Neglecting spatial autocorrelation causes underestimation of the error of sugarcane yield models

Ferraciolli, Matheus Agostini; Bocca, Felipe Ferreira; Rodrigues, Luiz Henrique Antunes

doi:10.1016/j.compag.2018.09.003

Cited by 15 publications

(7 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As a whole, KNN is the method that has the highest significant effect. KNN is one of the simplest algorithms for looking at the nearest neighbor value [71], even though KNN is considered a poor test on the IRIS dataset [72].…”

Section: Effect Of Synthetic Data On ML Classifier Performance Using ...mentioning

confidence: 99%

Application of the Synthetic Over-Sampling Method to Increase the Sensitivity of Algorithm Classification for Class Imbalance in Small Spatial Datasets

2022

IJIES

View full text Add to dashboard Cite

The difficulty of acquiring data from numerous intergovernmental agencies/institutions for prone road traffic accidents (PRTA) spatial datasets produces a small-scale dataset that causes dataset imbalance. Class imbalance in small-scale datasets causes uncertainty in the results of the modeling PRTA classification. The proposed research is a scenario-based case representation model on the pre-processing data stage to increase the sensitivity of algorithm classification in a small-scale dataset that causes dataset imbalance using machine learning (ML), the synthetic oversampling method. The retrieval of attributes from the spatial dataset is transformed into the raw dataset, the normalized dataset, the synthetic minority over-sampling technique (SMOTE) raw dataset, and SMOTE normalized dataset scenarios. Balancing datasets using four variants of SMOTE, namely ADASYN, Borderline-SMOTE, K-Means SMOTE, and SVM-SMOTE resampled. To evaluate how well the PRTA classification model performed, we utilized the hyper-parameters optimization technique and the genetic algorithm (GA) search cross-validation. The experiment was run with the ML classifier method, including the k-nearest neighbor (KNN), support vector machines (SVM), multilayer perceptron (MLP), naive bayes (NB), logistic regression (LR), and random forest (RF). The Area Under Curve (AUC) was used to evaluate the results of the experiments. The results of the dataset test in a predetermined scenario conclude that a single algorithm that is computationally light to produce an optimal classifier tends to use a raw dataset that is balanced using SMOTE. The KNN method as a single algorithm for classification based on the distance between samples is superior, with an AUC value of 0.89, which is included in the good classification category of all ML classifiers proposed to handle small data sets imbalanced classes using SMOTE raw datasets for K-Means variants SMOTE.

show abstract

Section: Effect Of Synthetic Data On ML Classifier Performance Using ...mentioning

confidence: 99%

Application of the Synthetic Over-Sampling Method to Increase the Sensitivity of Algorithm Classification for Class Imbalance in Small Spatial Datasets

2022

IJIES

View full text Add to dashboard Cite

show abstract

“…Additional preprocessing referred to avoiding the effect of spatial autocorrelation, described by Ferraciolli et al (2018). This effect means that fields that are close to each other share characteristics, which could compromise the independence of training and test sets in the cases in which close fields are present in both sets.…”

Section: Pre-processingmentioning

confidence: 99%

How good are the models available for estimating sugar content in sugarcane?

Oliveira

Rodrigues

2020

European Journal of Agronomy

View full text Add to dashboard Cite

“…In most of these studies, ML reduced prediction errors. The presence of spatial autocorrelation (SAC) in explanatory or predictive variables, however, has recently been suggested to induce the overfitting or the underestimation of the ML model (Rocha et al, 2018; Ferraciolli, Bocca, and Rodrigues, 2019; Sinha et al, 2019). This highlights that spatial dependency still has possible influences even on the ML models (Nikparvar and Thill, 2021).…”

Section: Introductionmentioning

confidence: 99%

Three Common Machine Learning Algorithms Neither Enhance Prediction Accuracy Nor Reduce Spatial Autocorrelation in Residuals: An Analysis of Twenty‐five Socioeconomic Data Sets

Song

Kim

2022

Geographical Analysis

View full text Add to dashboard Cite

Machine learning (ML) is being applied in an increasing volume of geographical research. However, the aspects of spatial autocorrelation (SAC) in the residuals produced by ML models have been understudied compared to the benefit of ML, namely, reduction of prediction errors. In this study, we examined the relationship between predictive accuracy and the reduction in the residual SAC for 597 variables from 25 geographical socio‐economic data sets using spatial and nonspatial cross‐validation of three ML algorithms such as random forests, support vector machine, and artificial neural network (ANN) to provide an extensive empirical diagnosis—but not a definitive theory—of the relationship between SAC and ML. Our results highlighted that the ML algorithms with tuned hyperparameters yielded marginal predictive accuracy gains and the minimal decreases in residual SAC. ANN revealed lower accuracy and higher reduction in the residual SAC than others. This implies ML algorithms in geographical research in socio‐economic domains would not always result in higher prediction accuracy. We suggest that ML in geographical research should be cautiously employed when the main objective is related to the residual SAC. We also showed that spatial cross‐validation neither improves predictive accuracy substantially nor reduce the residual SAC effectively.

show abstract

Neglecting spatial autocorrelation causes underestimation of the error of sugarcane yield models

Cited by 15 publications

References 21 publications

Application of the Synthetic Over-Sampling Method to Increase the Sensitivity of Algorithm Classification for Class Imbalance in Small Spatial Datasets

Application of the Synthetic Over-Sampling Method to Increase the Sensitivity of Algorithm Classification for Class Imbalance in Small Spatial Datasets

How good are the models available for estimating sugar content in sugarcane?

Three Common Machine Learning Algorithms Neither Enhance Prediction Accuracy Nor Reduce Spatial Autocorrelation in Residuals: An Analysis of Twenty‐five Socioeconomic Data Sets

Contact Info

Product

Resources

About