Hybrid deep learning algorithms on the dimensionally reduced dataset with optimized parameters for high-precision predictions of rainfall in Chhattisgarh State
DOI:
https://doi.org/10.54302/mausam.v75i3.6239Keywords:
Prediction, Precipitation, Genetic algorithm, Principal Component Analysis (PCA), Long Short Term memory network, OptimizationAbstract
Time series forecasting of multi-variant rainfall data was done using a sequential hybrid model. In this model, principal component analysis (PCA) was used to reduce the dimension of the dataset with minimal loss of the original information. The optimized value of window size and the number of Long Short Term Memory (LSTM) units to be used in the deep learning algorithm (LSTM) were estimated using the Genetic algorithm (GA). Thereafter, the dataset retrieved using PCA was inputted with the parameters optimized using GA. Because of these reasons; the model was named PCA-O-LSTM. A comprehensive, comparative study of various models, such as LSTM, PCA–LSTM, GA-LSTM, and PCA-O-LSTM was carried out. For a better interpretation of the results, each of the models was run for various epochs, like 10, 20, 50, 100 and 200. The quality of prediction done using the PCA-O-LSTM model was evaluated by different parameters like using determination coefficient (R2), mean square error (MSE), root-mean-square error (RMSE), mean absolute error (MAE), Normalized error (NORM), RMSE-observations standard deviation ratio (RSR) and cosine similarity (CS). The value of R2, were in the range of (0.962874, 0.972276), (0.970131-0.955826) and (0.950982- 0.972991) with the best value of the said parameter for 200, 200 and 100 epochs in case of GA-LSTM, PCA-LSTM and PCA-O-LSTM, respectively. The best possible value of R2 was seen in the case of the PCA-O-LSTM model in which a dimensional-reduced dataset along with GA optimized the window size and numbers of units were incorporated.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2024 MAUSAM
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
All articles published by MAUSAM are licensed under the Creative Commons Attribution 4.0 International License. This permits anyone.
Anyone is free:
- To Share - to copy, distribute and transmit the work
- To Remix - to adapt the work.
Under the following conditions:
- Share - copy and redistribute the material in any medium or format
- Adapt - remix, transform, and build upon the material for any purpose, even
commercially.