Prediction of Remaining Productive Life of Oil Palm Plantation Soil Using Support Vector Regression with Permutation-Based Feature Importance Analysis

Zara Yunizar Zainal; Nurdin Nurdin; Zharif Athaya Andarfi

doi:10.30871/jaic.v10i4.13229

Authors

Zara Yunizar Zainal Universitas Malikussaleh
Nurdin Nurdin Universitas Malikussaleh
Zharif Athaya Andarfi Universitas Malikussaleh

DOI:

https://doi.org/10.30871/jaic.v10i4.13229

Keywords:

Support Vector Regression (SVR), Remaining Productive Life (RPL), Oil Palm Plantation Soil, Permutation-Based Feature Importance, Soil Productivity Prediction

Abstract

Oil palm (Elaeis guineensis Jacq.) is a strategically vital crop in Southeast Asia, yet progressive soil degradation driven by prolonged monoculture, pathogen pressure, and intensive land use poses a critical threat to long-term plantation sustainability. Existing soil assessment methods deliver static fertility classifications without quantifying the remaining productive lifespan of a given plot. This study introduces Remaining Productive Life (RPL) as a novel regression target defined as the estimated number of years before plantation soil productivity falls below a critical economic threshold. A Support Vector Regression (SVR) model with Radial Basis Function (RBF) kernel, formulated as K(xi, xj) = exp(−γ‖xi − xj‖²) with γ = 0.0303 (= 1/n_features) and regularization parameter λ = 0.5, was applied to a realistic synthetic multi-year dataset comprising 14,400 observations across 800 plantation plots spanning five soil types (Ultisol, Inceptisol, Alfisol, Peat, Oxisol) and the period 2008–2025. Thirty-three soil physicochemical, biological, management, and economic indicators constituted the input feature set. The SVR model achieved R² = 0.9141, MAE = 0.5122 years, RMSE = 1.1026 years, and MAPE = 11.883% on the independent test set, with 92.0% of predictions yielding an absolute error below one year. Permutation-based feature importance analysis identified Degradation Rate (ΔMAE = 0.2864), Plantation Age (0.2560), and Soil Productivity Index (0.2530) as the three dominant predictors, while Ganoderma Risk Index ranked fourth (0.2138), revealing the pivotal contribution of biological soil health to long-term productivity prediction. These findings establish SVR-based RPL estimation as an effective, interpretable framework for precision plantation management and proactive soil sustainability planning.

Downloads

Download data is not yet available.

References

[1] X. Jia, Y. Fang, B. Hu, B. Yu, and Y. Zhou, “Development of Soil Fertility Index Using Machine Learning and Visible-Near-Infrared Spectroscopy,” Land, vol. 12, no. 12, 2023, doi: 10.3390/land12122155.

[2] J. Ratai et al., “Tropical peat soil changes across successive oil palm generations in Sarawak, Malaysia,” Heliyon, vol. 10, no. 18, p. e37754, 2024, doi: 10.1016/j.heliyon.2024.e37754.

[3] Y. Ang and Y. P. Lee, “Oil Palm Yield Prediction Across Blocks Using Multi-Source Data and Machine Learning,” Res. Sq., 2022, [Online]. Available: https://www.researchsquare.com/article/rs-1938493/v1?utm_source=researcher_app&utm_medium=referral&utm_campaign=RESR_MRKT_Researcher_inbound

[4] F. Setiajiati, D. R. Nurrochmat, B. W. Van Assen, and H. Purwawangsa, “Current status of Indonesia’s palm oil products and their competitiveness in the global market,” IOP Conf. Ser. Earth Environ. Sci., vol. 1379, no. 1, 2024, doi: 10.1088/1755-1315/1379/1/012022.

[5] I. A. H. Yousif et al., “Soil quality indicator-based land productivity modelling for agricultural sustainability,” PLoS One, vol. 20, no. 1, pp. 1–28, 2025, doi: 10.1371/journal.pone.0316840.

[6] Y. H. Haw et al., “Detection of Basal Stem Rot Disease Using Deep Learning,” IEEE Access, vol. 11, pp. 49846–49862, 2023, doi: 10.1109/ACCESS.2023.3276763.

[7] Y. Huang, R. Srivastava, C. Ngo, J. Gao, J. Wu, and S. Chiao, “Data-Driven Soil Analysis and Evaluation for Smart Farming Using Machine Learning Approaches,” Agric., vol. 13, no. 9, 2023, doi: 10.3390/agriculture13091777.

[8] Z. Zhu and Y. Xie, “Research on Evaluation Methods of Black Soil Farmland Productivity Based on Field Block Scale,” Appl. Sci., vol. 14, no. 7, 2024, doi: 10.3390/app14073130.

[9] M. Awais et al., “AI and machine learning for soil analysis: an assessment of sustainable agricultural practices,” Bioresour. Bioprocess., vol. 10, no. 1, 2023, doi: 10.1186/s40643-023-00710-y.

[10] E. Harsányi et al., “Assessment of Advanced Machine and Deep Learning Approaches for Predicting CO2 Emissions from Agricultural Lands: Insights Across Diverse Agroclimatic Zones,” Earth Syst. Environ., vol. 8, no. 4, pp. 1109–1125, 2024, doi: 10.1007/s41748-024-00424-x.

[11] N. A. Butt et al., “The Development of Intelligent Agents: A Case-Based Reasoning Approach to Achieve Human-Like Peculiarities via Playback of Human Traces,” IEEE Access, vol. 11, no. May, pp. 78693–78712, 2023, doi: 10.1109/ACCESS.2023.3274740.

[12] D. Fiantis et al., “Mapping peat thickness and carbon stock of a degraded peatland in West Sumatra, Indonesia,” Soil Use Manag., vol. 40, no. 1, pp. 1–13, 2024, doi: 10.1111/sum.12954.

[13] A. Fisher, C. Rudin, and F. Dominici, “All models are wrong, but many are useful: Learning a variable’s importance by studying an entire class of prediction models simultaneously,” J. Mach. Learn. Res., vol. 20, no. Vi, pp. 1–88, 2019.

[14] O. Folorunso et al., “Exploring Machine Learning Models for Soil Nutrient Properties Prediction: A Systematic Review,” Big Data Cogn. Comput., vol. 7, no. 2, 2023, doi: 10.3390/bdcc7020113.

[15] M. Khaleghian, “Organic Matter Prediction,” 2024.

[16] F. N. Thabit, O. I. A. Negim, M. A. E. AbdelRahman, A. Scopa, and A. R. A. Moursy, “Using Various Models for Predicting Soil Organic Carbon Based on DRIFT-FTIR and Chemical Analysis,” Soil Syst., vol. 8, no. 1, 2024, doi: 10.3390/soilsystems8010022.

[17] F. Wu, Q. Wu, Y. Tan, and X. Xu, “Remaining Useful Life Prediction Based on Deep Learning: A Survey,” Sensors, vol. 24, no. 11, pp. 1–30, 2024, doi: 10.3390/s24113454.

[18] M. Pérez-Sato et al., “Soil physicochemical properties change by age of the oil palm crop,” Heliyon, vol. 9, no. 6, 2023, doi: 10.1016/j.heliyon.2023.e16302.