Kth-nearest Neighbor Research Articles

Clustering in point processes on linear networks using nearest neighbour volumes

This study introduces a novel method specifically designed to detect clusters of points within linear networks. This method extends a classification approach used for point processes in spatial contexts. Unlike traditional methods that operate on planar spaces, our approach adapts to the unique geometric challenges of linear networks, where classical properties of point processes are altered, and intuitive data visualisation becomes more complex. Our method utilises the distribution of the Kth nearest neighbour volumes, extending planar-based clustering techniques to identify regions of increased point density within a network. This approach is particularly effective for distinguishing overlapping Poisson processes within the same linear network. We demonstrate the practical utility of our method through applications to road traffic accident data from two Colombian cities, Bogota and Medellin. Our results reveal distinct clusters of high-density points in road segments where severe traffic accidents (resulting in injuries or fatalities) are most likely to occur, highlighting areas of increased risk. These clusters were primarily located on major arterial roads with high traffic volumes. In contrast, low-density points corresponded to areas with fewer accidents, likely due to lower traffic flow or other mitigating factors. Our findings provide valuable insights for urban planning and road safety management.

Marine remote target signal extraction based on 128 line-array single photon LiDAR

LiDAR technology has garnered significant attention in recent years due to its superior directivity, high resolution, and precise 3D information acquisition capabilities, making it indispensable in navigation systems. Among its variants, single-photon LiDAR stands out for maritime applications, owing to its reduced power consumption and extended detection range. However, the high sensitivity of single-photon detectors often results in substantial noise, necessitating effective denoising before data can be utilized for identification, tracking, and other purposes. In this study, we present a novel 128-line, 1550 nm shipborne long-range single-photon LiDAR system, with data collected and analyzed in maritime environments. This system contends with challenges such as a large dynamic range, abundant noise photons, and the complexity of sea surface conditions. To address these issues, we propose an efficient and adaptive denoising algorithm based on the k-th nearest neighbor (KNN) methodology. By examining the distribution characteristics of signal and noise photons, our approach enables target extraction even under conditions of intense noise and sparse signals. Our method exhibits robust adaptability across various detection scenarios. Experimental evaluations demonstrate its efficacy, accurately identifying targets at distances of 3.2 km in clear weather and 1.6 km in foggy conditions.

A Real-Time Intrusion Detection System for DoS/DDoS Attack Classification in IoT Networks Using KNN-Neural Network Hybrid Technique

As more devices are connected to the Internet through the Internet of Things (IoT), there are huge security challenges. One of the major problems is Distributed Denial of Service (DoS) and Distributed Denial of Service (DDoS) attacks. These attacks floods networks with useless traffic, disrupting IoT services. There is a need for better security measures to handle this. Intrusion Detection Systems (IDS) is used to find suspicious activities, but many of them can't keep up with new types of attacks in real time. This study focuses on creating an efficient real time hybrid framework that uses the Kth Nearest Neighbor (KNN) algorithm and dense neural networks. This proposed framework aims to identify and categorize DoS/DDoS attacks in real time through the utilization of a simulation model and MQTT-IoT-IDS2020 dataset and compared with existing frameworks, our proposed framework excels in accuracy, precision, and recall.

The screening of cannabis addiction using machine learning, MoCA, and anxiety/depression tests

