Multi‐class classification using kernel density estimation on K ‐nearest neighbours

2016 ◽  
Vol 52 (8) ◽  
pp. 600-602 ◽  
Author(s):  
Xiaofeng Tang ◽  
Aiqiang Xu
Keyword(s):  
Density Estimation ◽  
Kernel Density Estimation ◽  
Kernel Density ◽  
Nearest Neighbours ◽  
Multi Class Classification

scholarly journals Ther varom mid j hia: Tracing linguistic diffusion in the history of Norwegian using kernel density estimation

2019 ◽  
Vol 27 (1) ◽  
pp. 5-34
Author(s):  
Tam Blaxter
Keyword(s):  
Density Estimation ◽  
Middle Ages ◽  
Kernel Density Estimation ◽  
Kernel Density ◽  
Gaussian Kernel ◽  
Historical Sources ◽  
Nearest Neighbours ◽  
Different Types ◽  
Language Data ◽  
History Of

Abstract Tracing the diffusion of linguistic innovations in space from historical sources is challenging. The complexity of the datasets needed in combination with the noisy reality of historical language data mean that it has not been practical until recently. However, bigger historical corpora with richer spatial and temporal information allow us to attempt it. This paper presents an investigation into changes affecting first person non-singular pronouns in the history of Norwegian: first, individual changes affecting the dual (vit > mit) and plural (vér > mér), followed by loss of the dual-plural distinction by merger into either form or replacement of both by Danish-Swedish vi. To create dynamic spatial visualisations of these changes, the use of kernel density estimation is proposed. This term covers a range of statistical tools depending on the kernel function. The paper argues for a Gaussian kernel in time and an adaptive uniform (k-nearest neighbours) kernel in space, allowing uncertainty or multiple localisation to be incorporated into calculations. The results for this dataset allow us to make a link between Modern Norwegian dialectological patterns and language use in the Middle Ages; they also exemplify different types of diffusion process in the spread of linguistic innovations.

scholarly journals Author Correction: An extended Weight Kernel Density Estimation model forecasts COVID-19 onset risk and identifies spatiotemporal variations of lockdown effects in China

2021 ◽  
Vol 4 (1) ◽  
Author(s):  
Wenzhong Shi ◽  
Chengzhuo Tong ◽  
Anshu Zhang ◽  
Bin Wang ◽  
Zhicheng Shi ◽  
...  
Keyword(s):  
Density Estimation ◽  
Kernel Density Estimation ◽  
Kernel Density ◽  
Spatiotemporal Variations ◽  
Estimation Model

A Correction to this paper has been published: https://doi.org/10.1038/s42003-021-01924-6

scholarly journals Identifying the density of grassland fire points with kernel density estimation based on spatial distribution characteristics

2021 ◽  
Vol 13 (1) ◽  
pp. 796-806
Author(s):  
Zhen Shuo ◽  
Zhang Jingyu ◽  
Zhang Zhengxiang ◽  
Zhao Jianjun
Keyword(s):  
Spatial Distribution ◽  
Density Estimation ◽  
Kernel Density Estimation ◽  
Function Method ◽  
Kernel Density ◽  
Fire Occurrence ◽  
Distribution Characteristics ◽  
Fire Point ◽  
Spatial Distribution Characteristics ◽  
Grassland Fire

Abstract Understanding the risk of grassland fire occurrence associated with historical fire point events is critical for implementing effective management of grasslands. This may require a model to convert the fire point records into continuous spatial distribution data. Kernel density estimation (KDE) can be used to represent the spatial distribution of grassland fire occurrences and decrease the influences historical records in point format with inaccurate positions. The bandwidth is the most important parameter because it dominates the amount of variation in the estimation of KDE. In this study, the spatial distribution characteristic of the points was considered to determine the bandwidth of KDE with the Ripley’s K function method. With high, medium, and low concentration scenes of grassland fire points, kernel density surfaces were produced by using the kernel function with four bandwidth parameter selection methods. For acquiring the best maps, the estimated density surfaces were compared by mean integrated squared error methods. The results show that Ripley’s K function method is the best bandwidth selection method for mapping and analyzing the risk of grassland fire occurrence with the dependent or inaccurate point variable, considering the spatial distribution characteristics.

Sign in / Sign up

Export Citation Format

Share Document