Appavu Suresh
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Appavu Suresh has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Muthuswamy Kalpana
Department of Computer Science, Tamil Nadu Agricultural University, Coimbatore 641 003, India
6 shared publications - Narasimma Bharathi Venkatesa Palanichamy
Agriculture College and Research Institute, Tamil Nadu Agricultural University, Coimbatore 641 003, Tamil Nadu, India
6 shared publications - Natarajan Balakrishnan
Department of Agriculture, Tamil Nadu Agricultural University, Coimbatore 641 003, Tamil Nadu, India
6 shared publications - Marimuthu Rajavel
Office of the Public Relations, Tamil Nadu Agricultural University, Coimbatore 641 003, Tamil Nadu, India
5 shared publications - Vasudevan Balamurugan
Office of the Dean (Agriculture), Tamil Nadu Agricultural University, Coimbatore 641 003, Tamil Nadu, India
5 shared publications - Rajaram Dhivya
Department of Business Administration, PSGR Krishnammal College for Women, Coimbatore 641 004, Tamil Nadu, India
4 shared publications
Latest Publications
- Open Access
- 17 October 2025
- Published by Horizon E-Publishing Group in Plant Science Today
- Vol. 12 (sp4)
Foreign and Domestic Institutional Investors: A comprehensive analysis of their dynamic relationship in the Indian capital market
The 1991 liberalization policy, which introduced Foreign Institutional Investors (FIIs) to the market, to recent years, revealed a shift in the market dynamics. While FIIs have historically played a dominant role, recent years have seen increased contributions from domestic giants like mutual funds. The study employs a comprehensive methodology involving Granger causality tests and the Threshold Generalized Autoregressive Conditional Heteroskedasticity (TGARCH) model to analyze the causal relationship and impact on market volatility. The analysis covers the period from January 2017 to November 2023, utilizing data from authoritative sources like the National Stock Exchange (NSE) and the National Securities Depository Limited (NSDL). The findings reveal a bidirectional causality between Domestic Institutional Investors (DII) and FII gross sales and a unidirectional causality in gross and net purchases/sales. The results of the TGARCH model demonstrate significant impacts of negative news on volatility in various indices, with indications of a leverage effect in some cases. The study concludes that while both FIIs and DIIs influence market volatility, negative news has a larger impact on volatility than positive news of the same magnitude. The recent news impacts high on the market volatility than the historical news which is observed from the β variable values. The findings contribute to understanding the intricate dynamics of institutional investors in the Indian capital market, providing valuable insights for investors and policymakers.
- 17 October 2025
- Published by Horizon E-Publishing Group in Plant Science Today
- Vol. 12 (sp4)
Open AccessForeign and Domestic Institutional Investors: A comprehensive analysis of their dynamic relationship in the Indian capital market
The 1991 liberalization policy, which introduced Foreign Institutional Investors (FIIs) to the market, to recent years, revealed a shift in the market dynamics. While FIIs have historically played a dominant role, recent years have seen increased contributions from domestic giants like mutual funds. The study employs a comprehensive methodology involving Granger causality tests and the Threshold Generalized Autoregressive Conditional Heteroskedasticity (TGARCH) model to analyze the causal relationship and impact on market volatility. The analysis covers the period from January 2017 to November 2023, utilizing data from authoritative sources like the National Stock Exchange (NSE) and the National Securities Depository Limited (NSDL). The findings reveal a bidirectional causality between Domestic Institutional Investors (DII) and FII gross sales and a unidirectional causality in gross and net purchases/sales. The results of the TGARCH model demonstrate significant impacts of negative news on volatility in various indices, with indications of a leverage effect in some cases. The study concludes that while both FIIs and DIIs influence market volatility, negative news has a larger impact on volatility than positive news of the same magnitude. The recent news impacts high on the market volatility than the historical news which is observed from the β variable values. The findings contribute to understanding the intricate dynamics of institutional investors in the Indian capital market, providing valuable insights for investors and policymakers.
- Open Access
- 30 September 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-20
Weather-driven groundnut price forecasting and profitability assessment of cropping patterns in Tamil Nadu using boosting algorithms
This research explores the integration of weather-based forecasting and profitability analysis to optimize groundnut-based cropping patterns in Tamil Nadu. Groundnut, a crucial oilseed crop, is significantly influenced by weather variability, which impacts its price and profitability. The study leverages advanced boosting algorithms, including Light Gradient Boost, XGBoost, HistGradientBoosting, and CatBoost, to forecast groundnut prices using a multivariate approach that incorporates weather parameters like Minimum and Maximum Temperature, Relative Humidity and Rainfall. Weather parameters were discretized using \(\:k\)-means clustering. Crop prices were decomposed using Seasonal-Trend decomposition based on LOESS and each component was forecasted separately. HistGradientBoosting consistently outperforms other models, achieving the lowest multivariate Mean Absolute Error (MAE) across most crops and districts, underscoring its capability to handle complex, high-dimensional data effectively. The results reveal a substantial improvement in forecasting accuracy with multivariate models compared to univariate ones, establishing the importance of integrating weather features. Groundnut-related cropping patterns were analyzed for profitability using forecasted prices, with patterns involving high-value crops, such as onion in Namakkal, achieving the highest benefit-cost ratio (BCR) of 2.18. Patterns involving black gram also consistently outperformed green gram in economic efficiency. The findings emphasize the need for region-specific, weather-informed cropping strategies to maximize returns for farmers while mitigating risks.- 30 September 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-20
Open AccessWeather-driven groundnut price forecasting and profitability assessment of cropping patterns in Tamil Nadu using boosting algorithms
This research explores the integration of weather-based forecasting and profitability analysis to optimize groundnut-based cropping patterns in Tamil Nadu. Groundnut, a crucial oilseed crop, is significantly influenced by weather variability, which impacts its price and profitability. The study leverages advanced boosting algorithms, including Light Gradient Boost, XGBoost, HistGradientBoosting, and CatBoost, to forecast groundnut prices using a multivariate approach that incorporates weather parameters like Minimum and Maximum Temperature, Relative Humidity and Rainfall. Weather parameters were discretized using \(\:k\)-means clustering. Crop prices were decomposed using Seasonal-Trend decomposition based on LOESS and each component was forecasted separately. HistGradientBoosting consistently outperforms other models, achieving the lowest multivariate Mean Absolute Error (MAE) across most crops and districts, underscoring its capability to handle complex, high-dimensional data effectively. The results reveal a substantial improvement in forecasting accuracy with multivariate models compared to univariate ones, establishing the importance of integrating weather features. Groundnut-related cropping patterns were analyzed for profitability using forecasted prices, with patterns involving high-value crops, such as onion in Namakkal, achieving the highest benefit-cost ratio (BCR) of 2.18. Patterns involving black gram also consistently outperformed green gram in economic efficiency. The findings emphasize the need for region-specific, weather-informed cropping strategies to maximize returns for farmers while mitigating risks. - Open Access
- 30 April 2025
- Published by Horizon E-Publishing Group in Plant Science Today
Using machine learning models to forecast methane emissions from agriculture in India
Methane (CH4) is a potent and powerful greenhouse gas with significant warming potential. CH4 has approximately 84 times the global warming potential (GWP) of CO2 over a 20-year period and 25 times over a 100-year period. Methane persists in the atmosphere for approximately a decade before breaking down through oxidation processes. In order to forecast methane emissions from agricultural and related activities, this study applied an evaluation approach. This study analyzed annual data from 1990 to 2021, collected from the FAOSTAT website, using 7 machine learning models: Random Forest, LASSO, Gradient Boosting, AdaBoost, XGB, Ridge Regression and Linear Regression. The results indicate that the Linear Regression model outperformed the other models in predicting methane emissions. The results show that linear regression is more effective than the various machine learning algorithms. The linear regression model (R² = 0.98, RMSE = 1.95, MSE = 3.86 and MAE = 1.45) achieved the best performance among all models in terms of accuracy. The comparative study's findings should yield a highly accurate assessment of the methane emissions from agricultural areas and help with the creation of laws or other policies aimed at reducing those emissions. These findings can assist the Indian government in formulating effective policies to mitigate methane emissions while maintaining agricultural productivity. Our approach estimates methane emissions from agriculture in India with an R² value of 0.99, indicating high predictive accuracy.
- 30 April 2025
- Published by Horizon E-Publishing Group in Plant Science Today
Open AccessUsing machine learning models to forecast methane emissions from agriculture in India
Methane (CH4) is a potent and powerful greenhouse gas with significant warming potential. CH4 has approximately 84 times the global warming potential (GWP) of CO2 over a 20-year period and 25 times over a 100-year period. Methane persists in the atmosphere for approximately a decade before breaking down through oxidation processes. In order to forecast methane emissions from agricultural and related activities, this study applied an evaluation approach. This study analyzed annual data from 1990 to 2021, collected from the FAOSTAT website, using 7 machine learning models: Random Forest, LASSO, Gradient Boosting, AdaBoost, XGB, Ridge Regression and Linear Regression. The results indicate that the Linear Regression model outperformed the other models in predicting methane emissions. The results show that linear regression is more effective than the various machine learning algorithms. The linear regression model (R² = 0.98, RMSE = 1.95, MSE = 3.86 and MAE = 1.45) achieved the best performance among all models in terms of accuracy. The comparative study's findings should yield a highly accurate assessment of the methane emissions from agricultural areas and help with the creation of laws or other policies aimed at reducing those emissions. These findings can assist the Indian government in formulating effective policies to mitigate methane emissions while maintaining agricultural productivity. Our approach estimates methane emissions from agriculture in India with an R² value of 0.99, indicating high predictive accuracy.
- Open Access
- 24 April 2025
- Published by Horizon E-Publishing Group in Plant Science Today
Predicting small millets productivity based on machine learning models: A comprehensive study
People in rural and urban areas prefer small millet-based food products and they are more popular in markets. Before the green revolution, little millets were widely cultivated for food and fodder, but dietary preferences changed and they were no longer widely planted. Machine learning techniques have been applied in agriculture recently to analyze and predict crop yields. A major concern for farmers during the growing season is estimating their expected yield. In this study, data on area and production for small millets in Tamil Nadu are collected over 50 years. The machine learning models are used to predict productivity with the available small millets dataset. Four different machine learning models are used to estimate small millets productivity. With an accuracy of 95.46 %, a mean absolute error (MAE) of 0.0709, a root mean square error (RMSE) of 0.014 and an R-square value of 0.94, the Random Forest regressor performed better than the other models. The current research helps the farmers in mitigating potential losses, as their financial stability and productivity output are closely related. Additionally, the study provides valuable insights for better planning and implementation. Furthermore, the Random Forest regressor offers insightful information to help farmers maximize their farming techniques and make well-informed judgments.
- 24 April 2025
- Published by Horizon E-Publishing Group in Plant Science Today
Open AccessPredicting small millets productivity based on machine learning models: A comprehensive study
People in rural and urban areas prefer small millet-based food products and they are more popular in markets. Before the green revolution, little millets were widely cultivated for food and fodder, but dietary preferences changed and they were no longer widely planted. Machine learning techniques have been applied in agriculture recently to analyze and predict crop yields. A major concern for farmers during the growing season is estimating their expected yield. In this study, data on area and production for small millets in Tamil Nadu are collected over 50 years. The machine learning models are used to predict productivity with the available small millets dataset. Four different machine learning models are used to estimate small millets productivity. With an accuracy of 95.46 %, a mean absolute error (MAE) of 0.0709, a root mean square error (RMSE) of 0.014 and an R-square value of 0.94, the Random Forest regressor performed better than the other models. The current research helps the farmers in mitigating potential losses, as their financial stability and productivity output are closely related. Additionally, the study provides valuable insights for better planning and implementation. Furthermore, the Random Forest regressor offers insightful information to help farmers maximize their farming techniques and make well-informed judgments.
- Open Access
- 15 April 2025
- Published by Horizon E-Publishing Group in Plant Science Today
Assessing the effectiveness of Artificial Neural Networks and PSLR models in predicting per capita food grain production in India
Every living thing needs food. In addition to fostering social progress and economic expansion, agriculture is essential to our everyday existence. India has made great strides toward guaranteeing a sufficient supply of food since gaining its freedom. While India's population has tripled, food grain production has more than quadrupled. Consequently, there are now substantially more food grains available per individual. To meet the country's food needs, rice and wheat production is essential. Decision-makers need access to accurate forecasts to identify this need, put appropriate plans into place and allocate the required administrative resources. Based on artificial neural networks and PSLR, the per capita availability of food grain production in India was estimated. Indiastat and the FAO provided the historical data for the country from 1951 to 2022. To examine the food grain per capita availability (Kgs. Per Year), we used two effective analytical methodologies, artificial neural networks (ANNs) and Partial least squares (PLS) regression. The models' performances were compared using four relevant performance criteria to determine which model is best for future forecasting. The results show that, when it comes to accurately predicting the per capita number of dietary grains, ANN outperforms the PLSR model. For the ANN method, the values of MAE, MSE, RMSE and R2 were 3272.11, 1.748, 4194.28 and 0.956, in that order. The study discovered that PLS also functioned well, with very little difference between the two models' performance indicators.
- 15 April 2025
- Published by Horizon E-Publishing Group in Plant Science Today
Open AccessAssessing the effectiveness of Artificial Neural Networks and PSLR models in predicting per capita food grain production in India
Every living thing needs food. In addition to fostering social progress and economic expansion, agriculture is essential to our everyday existence. India has made great strides toward guaranteeing a sufficient supply of food since gaining its freedom. While India's population has tripled, food grain production has more than quadrupled. Consequently, there are now substantially more food grains available per individual. To meet the country's food needs, rice and wheat production is essential. Decision-makers need access to accurate forecasts to identify this need, put appropriate plans into place and allocate the required administrative resources. Based on artificial neural networks and PSLR, the per capita availability of food grain production in India was estimated. Indiastat and the FAO provided the historical data for the country from 1951 to 2022. To examine the food grain per capita availability (Kgs. Per Year), we used two effective analytical methodologies, artificial neural networks (ANNs) and Partial least squares (PLS) regression. The models' performances were compared using four relevant performance criteria to determine which model is best for future forecasting. The results show that, when it comes to accurately predicting the per capita number of dietary grains, ANN outperforms the PLSR model. For the ANN method, the values of MAE, MSE, RMSE and R2 were 3272.11, 1.748, 4194.28 and 0.956, in that order. The study discovered that PLS also functioned well, with very little difference between the two models' performance indicators.
- Open Access
- 6 March 2025
- Published by Horizon E-Publishing Group in Plant Science Today
A novel approach for predicting net irrigated area in India using hybrid deep learning architectures
Studying irrigation systems is crucial to ensuring efficient freshwater utilization and conservation. This study examines the efficacy of forecasting the net irrigated area for future generations to create a model of prediction that can efficiently exchange water demand. To improve the forecast, we generate a model using two-hybrid deep learning techniques to predict irrigation demands: Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) and Convolutional Neural Network-Gated Recurrent Unit (CNN-GRU). These models effectively capture complex variables from diverse data sources, including rainfall patterns, irrigated area statistics and various irrigation system parameters. The main ideas, noteworthy contributions and crucial quantitative results from our study on net irrigated area projection are outlined in this publication. Our main contribution is the development of unique hybrid deep learning approaches that effectively integrate the CNN-LSTM and CNN-GRU architectures. Better predictions are made possible by the models’ design, which consists of parallel CNN layers that independently interpret certain input features. Thorough examinations of these situations validated the models’ effectiveness and led to notable decreases in important evaluation parameters, such as the RMSE, MSE, MAE and R2. Regarding excellent accuracy in predicting and overall performance, our CNN-GRU hybrid deep learning model outperformed the other models in the present research.
- 6 March 2025
- Published by Horizon E-Publishing Group in Plant Science Today
Open AccessA novel approach for predicting net irrigated area in India using hybrid deep learning architectures
Studying irrigation systems is crucial to ensuring efficient freshwater utilization and conservation. This study examines the efficacy of forecasting the net irrigated area for future generations to create a model of prediction that can efficiently exchange water demand. To improve the forecast, we generate a model using two-hybrid deep learning techniques to predict irrigation demands: Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) and Convolutional Neural Network-Gated Recurrent Unit (CNN-GRU). These models effectively capture complex variables from diverse data sources, including rainfall patterns, irrigated area statistics and various irrigation system parameters. The main ideas, noteworthy contributions and crucial quantitative results from our study on net irrigated area projection are outlined in this publication. Our main contribution is the development of unique hybrid deep learning approaches that effectively integrate the CNN-LSTM and CNN-GRU architectures. Better predictions are made possible by the models’ design, which consists of parallel CNN layers that independently interpret certain input features. Thorough examinations of these situations validated the models’ effectiveness and led to notable decreases in important evaluation parameters, such as the RMSE, MSE, MAE and R2. Regarding excellent accuracy in predicting and overall performance, our CNN-GRU hybrid deep learning model outperformed the other models in the present research.