Evaluating Supervised Learning Classifier Performance for OFDM Communication in AWGN-Impacted Systems

• Supervised Learning in OFDM Systems : The study evaluates the performance of supervised learning classifiers in Orthogonal Frequency Division Multiplexing (OFDM) systems, focusing on their ability to identify and mitigate signal impairments caused by Additive White Gaussian Noise (AWGN). • Classifier Comparison : Various supervised learning algorithms, such as Support Vector Machines (SVM), Decision Trees, and Neural Networks, are analyzed to assess their effectiveness in classifying and decoding data symbols in noisy environments. • Impact of AWGN on Communication : The paper highlights the challenges posed by AWGN in OFDM systems, including signal distortion and bit errors, and explores how machine learning-based classifiers can enhance communication reliability. • Performance Metrics : The study uses key performance indicators such as accuracy, precision, recall, and Bit Error Rate (BER) to benchmark the classifiers, providing a comprehensive analysis of their efficiency in noise-impacted scenarios. • Optimization Potential : The research underscores the potential of optimizing supervised learning models through feature engineering and hyperparameter tuning, enabling robust performance even under severe noise conditions in OFDM communication systems. The evolution of various communication and needs in the everyday usage has fetched an avenue for the major researchers to evaluate and enhance the quality of communication for efficient data delivery with seamless boundaries even through noisy communication channels. Hence the present research work is an attempt to recommend a Machine Learning (ML) methodology for the performance analysis of multiplexed communication channel under Orthogonal frequency division technique. The accuracy of the proposed model is subjected to various metrics considered to measure the performance of the model with limited analysis viz., Peak Signal to Noise Ratio (PSNR), Power to Average Power Ratio (PAPR) and Bit Error Ratio (BER) and Accuracy, in the perspective of the AWGN channel Mean and Variance are considered as variables from 0.4 to 1 and 0.7 ± 0.3 respectively. The ML algorithms are used as a tool for classifier at the receiver and significantly three supervised learning algorithms are used such as K Nearest Neighbours (KNN), Support Vector Machine (SVM) and Random Forest (RF) due to its advantages over the other existing methods and unsupervised learning methods are reserved for the further usage of metric calculation. The data to be transmitted is modulated and subjected to a kind of gaussian noise which is additive in nature and visible throughout the channel making it as noisy channel. With the obtained results we have proposed an hybrid model by implementing various ML algorithm for existing communication system pertaining to the receiver for reconstruction of the received signals by training the system using the data transmitted and received data is subjected to testing proposed ML algorithm. The present research context for ML classifiers - KNN, RF, and SVM - for OFDM systems under AWGN noise, tabulated results show that as noise variance increases, BER and PAPR shoot up, while accuracy and PSNR deteriorates. KNN method performed poorly with a high BER of 0.88 and low PSNR, making it the least effective. RF achieved strong performance with a BER of 0.0102 and high PSNR. SVM outperformed others with better power efficiency (PAPR 9.3–10.56) and PSNR (14.0–25.6), making it ideal for moderate noise environments. The model is adaptable to Free Space Optics and mmWave systems and supports future work in unsupervised and hybrid ML models.