Refining Sorghum Weed Classification with Dimension Reduction Techniques in Machine Learning

This paper explores the refinement of sorghum weed classification by applying dimension-reduction techniques in machine learning. Utilizing a comprehensive dataset of sorghum weed types, the study aims to enhance classification accuracy by integrating machine learning models with dimensionality reduction methods. Various algorithms are employed to assess model performance, including Logistic Regression, Support Vector Machines, Decision Trees, and Random Forests. In parallel, dimension reduction techniques such as Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), t-distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation and Projection (UMAP) are applied to determine their impact on classification outcomes. The results highlight that LDA significantly enhances decision tree performance, leading to a 16.92% improvement in accuracy. This research highlights the potential of combining machine learning and dimensionality reduction to improve the precision of weed classification, ultimately contributing to more effective agricultural management and weed control strategies.