Authentication of honey’s authenticity, type, and botanical origin is important from the point of view of its suitability for human consumption as well as human health. There is a global trend towards unifying analytical methods and introducing modern, accurate, and fast methods with high reproducibility for honey analysis. The International Honey Commission has proposed such methods, which are officially recognized and used in quality control. These methods are also included in the Bulgarian Regulation for sampling and analysis methods for bee honey control. Pollen analysis is used to identify the predominant source of nectar and determine the type of the particular honey. A total of 89 honey samples (59 monofloral and 30 polyfloral) were analyzed using spectral analysis in the near-infrared region. Principal component analysis was first applied to the spectral data to reduce the input vector to ten parameters. Experimental studies were conducted, and seven classification regression models (tree, optimizable tree, three neural networks, quadratic SVM, and kernel SVM) were developed to determine and predict the type of honey sample. Trees and neural networks showed the best results for predicting honey type with high accuracy. The achieved accuracy is 69.3% for the polyfloral honey class. The accuracy of these procedures can be improved by optimizing the models and increasing the honey sample sets. Regression models based on informative characteristics and honey content will form the basis for developing a small device to determine quality indicators and automatically evaluate commercial honey types.