Kernel Density Estimation Model Research Articles

Application of a DPSIR Modeling Framework to Assess Spatial–Temporal Differences of Water Poverty in China

AbstractWater shortage is a common problem around the world, especially in developing countries. Water shortage is closely linked to natural and social conditions, but the linkages between these natural and social conditions and their underlying temporal and spatial variation are less well explored. This paper details an application of the Driving‐Force‐Pressure‐State‐Impact‐Response (DPSIR) model, a holistic and sustainable tool for resources planning and management, and uses comprehensive weights to evaluate the water poverty (wp) in China from 1997 to 2014. This study applies the Kernel density estimation model to analyze the temporal variation trend and uses the least square error model to analyze the spatial pattern of wp. The results show the level of wp is gradually declining over time and the improvements in the coastal and inland wp situation are not spatially harmonious, and there are four primary types of wp in China based on drivers and causal mechanisms: D‐P‐I, D‐P‐I‐R, D‐P‐S‐I, and D‐P‐S‐I‐R. Furthermore, we analyze the main causes of spatial difference of wp and put forward corresponding countermeasures. The research findings are intended to provide a new insight for the evaluation of wp in the context of sustainable development, breaking past limitations that arise in simplified analyses using a single method, and to provide a strategy for regional water resources management to relieve wp.

Wind Power Prediction Based on LSTM Networks and Nonparametric Kernel Density Estimation

Wind energy is a kind of sustainable energy with strong uncertainty. With a large amount of wind power injected into the power grid, it will inevitably affect the security, stability and economic operation of the power grid. High-precision wind power spot prediction and fluctuation interval information can provide more adequate decision-making support for grid scheduling and optimization. Hence, this paper proposes a K-Means-long short-term memory (K-Means-LSTM) network model for wind power spot prediction, and a nonparametric kernel density estimation (KDE) model with bandwidth optimization for wind power probabilistic interval prediction. The long short-term memory (LSTM) network has a strong memory function, which can establish the correlation between the data before and after, so as to improve the prediction accuracy. The K-Means clustering method forms different clusters of wind power impact factors to generate a new LSTM sub-prediction model. The optimization of the bandwidth in the nonparametric KDE is implemented by the mean integrated squared error criterion. In addition, a part of the dataset is deliberately demarcated from the wind power historical dataset to generate reasonable wind power prediction errors. The simulation results show that the proposed K-Means-LSTM network model has higher prediction accuracy than the back propagation (BP) neural networks, Elman neural networks, support vector regression (SVR) and LSTM network models. Compared with the KDE model with random bandwidth and the Gaussian distribution model, the bandwidth optimization model proposed in this paper has more narrow prediction intervals with higher interval coverage rates.

Personalized Check-in Prediction Model Based on User’s Dissimilarity and Regression

To solve the problem that the user check-in prediction model is difficult to provide personalized check-in services, this paper proposes a novel hybrid model, called personalized check-in prediction model based on user's dissimilarity and regression (UDR). The UDR is mainly composed of two sub-models: user's regression location prediction model (UR) and user's dissimilarity location prediction model (UD). In UR, considering the personalization of user check-ins, we propose a hybrid weighted Markov model, which combines the whole check-in data and individual check-ins. Different from other methods, for the prediction of individual check-ins, we not only consider the preference of individual users, but also the influence of friend relationships. Meanwhile, the Hidden Markov model(HMM) is used to determine users' next check-in location by using time series feature (week-hour) and location sequence. In addition, by improving the kernel density estimation, we propose a multi-level hybrid kernel density estimation model, which is built based on the individual, city and region layers, and smoothes the over-fitting phenomenon caused by few check-ins. In UD, we take into account the weather factors that most existing methods did not consider. By defining the “cold and hot spot transference” and weather similarity features, we explore the influence of weather on user's check-ins and also propose a method used to calculate the similarity between user check-in weather preferences and location weathers. At the same time, the influence of social, time, and space factors are also considered. The experiments on two LBSN datasets demonstrate that the performance of UDR is superior to the state-of-the-art check-in prediction methods.

