MétaCan
Menu
Back to cohort
Record W7116877000 · doi:10.1016/j.fcr.2025.110311

Utilizing soil characteristics and hybrid machine learning for interpretable potato yield prediction: A study with satin-bowerbird optimization and deep neural network

2025· article· en· W7116877000 on OpenAlex
Masoud Karbasi, Gurjit S. Randhawa, Aitazaz A. Farooque, Mumtaz Ali, Mehdi Jamei, Khabat Khosravi, Hassan Afzal, Anurag Malik, Qamar U. Zaman

Why this work is in the frame

A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.

affAt least one author lists a Canadian institution in the pinned OpenAlex snapshot.
fundA Canadian funder is recorded on the work.
aboutThe title or abstract carries a Canadian signal from the geographic lexicon.

Bibliographic record

VenueField Crops Research · 2025
Typearticle
Languageen
FieldAgricultural and Biological Sciences
TopicSmart Agriculture and AI
Canadian institutionsDalhousie UniversityUniversity of GuelphUniversity of Prince Edward Island
FundersNatural Sciences and Engineering Research Council of CanadaUniversity of Prince Edward Island
KeywordsInterpretabilityArtificial neural networkFeature selectionSupport vector machineMean squared errorRegressionSelection (genetic algorithm)Correlation coefficientFeature (linguistics)Predictive modelling

Abstract

fetched live from OpenAlex

Context Yield forecasting is crucial to the agricultural planning enterprise, such as input control, farm logistics and reduction of economic risks. The soils in the Maritime provinces of Canada have a great difference in their properties which affect the productivity of crops. Such variability requires a strong prediction model that could address the different characteristics of soil. Objective This research proposal is expected to establish a stable potato yield prediction model based on the soil property data of New Brunswick and Prince Edward Island and determine whether the application of optimization techniques with deep learning can enhance the prediction accuracy over the conventional machine learning approach. Methods Soil samples were taken at eight experimental sites in the 2017 and 2018 growing seasons, with 18 soil properties being captured. The feature selection techniques were used to create three input scenarios (Comb1, Comb2, Comb3). To optimize the selection of input variables, a hybrid prediction model, DNN-SBO (Deep Neural Network -Satin Bowerbird Optimization), was suggested and refined with the Boruta feature selection and Best Subset Regression-WASPAS. The performance of the model was tested in comparison with Kernel Ridge Regression (KRR), Elastic Net, K-Nearest Neighbors (KNN) and Support Vector Regression (SVR), on the evaluation metrics of Correlation Coefficient (R), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The model interpretability was done using SHAP (Shapley Additive exPlanation) analysis. Results and Conclusions Comb2 was the best input scenario that consisted of Total Base Saturation, Sulfur, Magnesium, Potash, Aluminum, Zinc, Phosphate, Manganese, Organic Matter, Iron, and Copper. DNN-SBO model had the best predictive power with R= 0.903 (train) and RMSE= 4.165 t/ha and MAPE= 6.766 % and R= 0.853(test) and RMSE= 5.522 t/ha and MAPE= 9.707 %. The SHAP analysis has shown that Iron was the most significant predictor (mean SHAP = +5.49), next was Copper, Zinc, Phosphorus, and Organic Matter. Significance The paper sheds light on the promise of deep learning that is based on bio-inspired optimization and feature selection techniques in order to achieve a significant increase in crop yield prediction. The findings can lead to the wider use of the similar methods in precision agriculture, which will result in smarter and data-driven farming in variably soiled areas.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.000
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Observational · Consensus signal: Observational
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.358
Threshold uncertainty score0.471

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0000.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0000.000
Science and technology studies0.0010.000
Scholarly communication0.0000.000
Open science0.0000.000
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.027
GPT teacher head0.274
Teacher spread0.247 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it