Faculty Publications

Now showing 1 - 3 of 3

Wavelet coupled MARS and M5 Model Tree approaches for groundwater level forecasting
(Elsevier B.V., 2017) Rezaie-Balf, M.; Naganna, S.R.; Ghaemi, A.; Deka, P.C.
In this study, two different machine learning models, Multivariate Adaptive Regression Splines (MARS) and M5 Model Trees (MT) have been applied to simulate the groundwater level (GWL) fluctuations of three shallow open wells within diverse unconfined aquifers. The Wavelet coupled MARS and MT hybrid models were developed in an attempt to further increase the GWL forecast accuracy. The Discrete Wavelet Transform (DWT) which is particularly effective in dealing with non-stationary time-series data was employed to decompose the input time series into various sub-series components. Historical data of 10 years (August-1996 to July-2006) comprising monthly groundwater level, rainfall, and temperature were used to calibrate and validate the models. The models were calibrated and tested for one, three and six months ahead forecast horizons. The wavelet coupled MARS and MT models were compared with their simple counterpart using standard statistical performance evaluation measures such as Root Mean Square Error (RMSE), Normalized Nash-Sutcliffe Efficiency (NNSE) and Coefficient of Determination (R2). The wavelet coupled MARS and MT models developed using multi-scale input data performed better compared to their simple counterpart and the forecast accuracy of W-MARS models were superior to that of W-MT models. Specifically, the DWT offered a better discrimination of non-linear and non-stationary trends that were present at various scales in the time series of the input variables thus crafting the W-MARS models to provide more accurate GWL forecasts. © 2017 Elsevier B.V.
Artificial intelligence approaches for spatial modeling of streambed hydraulic conductivity
(Springer International Publishing, 2019) Naganna, S.R.; Deka, P.C.
Saturated hydraulic conductivity (Ks) describes the water movement through saturated porous media. The hydraulic conductivity of streambed varies spatially owing to the variations in sediment distribution profiles all along the course of the stream. The artificial intelligence (AI) based spatial modeling schemes were instituted and tested to predict the spatial patterns of streambed hydraulic conductivity. The geographical coordinates (i.e., latitude and longitude) of the sampled locations from where the in situ hydraulic conductivity measurements were determined were used as model inputs to predict streambed Ks over spatial scale using artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS) and support vector machine (SVM) paradigms. The statistical measures computed by using the actual versus predicted streambed Ks values of individual models were comparatively evaluated. The AI-based spatial models provided superior spatial Ks prediction efficiencies with respect to both the strategies/schemes considered. The model efficiencies of spatial modeling scheme 1 (i.e., Strategy 1) were better compared to Strategy 2 due to the incorporation of more number of sampling points for model training. For instance, the SVM model with NSE = 0.941 (Strategy 1) and NSE = 0.895 (Strategy 2) were the best among all the models for 2016 data. Based on the scatter plots and Taylor diagrams plotted, the SVM model predictions were found to be much efficient even though, the ANFIS predictions were less biased. Although ANN and ANFIS models provided a satisfactory level of predictions, the SVM model provided virtuous streambed Ks patterns owing to its inherent capability to adapt to input data that are non-monotone and nonlinearly separable. The tuning of SVM parameters via 3D grid search was responsible for higher efficiencies of SVM models. © 2019, Institute of Geophysics, Polish Academy of Sciences & Polish Academy of Sciences.
ANFIS-based soft computing models for forecasting effective drought index over an arid region of India
(IWA Publishing, 2023) Kikon, A.; Dodamani, B.M.; Deb Barma, S.D.; Naganna, S.R.
Drought is a natural hazard that is characterized by a low amount of precipitation in a region. In order to evaluate the drought-related issues that cause chaos for human well-being, drought indices have become increasingly important. In this study, the monthly precipitation data from 1964 to 2013 (about 50 years) of the Jodhpur district in the drought-prone Rajasthan state of India was used to derive the effective drought index (EDI). The machine learning models hybridized with evolutionary optimizers such as the genetic algorithm adaptive neuro-fuzzy inference system (GA-ANFIS) and particle swarm optimization ANFIS (PSO-ANFIS) were used in addition to the generalized regression neural network (GRNN) to predict the EDI index. Using the partial autocorrelation function (PACF), models for forecasting the monthly EDI were constructed with 2-, 3- and 5-input combinations to evaluate their outcomes based on various performance indices. The results of the different combination models were compared. With reference to 2-input and 3-input combination models, both GA-ANFIS and PSO-ANFIS show better performance results with R2 ¼ 0.75, while among the models with 5-input combination, GA-ANFIS depicts better performance results compared to other models with R2 ¼ 0.78. The results are presented suitably with the aid of scatter plots, Taylor’s diagram and violin plots. Overall, the GA-ANFIS and PSO-ANFIS models outperformed the GRNN model. © 2023 The Authors.

Faculty Publications

Browse

Filters

Settings

Sort By

Results per page

Search Results