Background Modeling with Extracted Dynamic Pixels for Pumping Unit Surveillance

Background modeling plays an important role in the application of intelligent video surveillance. Researchers have presented diverse approaches to support the development of dynamic background modeling. However, in the case of pumping unit surveillance, traditional background modeling methods often mistakenly detect the periodic rotational pumping unit as the foreground object. To address this problem here, we propose a novel background modeling method for foreground segmentation, particularly in dynamic scenes that include a rotational pumping unit. In the proposed method, the ViBe method is employed to extract possible foreground pixels from the sequence frames and then segment the video image into dynamic and static regions. Subsequently, the kernel density estimation (KDE) method is used to build a background model with dynamic samples of each pixel. The bandwidth and threshold of the KDE model are calculated according to the sample distribution and extremum of each dynamic pixel. In addition, the strategy of sample adjustment combines regular and real-time updates. The performance of the proposed method is evaluated against several state-of-the-art methods applied to complex dynamic scenes consisting of a rotational pumping unit. Experimental results show that the proposed method is available for periodic object motion scenario monitoring applications.

An mRMR-SVM Approach for Opto-Fluidic Microorganism Classification.

The detection of microorganisms is important in numerous applications such as water quality monitoring, blood analysis, and food testing. The conventional detection methods are tedious and labour-intensive. Establish methods involve culturing, counting and identification of the pathogen by an experienced technician which typically can take several days. The use of opto-fluidic technology to capture microorganism images offers 0 route to reduce the overall assay time. However, the detection still requires a trained technician. This paper proposes an image processing method that can be used to classify microorganism images captured by an opto-fluidic set up in an automatic manner. The proposed algorithm incorporates some of the features used in other microorganism image detection methods and proposes two new features-Entropy of Histogram of Oriented Gradients (HOG) and the filtered intensities. In addition, we propose to apply the minimal-Redundancy-Maximal-Relevance (mRMR) criterion to select and rank these features. The probability and joint probability distribution functions of the mRMR are estimated using a Gaussian model and the Kernel Density Estimation model. The performance of the proposed method was validated using SVM and data collected from an experimental setup. The results show that our proposed method outperforms existing methods and is capable of achieving a classification accuracy up to 95.8%.

Empirical safety stock estimation based on kernel and GARCH models

Supply chain risk management has drawn the attention of practitioners and academics alike. One source of risk is demand uncertainty. Demand forecasting and safety stock levels are employed to address this risk. Most previous work has focused on point demand forecasting, given that the forecast errors satisfy the typical normal i.i.d. assumption. However, the real demand for products is difficult to forecast accurately, which means that—at minimum—the i.i.d. assumption should be questioned. This work analyzes the effects of possible deviations from the i.i.d. assumption and proposes empirical methods based on kernel density estimation (non-parametric) and GARCH(1,1) models (parametric), among others, for computing the safety stock levels. The results suggest that for shorter lead times, the normality deviation is more important, and kernel density estimation is most suitable. By contrast, for longer lead times, GARCH models are more appropriate because the autocorrelation of the variance of the forecast errors is the most important deviation. In fact, even when no autocorrelation is present in the original demand, such autocorrelation can be present as a consequence of the overlapping process used to compute the lead time forecasts and the uncertainties arising in the estimation of the parameters of the forecasting model. Improvements are shown in terms of cycle service level, inventory investment and backorder volume. Simulations and real demand data from a manufacturer are used to illustrate our methodology.

Estimation and application of semiparametric stochastic volatility models based on kernel density estimation and hidden Markov models

Discrete‐time stochastic volatility models play a key role in the analysis of financial time series. However, the parametric assumption of conditional distribution for asset returns, given the volatility, has been questioned. When the conditional distribution is unknown and unspecified, in this paper, a maximum‐likelihood estimation approach for the semiparametric stochastic volatility models is proposed based on kernel density estimation and hidden Markov models. Several numerical studies are conducted to evaluate the finite sample performance of the proposed estimation method. Implementation on empirical studies also illustrates the validity of the proposed method in practice.

