Drinking alcohol can cause many problems. Social problems like family disputes and trouble at work may result from it. Alcohol can also result in accidents. Additionally, it can create economic and health problems. Detecting alcohol consumption early is very important. It helps prevent accidents and health problems. Drinking can cause speech changes such as slurring or shaking and altered vowel sounds due to poor control. This study presents an Intelligent Alcohol Consumption Detection system using a Hyperparameter-Tuned Multi-Layer Perceptron (MLP), enhanced with a Whale Optimization Algorithm (WOA), applied to a vowelized (/a/e/i/o/u) dataset for precise identification of intoxication. The study compares the performance of the MLP model with and without WOA optimization. It evaluates additional machine learning (ML) models, including SVM (RBF and Polynomial kernels), Naive (NB) Bayes, and K-Nearest Neighbors (KNN). The experimental results show that the WOA+MLP model outperforms others in accuracy (ACC), recall (RECA), F1-score, precision (PREC), and AUC score. Specifically, the WOA+MLP model achieved an ACC of 93.14%, a PREC of 93.33%, and an AUC score of 0.9281, superior to the standard MLP model and other classifiers in detecting alcohol consumers and non-consumers. These findings highlight the effectiveness of WOA in fine-tuning machine-learning models for improved classification ACC in alcohol detection scenarios. This research focuses on intelligent alcohol detection. It introduces a non-invasive system that uses speech to detect alcohol consumers easily. This system can be easily used in real-life situations. It aims to improve public safety and help with decision-making.