A new cannabis addiction screening approach is developed based on the use of Machine Learning (ML) alongside with psychological and cognitive assessment tests. The Hospital Anxiety and Depression (HAD) test is used to determine the Absence of Anxiety and Depression “No Disorder” or the presence of “Anxiety Only”, “Depression Only”, and both “Anxiety and Depression”. These features are considered to evaluate the psychological effects of cannabis use. In addition to that, the Montreal Cognitive Assessment (MoCA) test is chosen as a feature to assess the cognitive state of the participants. Also, the Age of First Cannabis Use (AFCU) is included as a feature to incorporate participants with no prior cannabis use alongside with those who have used cannabis before. The portion of participants with prior experience consuming cannabis comprises a subset containing passive users and a subset of users that have developed cannabis addiction. This 6-month study was conducted in Marrakech, Morocco in a center affiliated to the National Association of Drug-Risk Reduction of Morocco known as RdR-Maroc with the participation of 146 subjects. The participants in this study are grouped into two groups of 73 participants each. Both case and control groups were clinically assessed by a professional. In this paper, the classical statistical data analysis for the AFCU (M=15.6; SD=2.4), MoCA (M=25.0; SD=2.4), and HAD (58 cases with no disorder, 88 cases with disorder) failed to separate between the two classes. Also, the feature selection for modeling is performed using Spearman correlation, Kendall correlation, and the Odds Ratio (OR). These statistics showed that the selected features were associated to addiction except for “Anxiety Only”. Moreover, ML models such as Logistic Regression (LR), Support Vector Machine (SVM), Artificial Neural Network (ANN), kth-Nearest Neighbor (KNN), and Decision Tree (DT) were used to classify participants into two categories (Addict vs. Non-Addict) based on the selected features mentioned earlier. The Receiver Operating Characteristic (ROC) curve is used to overview the quality of the models and showed that they all captured the nonlinearities underlined in the data. On the other hand, the Area Under the Curve (AUC) allowed the ranking of these models and indicated that the SVM model outperformed all other models with an AUC of 0.97. After that, the SVM model was compared to the Cannabis Use Disorder (CUD) based screening method issued by the fifth edition of the Diagnostic and Statistical Manual of mental disorders (DSM-5). The SVM model's accuracy of 96.6 % outperformed the CUD questionnaires accuracy of 87.7 %. The validity of the proposed approach was investigated where the SVM model presented high validity characteristics (sensitivity=1, specificity=0.93) compared to the CUD questionnaire (sensitivity=0.75, specificity=1). Hence, the proposed approach provided satisfactory results compared to current adopted CUD based screening test. Our findings expand current research on the development of simple and powerful tools for cannabis addiction screening with the purpose of early detection in order to establish preventive and therapeutic strategies that could potentially mitigate the prevalence of cannabis addiction.

Use of the k-nearest neighbour and its analysis for fall detection on Systems on a Chip for multiple datasets

Fall of an elderly person often leads to serious injuries and death. Many falls occur in the home environment, and hence a reliable fall detection system that can raise alarms with minimum latency is a necessity. Wrist-worn accelerometer-based fall detection systems and multiple datasets are available, but no attempt has been made to analyze the accuracy and precision. Wherever the comparison does exist, it has been run on a cloud. No analysis of the models, convergence, and dataset analysis on Systems on a Chip (SoCs) has ever been attempted. In this paper, we attempt to present why Machine Learning (ML) algorithms in their current state cannot be run on existing SoCs. We have used Snapdragon 410c SoC to do our analytics. In this paper, we have used the kth-nearest neighbour to prove that ML cannot be directly run on SoCs. We have looked at the effect of distance metrics and neighbors as well as the effect of feature extraction on the accuracies and the latencies. In this paper, we establish the need for model compression and data pruning for fall detection using ML/Deep Learning algorithms on SoCs. We have done this by analyzing various datasets on varying architectural parameters.

Performance Analysis of Different Machine Learning Algorithms on Credit Card Fraud Detection

Machine learning (ML) is a logical investigation of various algorithms and factual models that PCs utilize to carry out particular operations that are not clearly programmed. This paper aims to statistically analyze different machine learning algorithms, and compare and contrast their performance for credit card fraud detection. Algorithms used are Artificial Neural Networks(ANN), Sample Vector Machine (SVM), and Kth Nearest Neighbour (KNN), Decision Tree, Logistic Regression and Random Forest. All these above mentioned algorithms are compared on basis of performance measures. It is deduced that the random forest algorithm is the best algorithm.

Large nearest neighbour balls in hyperbolic stochastic geometry

Consider a stationary Poisson process in a d-dimensional hyperbolic space. For R>0 define the point process xi _R^{(k)} of exceedance heights over a suitable threshold of the hyperbolic volumes of kth nearest neighbour balls centred around the points of the Poisson process within a hyperbolic ball of radius R centred at a fixed point. The point process xi _R^{(k)} is compared to an inhomogeneous Poisson process on the real line with intensity function e^{-u} and point process convergence in the Kantorovich-Rubinstein distance is shown. From this, a quantitative limit theorem for the hyperbolic maximum kth nearest neighbour ball with a limiting Gumbel distribution is derived.

Developing AI/ML Based Predictive Capabilities for a Compression Ignition Engine Using Pseudo Dynamometer Data

Energy and power demands for military operations continue to rise as autonomous air, land, and sea platforms are developed and deployed with increasingly energetic weapon systems. The primary limiting capability hindering full integration of such systems is the need to effectively and efficiently manage, generate, and transmit energy across the battlefield. Energy efficiency is primarily dictated by the number of dissimilar energy conversion processes in the system. After combustion, a Compression Ignition (CI) engine must periodically continue to inject fuel to produce mechanical energy, simultaneously generating thermal, acoustic, and fluid energy (in the form of unburnt hydrocarbons, engine coolant, and engine oil). In this paper, we present multiple sets of Shallow Artificial Neural Networks (SANNs), Convolutional Neural Network (CNNs), and K-th Nearest Neighbor (KNN) classifiers, capable of approximating the in-cylinder conditions and informing future optimization and control efforts. The neural networks provide outstanding predictive capabilities of the variables of interest and improve understanding of the energy and power management of a CI engine, leading to improved awareness, efficiency, and resilience at the device and system level.

A deep neural network-based approach for fake news detection in regional language

Purpose The current natural language processing algorithms are still lacking in judgment criteria, and these approaches often require deep knowledge of political or social contexts. Seeing the damage done by the spreading of fake news in various sectors have attracted the attention of several low-level regional communities. However, such methods are widely developed for English language and low-resource languages remain unfocused. This study aims to provide analysis of Hindi fake news and develop a referral system with advanced techniques to identify fake news in Hindi. Design/methodology/approach The technique deployed in this model uses bidirectional long short-term memory (B-LSTM) as compared with other models like naïve bayes, logistic regression, random forest, support vector machine, decision tree classifier, kth nearest neighbor, gated recurrent unit and long short-term models. Findings The deep learning model such as B-LSTM yields an accuracy of 95.01%. Originality/value This study anticipates that this model will be a beneficial resource for building technologies to prevent the spreading of fake news and contribute to research with low resource languages.

Limit laws for large th-nearest neighbor balls

AbstractLet $X_1,X_2, \ldots, X_n$ be a sequence of independent random points in $\mathbb{R}^d$ with common Lebesgue density f. Under some conditions on f, we obtain a Poisson limit theorem, as $n \to \infty$ , for the number of large probability kth-nearest neighbor balls of $X_1,\ldots, X_n$ . Our result generalizes Theorem 2.2 of [11], which refers to the special case $k=1$ . Our proof is completely different since it employs the Chen–Stein method instead of the method of moments. Moreover, we obtain a rate of convergence for the Poisson approximation.

Doubt on the statistical significance of the Giant GRB Ring

AbstractBalázs et al. discovered a gigaparsec‐scale ring‐like cluster of gamma‐ray bursts (GRBs) at redshift , the so‐called Giant GRB Ring (GGR), and claimed that the existence of such a gigantic structure challenges the standard cosmological paradigm. We show that the kth nearest neighbor analysis of Balázs et al., by which they estimated the probability of finding by chance a GRB cluster comparable to the GGR to be only 0.05%, has several flaws in the statistical reasoning. After correcting these flaws, the above probability becomes 5–20%, considerably higher than the original estimate and large enough to discard the hypothesis that the GGR has a physical origin. Moreover, this probability is even higher, 60–90%, when using the most recent compilation of GRB data whose redshifts have been measured. We conclude that the GGR is merely a pattern that occurred by chance in the random distribution of GRBs.

