Answers are grounded in this paper and cite their sources.
Imbalanced dataset handling in real time is one of the most challenging tasks in predictive modelling. This work handles the critical issues arising in imbalanced dataset with implementation of artificial neural network and deep neural network architecture. The usual machine learning algorithms fails to achieve desired throughput with certain input circumstances due to mismatched class ratios in the sample dataset. Dealing with imbalanced dataset leads to performance degradation and interpretability issue in traditional ML architectures. For regression tasks, where the target variable is continuous, the skewed data distribution is major issue. In this study, we have investigated a detailed comparison of traditional ML algorithms and neural networks with dimensionality reduction method to overcome this problem. Principle component analysis has been used for feature selection and analysis on real time satellite-based air pollution dataset. Five regression algorithms Multilinear, Ridge, Lasso, Elastic Net and SVM regression is combined with PCA and non PCA to interpret the outcome. To address unbalanced datasets in real-time, deep neural networks and artificial neural network architectures have been developed. Each model's experiments and mathematical modelling is done independently. The Deep neural network is superior compared to other conventional models for performance measures of target variable in imbalanced datasets.
No takes yet. Share an insight, caveat, or question.
Anindya Sen (2024) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: