Assessing the Impact of Agricultural Chemicals on Crop Yield and Human Health Using Machine Learning Models: A Case Study of Madhya Pradesh, India

Abstract

The excessive application of chemical fertilizers and pesticides in Indian agriculture - particularly in Madhya Pradesh - has raised critical concerns about long-term impacts on edible crop yield and community health. This paper presents an integrated machine learning (ML) framework that simultaneously predicts crop yield and assesses health risk from agricultural chemical exposure, using multi-source data collected from ten districts of Madhya Pradesh over the period 2010–2023. The dataset combines agricultural records (ICAR/FAO), National Health Mission data, satellite imagery (Sentinel-2), environmental monitoring data, and socioeconomic indicators comprising 1,430 district-year observations with 47 input features. The proposed CNN-LSTM hybrid model achieves 95.1% accuracy and R² = 0.97 for crop yield prediction, significantly outperforming six baseline models. Correlation analysis reveals Pearson r = 0.94 between pesticide use intensity and cancer incidence and r = 0.96 for respiratory disease prevalence across districts. A spatial health risk heatmap identifies Gwalior and Indore as high-risk priority zones. The paper also presents an Integrated Decision Support System (IDSS) that provides district-level optimised recommendations for balancing agricultural productivity with community health protection. Results demonstrate that deep learning ensemble architectures offer a scalable and robust platform for sustainable agricultural health governance.

Introduction

Agriculture in India heavily depends on chemical fertilizers and pesticides to maintain productivity, but this increasing use has led to serious environmental and public health concerns. India is among the largest pesticide producers and users globally, with states like Madhya Pradesh showing a steady rise in chemical fertilizer consumption, which coincides with increasing cases of diseases such as cancer, neurological disorders, respiratory illnesses, and reproductive health issues in farming communities.

The study highlights that traditional statistical methods are not sufficient to capture the complex, non-linear relationship between agrochemical exposure and health outcomes. Therefore, it proposes the use of machine learning (ML) and deep learning techniques to analyze large, multi-source datasets. The main aim is to jointly predict crop yield and health risks, map spatial exposure patterns, and build a decision support system that balances agricultural productivity with public health protection in Madhya Pradesh.

A major part of the research reviews recent literature (2020–2023), which consistently shows links between pesticide and fertilizer use and various health problems, including cancer, neurological damage, endocrine disruption, and respiratory diseases. However, existing studies mostly focus either on crop yield or health impacts separately, not both together in an integrated framework.

The dataset used combines agricultural, health, environmental, socioeconomic, and satellite data from 10 districts of Madhya Pradesh over 2010–2023. After extensive preprocessing (handling missing data, normalization, encoding, and dimensionality reduction), the final dataset includes 1,430 observations with 47 features.

The proposed model is a CNN–LSTM hybrid system that processes time-series agricultural and environmental data. It produces dual outputs: crop yield prediction and health risk classification across multiple disease categories. It is compared with several machine learning models (like Random Forest, SVM, ANN, and standalone LSTM).

Results show that the CNN–LSTM model performs best, achieving about 95.1% accuracy and strong predictive power (R² = 0.97), outperforming all baseline models. It also successfully captures real-world trends in crop yield and identifies correlations between pesticide use and health risks.

Conclusion

This paper presented a comprehensive machine learning framework for simultaneously predicting crop yield and assessing population health risk from agricultural chemical exposure, with application to ten districts of Madhya Pradesh over 2010–2023. The proposed CNN-LSTM hybrid model achieved 95.1% accuracy and R² = 0.97, significantly outperforming five baseline models including standalone LSTM, ANN, Random Forest, SVM, and Linear Regression. Correlation analysis confirmed strong positive associations (r > 0.85, p < 0.001) between pesticide use intensity and all five chronic disease categories studied. Spatial risk mapping identified Gwalior and Indore as high-priority intervention zones. The proposed IDSS demonstrated that meaningful community health improvements are achievable with modest reductions in chemical inputs and targeted pesticide substitutions, with minimal impact on agricultural productivity. The proposed framework provides a scalable, data-driven foundation for sustainable agricultural governance in India. It equips policymakers, agricultural extension officers, and public health agencies with evidence-based tools grounded in district-specific data. Planned future extensions include integration of real-time IoT soil sensor data, individual-level clinical records from NHM cohorts, transformer-based attention architectures, and GNN-based spatial spillover modeling to capture inter-district contamination dynamics.

References

[1] N. Sharma and R. Singhvi, \"Effects of chemical fertilizers and pesticides on human health and environment: a review,\" Int. J. Agric. Environ. Biotechnol., vol. 10, no. 6, pp. 675–679, 2017. [2] Government of India, \"Trend in Consumption of Chemical Fertilizers in India,\" Ministry of Agriculture & Farmers Welfare, New Delhi, 2021. [Online]. Available: https://factly.in/data-chemical-fertilizer-consumption [3] E. Elbasi and C. Zaki, \"Crop prediction model using machine learning algorithms,\" Appl. Sci., vol. 13, no. 16, p. 9288, Aug. 2023, doi: 10.3390/app13169288. [4] P. Sharma et al., \"Fertilizer use and its impact on groundwater contamination in Indian agricultural regions,\" Indian J. Environ. Sci., 2023. [5] S. Verma et al., \"Chronic health impacts of pesticide exposure among Indian farmers,\" Indian J. Occup. Health, 2022. [6] L. Zhang and C. Yan, \"The impact of agricultural chemical inputs on environment: global evidence from informetrics analysis,\" Int. J. Low-Carbon Technol., doi: 10.1093/ijlct/cty039, 2018. [7] L. Li et al., \"Developing machine learning models with multi-source environmental data to predict wheat yield in China,\" Comput. Electron. Agric., vol. 194, p. 106790, 2022. [8] European Food Safety Authority (EFSA), \"Pesticide residues in food — annual report 2020,\" EFSA J., 2020. [9] R. Katarya and A. Raturi, \"Impact of machine learning techniques in precision agriculture,\" in Proc. ICETCE 2020, IEEE, doi: 10.1109/ICETCE. [10] P. Krasilnikov, M. A. Taboada, and Amanullah, \"Fertilizer use, soil health and agricultural sustainability,\" Agriculture, vol. 12, p. 462, 2022, doi: 10.3390/agriculture12040462. [11] M. Yousaf et al., \"Effects of fertilization on crop production under rice-oilseed rape rotation,\" Sci. Rep., vol. 7, p. 1270, 2017. [12] R. Kumar et al., \"Crop selection method to maximize crop yield rate using machine learning,\" in Proc. ICSTM 2015, pp. 138–145, doi: 10.1109/ICSTM.2015.7225403. [13] C. Patil et al., \"Pesticide exposure and child development in farming communities: a cohort study,\" J. Pediatr. Health, 2021. [14] S. Singh et al., \"Pesticide exposure and cancer risk in rural Indian populations,\" Cancer Epidemiol., 2021. [15] P. Agarwal et al., \"Reproductive health effects of pesticide exposure among rural women in India,\" Reprod. Toxicol., 2021. [16] V. Reddy et al., \"Fertilizer runoff, eutrophication and its effect on community health,\" Water Res., 2021. [17] IEEE Xplore, \"Identifying the type of chronic kidney disease based on heavy metals in soil using ANN,\" ieeexplore.ieee.org/document/9667974, 2022. [18] ScienceDirect, \"Site-specific fertilizer recommendation using data-driven machine learning for wheat productivity,\" Field Crops Res., 2024. [19] A. Gehlot et al., \"Technical analysis of crop production prediction using machine learning and deep learning,\" in Proc. ICSES, 2022, pp. 1–5. [20] P. Reddy et al., \"Machine learning approaches for predicting crop yield under varying chemical inputs,\" Comput. Electron. Agric., 2023.

Copyright

Copyright © 2026 Jayram Dwivedi, Sitesh Kumar Sinha, Amureesh Kumar Saxena. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.