Entropy-based test for generalised Gaussian distributions

The proof of L2 consistency for the kth nearest neighbour distance estimator of the Shannon entropy for an arbitrary fixed k≥1 is provided. It is constructed the non-parametric test of goodness-of-fit for a class of introduced generalised multivariate Gaussian distributions based on a maximum entropy principle. The theoretical results are followed by numerical studies on simulated samples. It is shown that increasing of k improves the power of the introduced goodness of fit tests. The asymptotic normality of the test statistics is experimentally proven.

Nearest neighbor estimates of Kaniadakis entropy

Abstract The aim of this paper is to develop new nonparametric estimators of entropy based on the kth nearest neighbor distances that are considered between n sample points, k ≤ (n − 1) being a positive integer, fixed. The Method consists in using the new estimators which were useful in order to evaluate the entropies for random vectors. As results, using the Kaniadakis entropy measure, the asymptotic unbiasedness and consistency of the estimators are proven.

Using probabilistic solar power forecasts to inform flexible ramp product procurement for the California ISO

How can independent system operators (ISOs) take advantage of probabilistic solar forecasts to lower generation costs and improve reliability of power systems? We discuss one three-step approach for doing so, focusing on how such forecasts might help the California Independent System Operator (CAISO) prepare unexpected net load ramps, where net load equals gross demand minus wind and solar production. First, we enhance an existing solar forecasting system to provide well-calibrated hours-ahead probabilistic forecasts. We then relate the degree of uncertainty reflected in the forecasted prediction intervals (independent variables) to error distributions for net load ramp forecasts for the CAISO real-time market (dependent variable) using machine learning and quantile regression. Projected ramp forecast errors conditioned on solar uncertainty are translated into flexible ramp requirements that therefore reflect real-time meteorological and solar conditions, improving on typical ISO procedures. Detailed descriptions are provided on the quantile regression and kth-nearest neighbor categorization methods for accomplishing that translation. Finally, a multiple time-scale look-ahead market simulation model is applied to a 118-bus IEEE Reliability Test System, modified to represent the CAISO generation mix and demand distributions. The model runs quantify how solar-conditioned ramp requirements can, first, decrease operating costs by reducing requirements compared to often conservative unconditional methods and, second, decrease generation scarcity events and consequently improve reliability by increasing flexibility requirements at times when unconditional forecast-based requirements understate actual ramp uncertainty. Solar-conditioned ramp requirements are found to reduce generation operating costs by about 2% for the test system (which would be equivalent to over $100 million per year for a CAISO-size system).

A Novel Grid-Based Clustering Algorithm

Abstract Data clustering is an important method used to discover naturally occurring structures in datasets. One of the most popular approaches is the grid-based concept of clustering algorithms. This kind of method is characterized by a fast processing time and it can also discover clusters of arbitrary shapes in datasets. These properties allow these methods to be used in many different applications. Researchers have created many versions of the clustering method using the grid-based approach. However, the key issue is the right choice of the number of grid cells. This paper proposes a novel grid-based algorithm which uses a method for an automatic determining of the number of grid cells. This method is based on the kdist function which computes the distance between each element of a dataset and its kth nearest neighbor. Experimental results have been obtained for several different datasets and they confirm a very good performance of the newly proposed method.

Comparison of outlier detection techniques in non-stationary time series data

This study examined the performance of six outlier detection techniques using a non-stationary time series dataset. Two key issues were of interest. Scenario one was the method that could correctly detect the number of outliers introduced into the dataset whiles scenario two was to find the technique that would over detect the number of outliers introduced into the dataset, when a dataset contains only extreme maxima values, extreme minima values or both. Air passenger dataset was used with different outliers or extreme values ranging from 1 to 10 and 40. The six outlier detection techniques used in this study were Mahalanobis distance, depth-based, robust kernel-based outlier factor (RKOF), generalized dispersion, Kth nearest neighbors distance (KNND), and principal component (PC) methods. When detecting extreme maxima, the Mahalanobis and the principal component methods performed better in correctly detecting outliers in the dataset. Also, the Mahalanobis method could identify more outliers than the others, making it the "best" method for the extreme minima category. The kth nearest neighbor distance method was the "best" method for not over-detecting the number of outliers for extreme minima. However, the Mahalanobis distance and the principal component methods were the "best" performed methods for not over-detecting the number of outliers for the extreme maxima category. Therefore, the Mahalanobis outlier detection technique is recommended for detecting outlier in nonstationary time series data.

