Seed classification, particularly for dry bean seeds, is crucial for ensuring high agricultural productivity and efficiency. This paper investigates the impact of data preprocessing techniques on the performance of machine learning models in the classification of dry bean seeds. Using a data set from the UCI Machine Learning Repository, derived from an experiment by Koklu and Ozkan [9], various pre-processing techniques, such as missing value imputation, outlier removal, and data normalization, were applied. The k-Nearest Neighbors (kNN) classification methods and Multi-Layer Perceptron (MLP) were used to evaluate the effectiveness of these preprocessing techniques. In addition, an enhanced MLP model with optimized parameters was proposed, including the learning rate and hidden layer configuration. The experimental results demonstrate the critical role of data normalization, with Z-Score normalization yielding the best performance improvements. The enhanced MLP model significantly outperformed the baseline model, highlighting the importance of robust preprocessing and careful model optimization. These findings underscore the necessity of comprehensive data preprocessing and fine-tuning of machine learning models to achieve high classification accuracy and efficiency.