Accurate classification of dual-band Wi-Fi signals is essential for improving adaptive band selection and maintaining quality of service in complex indoor wireless environments. Although several efforts have addressed propagation modeling, only few works simultaneously examined dual-band classification across both 2.4 GHz and 5 GHz frequencies in realistic scenarios. In this work, we use the measurements data conducted in the Deutsches Museum Bonn, which captures both line-of-sight (LoS) and non-LoS (NLoS) propagation conditions in a complex indoor environment. Ten statistical features are extracted from the received signal data, including mean, standard deviation, and skewness. To classify the signals, multiple machine learning models are evaluated, including k-nearest neighbors, support vector machines, and two deep learning architectures. Among these, model 3A, which is a fully connected neural network comprising three hidden layers using ReLU activation with 64, 32, and 16 neurons, respectively, and a softmax output layer, achieves the best performance. Trained with the Adam optimizer and categorical cross-entropy loss, model 3A attains an overall classification accuracy of 93 \% at the optimal window, thus outperforming the baseline models in terms of precision, recall, and F1-score across all classes. These results highlight the model’s robustness for simultaneous dual-band classification and its potential application in intelligent band selection for next generation Wi-Fi systems.