developed real-coded genetic algorithm PLS (RCGA-PLS) model is reliable, yielding low RMSEP and RMSECV values of 0.939 and 0.6523, respectively. Utilizing 40 selected variables, this model achieved a correlation coefficient of 0.987 for predicting adulterants in cinnamon powder. This study highlights that FT-IR spectroscopy, paired with the RCGA-PLS method as a machine learning approach for variable selection, serves as a powerful technique for the rapid detection of adulteration in cinnamon samples. Our findings demonstrate the potential of integrating genetic algorithms with spectroscopic analysis for food authenticity assessment, paving the way for intelligent, data-driven quality control solutions in the food industry

Keywords