Imbalanced Dataset Classification using Extreme Learning Machine

Classification of imbalanced datasets using Extreme Learning Machine (ELM) has become a prominent research topic in the classification field. ELM has gained significant popularity in both academic and industrial communities due to its rapid learning capability and outstanding performance. The algorithm can provide quick results when processing large-scale datasets and rapidly train models without requiring complex parameter tuning. For imbalanced dataset classification, ELM optimization algorithms can learn features from the hidden layer and effectively identify minority classes, thereby improving classification accuracy. A key implementation aspect involves randomly initializing input weights and biases while analytically determining output weights through Moore-Penrose generalized inverse computation. The algorithm's architecture typically includes a single-hidden-layer feedforward network where the hidden layer activation function (such as sigmoid or RBF) transforms input data nonlinearly. Additionally, ELM demonstrates strong performance when handling nonlinear problems, with its efficiency stemming from the absence of iterative tuning in the hidden layer. Code implementation often involves matrix operations for weight calculation and can incorporate techniques like weighted ELM or sampling methods to address class imbalance.