Overcoming Computational Challenges of K-Nearest Neighbors: A Memory-Efficient Approach using Gray Level Co-Occurrence Matrix and Firefly Optimization Technique

Feature extraction is a critical step in machine learning, especially when dealing with high-dimensional datasets such as medical data. The k-Nearest Neighbors (KNN) algorithm is widely used for feature extraction due to its simplicity and effectiveness, but it often suffers from high memory consumption, particularly with large datasets such as breast cancer dataset. This work addresses the problem of optimizing memory efficiency of KNN, which can limit its applicability in resource-constrained environments, by proposing a novel approach through integrating the Gray Level Co-Occurrence Matrix (GLCM) for feature extraction with the Firefly optimization technique. GLCM is was employed to reduce the memory requirements of the dataset by extracting texture features, thus minimizing the computational cost during classification. The Firefly algorithm is was then utilized to optimize the selection of relevant features, further improving the memory efficiency and scalability of KNN. Datasets of over 1,000 locally collected breast cancer records were used. The datasets were partitioned into five folds to assess the performance of the models. The memory efficiency of standard KNN was compared with that of KNN optimized with the GLCM and Firefly algorithm (GLCM-FA-KNN). Results show that the GLCM-FA-KNN significantly outperformed the standard KNN in terms of memory efficiency, with an overall average of 5.31KB compared to 9.90KB for KNN. These findings demonstrate that GLCM-FA optimization improves KNN’s memory usage, making it a more suitable choice for large-scale data applications.