Forecasting SO2 air pollution in Salamanca, Mexico using an ADALINE.
A comparison between a linear regression model and a Non-linear regression model is presented in this work for forecasting of pollution levels due to SO2 in Salamanca city, Gto. Prediction is performed by means of an Adaptive Linear Neural Network (ADALINE) and a Generalized Regression Neural Network (GRNN). Prediction experiments are realized for 1, 12 and 24 hours in advance, and the results for linear regression have been satisfactory. The performance estimation of both models are determined using the Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE). Obtained results are compared. The final results indicated that ADALINE outperforms the past approach using GRNN.
| Attachment | Size |
|---|---|
| PresentacionF_IPROMS2007.wmv | 2.15 MB |
1.
I am not clear why some Fig.(Fig. 2, 3, 4, 6, 7 and 9) have time-shift and some Fig.(Fig. 5 and 8) don't.
2. According to your paper, Linear Regression Neural Network(ADALINE) is better than Non-Linear Regression Network(GRNN)to forecast concentration levels of SO2.
How can I know what Linear or Non-Linear is?
1.
Because we used two different pattern schemes to train the ADA, the result depends on the accommodation of the input patterns in training phase. In ADA I Scheme was use the first pattern by zero, since for the first data, we had not apriori information. In ADA II Scheme was not use the first pattern like zero.
2.Yes, according to the results ADALINE is better than GRNN, but is necessary make more test with other days or months, because we know that dispersion and pollutant concentration depends of the meteorological variables , maybe if we add meteorological variables could obtain the best result
Who wants t know the pollution in one city ?try calculating it for all the world for the last 100 years...and see...the result...all because of machines and man like mexico helicopter tours
Dear Author,
Thank you for submitting your paper and presentation.
However, it would be better to understand your work, if you explain your work with voice.