Predictive Analytics of Varieties of Potatoes

TL;DR

This paper treats potato clone selection as a binary classification task to accelerate breeding progress in Oregon’s multi-year trials. It systematically compares nonlinear models (Neural Net, HGBC, SVM) with and without data imputation, employing forward feature selection and a comprehensive simulation ADEMP framework to assess robustness across varied data-generating mechanisms, using metrics like $AUC$-$ROC$ and $MCC$ to handle class imbalance. Key findings show that non-linear methods generally outperform linear baselines, with SVM offering the most robust generalization across simulated distributions, and forward feature selection identifying yield, quality traits, and trial/environmental variables as primary drivers. Practically, the approach offers time and cost savings by improving early decisions on clone advancement while complementing breeder expertise and enabling scalable, data-informed selection in complex, multi-location breeding programs. The work highlights that careful imputation, feature engineering, and model choice are crucial for reliable predictions in agricultural trials, and it calls for ongoing model validation as new trial data accumulate.

Abstract

We explore the application of machine learning algorithms specifically to enhance the selection process of Russet potato clones in breeding trials by predicting their suitability for advancement. This study addresses the challenge of efficiently identifying high-yield, disease-resistant, and climate-resilient potato varieties that meet processing industry standards. Leveraging manually collected data from trials in the state of Oregon, we investigate the potential of a wide variety of state-of-the-art binary classification models. The dataset includes 1086 clones, with data on 38 attributes recorded for each clone, focusing on yield, size, appearance, and frying characteristics, with several control varieties planted consistently across four Oregon regions from 2013-2021. We conduct a comprehensive analysis of the dataset that includes preprocessing, feature engineering, and imputation to address missing values. We focus on several key metrics such as accuracy, F1-score, and Matthews correlation coefficient (MCC) for model evaluation. The top-performing models, namely a neural network classifier (Neural Net), histogram-based gradient boosting classifier (HGBC), and a support vector machine classifier (SVM), demonstrate consistent and significant results. To further validate our findings, we conduct a simulation study. By simulating different data-generating scenarios, we assess model robustness and performance through true positive, true negative, false positive, and false negative distributions, area under the receiver operating characteristic curve (AUC-ROC) and MCC. The simulation results highlight that non-linear models like SVM and HGBC consistently show higher AUC-ROC and MCC than logistic regression (LR), thus outperforming the traditional linear model across various distributions, and emphasizing the importance of model selection and tuning in agricultural trials.

Figures (2)

• Figure 1: US State of Oregon. Points represent the cities in which the potato trials were conducted: Corvallis, Hermiston, Ontario, and Klamath Falls. From each of these states, russet potatoes such as the one depicted at the center of the map potato_pic were collected and data of each clone's characteristics were recorded for analysis.
• Figure 3: Confusion matrices for the top performing models applied to both imputed and non-imputed datasets: feedforward neural network classifier (Neural Net) in first row, histogram-based gradient boosting classifier (HGBC) in middle row, and support vector machine (SVM) classifier in the last row. Variable selection was not performed here, however variables containing at least 400 NA's were dropped before training. Numerical values within the confusion matrices represent prediction proportions.