A genetic algorithm approach for predicting ribonucleic acid sequencing data classification using KNN and decision tree

Malaria larvae accept explosive variable lifecycle as they spread across numerous mosquito vector stratosphere. Transcriptomes arise in thousands of diverse parasites. Ribonucleic acid sequencing (RNA-seq) is a prevalent gene expression that has led to enhanced understanding of genetic queries. RNA-seq tests transcript of gene expression, and provides methodological enhancements to machine learning procedures. Researchers have proposed several methods in evaluating and learning biological data. Genetic algorithm (GA) as a feature selection process is used in this study to fetch relevant information from the RNA-Seq Mosquito Anopheles gambiae malaria vector dataset, and evaluates the results using kth nearest neighbor (KNN) and decision tree classification algorithms. The experimental results obtained a classification accuracy of 88.3 and 98.3 percents respectively.

Formula omitted]th distance distributions for generalized Gauss-Poisson process in [formula omitted

For a point process (PP), the kth contact distance refers to the distance of kth closest point from an arbitrary location and the kth nearest neighbor distance refers to the distance of kth nearest neighbor from an arbitrary point of the PP. We consider the generalized n-dimensional Gauss-Poisson process and derive the closed-form expressions for the cumulative distribution functions (CDFs) of these two distances for the general k. We also validate our analysis via numerical simulations and provide various insights using the presented analysis.

Temporal Series Crop Classification Study in Rural China Based on Sentinel-1 SAR Data

Crop classification is one of the focused topics in remote sensing study nowadays. Optical imageries, although providing much information, are often contaminated with cloud or other weather effects while SAR imageries are more resilient to those. Temporal series data are often used to improve classification accuracy, especially in crop classification. This article investigates the usage of temporal series SAR imagery on crop classification in vast rural areas of China. The selected area of interest has a complicated, heavily mixed agriculture as well as lots of nonagricultural landcovers. Total ten classes are considered, six of them being crop types. A pixel-based classifier using subspace k th nearest neighbor (KNN) algorithm is applied to open source Sentinel-1 polarized SAR data. Discussion includes analyzing temporal features of the SAR observation, time domain smoothing, feature engineering, different classification algorithms, and selection of temporal series. The study results in an overall accuracy of 98.2% for the ten classes in fivefold cross validation, indicating a propitious application for agricultural monitoring using SAR data.

Mental imagery classification using one-dimensional convolutional neural network for target selection in single-channel BCI-controlled mobile robot

This paper introduces the use of the one-dimensional convolutional neural network (1D-CNN) for end-to-end EEG decoding with application towards a BCI system with a shared control scheme. In general, subjects wearing a single-channel EEG electrode located at F8 (10–20 international standards) were required to perform mental tasks by mentally visualising the rotation of a star and mind relaxation at a specific time and by robot orientation. The visualisation of a rotating star suggests that the mobile robot is currently oriented towards a target, thus enabling target selection. We showed that proposed classifier obtained the best accuracy of 92.09% in classifying the subject’s performing mental rotation task or mental relaxation when compared with conventional classification methods such as support vector machine—75.69%, Kth-nearest neighbour—65.50% and linear discriminant analysis—65.20%. Furthermore, different from conventional methods, the use of 1D-CNN enables end-to-end learning, that is the automatic decoding of EEG signals without requiring feature selection or extraction. To validate that the proposed classifier performs better than conventional methods, the extracted kernel weights of proposed 1D-CNN filters were visualised as a temporal plot, and spectral analysis was performed on the extracted weights. The obtained results confirmed that the proposed 1D-CNN was able to generate filters that resemble the EEG wave patterns of different frequencies and spectral analysis confirmed that the filters exploited information from multiple frequency bands (such as alpha band and beta band) that are often associated with a heightened mental state when performing mental tasks.