Principal component analysis and neurocomputing-based models for total ozone concentration over different urban regions of India |
| |
Authors: | Goutami Chattopadhyay Surajit Chattopadhyay Parthasarathi Chakraborthy |
| |
Institution: | 1. Institute of Radiophysics and Electronics, University of Calcutta, Kolkata, 700 009, India 2. Department of Computer Application, Pailan College of Management and Technology, Kolkata, 700 104, India 3. Department of Business Administration, Pailan College of Management and Technology, Kolkata, 700 104, India
|
| |
Abstract: | The present study deals with daily total ozone concentration time series over four metro cities of India namely Kolkata, Mumbai, Chennai, and New Delhi in the multivariate environment. Using the Kaiser–Meyer–Olkin measure, it is established that the data set under consideration are suitable for principal component analysis. Subsequently, by introducing rotated component matrix for the principal components, the predictors suitable for generating artificial neural network (ANN) for daily total ozone prediction are identified. The multicollinearity is removed in this way. Models of ANN in the form of multilayer perceptron trained through backpropagation learning are generated for all of the study zones, and the model outcomes are assessed statistically. Measuring various statistics like Pearson correlation coefficients, Willmott’s indices, percentage errors of prediction, and mean absolute errors, it is observed that for Mumbai and Kolkata the proposed ANN model generates very good predictions. The results are supported by the linearly distributed coordinates in the scatterplots. |
| |
Keywords: | |
本文献已被 SpringerLink 等数据库收录! |
|