The<i>k</i>-nearest neighbor technique with local linear regression

Magnussen, Steen; Tomppo, Erkki

doi:10.1080/02827581.2013.878744

Cited by 12 publications

(2 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zhang (2012) changed the distance measure to grey distance and found its advantage in capturing the proximity relationship. Magnussen and Tomppo (2014) calibrated KNN imputation with local linear regression in the context of forest science. The new technique presented improved correlation between imputation and its real value.…”

Section: Knn Imputation Improvementmentioning

confidence: 99%

Cross-validation based K nearest neighbor imputation for software quality datasets: An empirical study

Huang

Keung

Sarro

et al. 2017

Journal of Systems and Software

View full text Add to dashboard Cite

Being able to predict software quality is essential, but also it pose significant challenges in software engineering. Historical software project datasets are often being utilized together with various machine learning algorithms for fault-proneness classification.Unfortunately, the missing values in datasets have negative impacts on the estimation accuracy and therefore, could lead to inconsistent results. As a method handling missing data, K nearest neighbor (KNN) imputation gradually gains acceptance in empirical studies by its exemplary performance and simplicity. To date, researchers still call for optimized parameter setting for KNN imputation to further improve its performance. In the work, we develop a novel incomplete-instance based KNN imputation technique, which utilizes a cross-validation scheme to optimize the parameters for each missing value. An experimental assessment is conducted on eight quality datasets under various missingness scenarios. The study also compared the proposed imputation approach with mean imputation and other three KNN imputation approaches. The results show that our proposed approach is superior to others in general. The relatively optimal fixed parameter settings for KNN imputation for software quality data is also 2 determined. It is observed that the classification accuracy is improved or at least maintained by using our approach for missing data imputation.

show abstract

Section: Knn Imputation Improvementmentioning

confidence: 99%

Cross-validation based K nearest neighbor imputation for software quality datasets: An empirical study

Huang

Keung

Sarro

et al. 2017

Journal of Systems and Software

View full text Add to dashboard Cite

show abstract

“…A subset of the elements in N have paired observations of both X and Y and are referred to as the reference dataset. Imputation for a target element is estimated as a function of k Y values in the reference dataset that have X values that are closest, using a measure of statistical proximity, to the X values of the target element (Magnussen and Tomppo 2014).…”

Section: Non-parametric Modellingmentioning

confidence: 99%

Multi-sensor modelling of a forest productivity index for radiata pine plantations

Watt

Dash

Watt³

et al. 2016

N.Z. j. of For. Sci.

View full text Add to dashboard Cite

Background: An understanding of how plantation productivity varies spatially is important for forest planning, management and projection of future plantation yields and returns. The 300 Index is a volume productivity index developed for Pinus radiata D.Don that has been widely used within New Zealand to assess site productivity. Although the 300 Index is routinely characterised at the stand level, little research has investigated if remotely sensed data sources can be used in combination with environmental layers to precisely predict this metric at fine spatial resolution. Methods: This study uses an extensive dataset obtained from P. radiata plantations in the central North Island, New Zealand. Using this dataset, the objective of this research was to compare the precision of parametric and non-parametric models of the 300 Index that included explanatory variables extracted from aerially acquired light detection and ranging (LiDAR), satellite imagery (RapidEye) at 5-m resolution or environmental layers and combinations of these three data sources. Models were constructed both with and without stand age as an explanatory variable as managers may not always have access to stand age. A total of 28 models (14 data sources × two model methods) were constructed using data from 433 plots. Precision and bias of these models was determined using an independent dataset of 60 plots. Results: Of the non-parametric methods tested (k-most similar neighbour (k-MSN), k-nearest neighbour (k-NN)), k-NN using an optimised value of k-most precisely predicted the 300 Index for 11 of the 14 constructed models. The use of k-NN was found to be more precise than parametric models when age was not available but of overall similar precision to parametric models when stand age was available as a predictor. For models including stand age, the inclusion of LiDAR resulted in the most precise model (mean R 2 = 0.789; root mean square error (RMSE) = 2.48 m 3 ha −1 year

show abstract

Machine Learning at the Edge

Levy

Naiser

2019

Software Engineering for Embedded Systems

View full text Add to dashboard Cite

Thek-nearest neighbor technique with local linear regression

Cited by 12 publications

References 56 publications

Cross-validation based K nearest neighbor imputation for software quality datasets: An empirical study

Cross-validation based K nearest neighbor imputation for software quality datasets: An empirical study

Multi-sensor modelling of a forest productivity index for radiata pine plantations

Machine Learning at the Edge

Contact Info

Product

Resources

About