Spatial characteristics of tourism resources and the development pattern in jurisdiction of Hulunbuir Agricultural Reclamation Group

The question of how tourism elements can be spatially organized continues to be challenging, particularly for tourism in agricultural reclamation area (TAR) in jurisdiction of Hulunbuir Agricultural Reclamation Group(HARG), an emerging tourism destination. To date, the spatial characteristics of TAR resources has not been adequately addressed and few studies have considered spatial development pattern of TAR. Point-axis theory is a useful framework to analysis the spatial elements, structure and evolution. This paper uses the point-axis theory to propose the spatial development pattern of TAR in Hulunbuir. We constructed the point, axis and zone space of tourism based on spatial characteristics of tourism resources identified by kernel density estimation and local spatial autocorrelation model. The study finds that, Hulunbuir Agricultural Reclamation Group owns abundant tourism resources. The unique feature of location and advantage of it mainly on architecture and water landscapes. The spatial distribution of these tourism attractions was uneven. The tourism attractions have been more in the western part and less in the eastern part, also highly concentrated in South and scattered in North. The tourism attractions mainly distributed on either side of the Greater Khingan Mountains, where had formed four accumulation areas. In addition, these tourism attractions concentrate on area along the traffic line and river. Common tourism resources account for principal parts of tourism resources of Hulunbuir Agricultural Reclamation Group. From the perspective of quality level, different type of tourism attractions are concentrate in different way, with the high-high cluster, high-low cluster and low-high cluster were the majority, and the low-low cluster is scarce. Based on that, the spatial development patterns was proposed. In general, the spatial distribution of TAR’s resources can be characterized as a “point-line-area” structure. Convenient links provided between the points and zones via lines, forming a string of cluster of elements. In terms of types, biological landscape, tourism commodity and water sceneries mainly concentrated at the northwest of the Greater Khingan (GK), Architecture and facilities center on the northwest and southwest of the GK and humanity activities mainly distributed in the north of GK. The study find a small differences of distribution exists in site, relic and meteorological landscapes. In terms of quality, the overall level of quality in western GK is higher than which of eastern GK. There are two high-level clusters in Hulunbuir agricultural reclamation area, one is in Taohai, Xiertala and another is in Chaoerhe. This paper presents a point-axis-zone pattern for TAR development in Hulunbuir. The pattern includes (1) Taohai wetland, Xiertala wetland and Chaoerhe wetland as the key tourism destination and Ganhe farms, Najitun farm, Taohai farm and Labudalin farm as tourist distributing center, (2) 4 transport corridors—G111, G301, S201, S301 as the main axis, the galleries of northwestern, southwestern and southern of GK as the secondary axis, and (3) the area of northeastern, southeastern, northwestern and southwest of GK as the funtional zone.

Multi-scale analysis on spatial morphology differentiation and formation mechanism of rural residential land: A case study in Shandong Province, China

Using the GIS space “hot spot” detection and kernel density estimation model, the spatial differentiation characteristics of rural residential land in Shandong Province are revealed from the perspective of multi-scale feature units of “point-line-surface”. Then, a geographical detector is used to analyze its factors and mechanisms. Results show that on the domain-wide scale, the rural residential land has obvious characteristics of spatial agglomeration that are sparsely distributed in a ladder from west to east; on the transect scale, the scale and distribution density of rural residential land present a multi-peak distribution, while the plaque shapes are stochastic equilibrium. On the point scale, the rural residential monomer takes on the feature of banded and serrated diversification. The differentiation is the result of scale difference of different influencing factors. Natural geographical conditions are influencing factors of multi-scale stability, especially in macro scale, which plays a comprehensive role in controlling the formation and changes. The characteristics of rural residential land in meso scale are mainly formed under the influence of dynamic equilibrium of economic and social conditions, the periodic change of institutional policy environment has a profound influence on micro-scale rural residential from rigid constraints and timeliness.

A multi-granularity perspective for spatial profiling of mobile apps

The prevalence of mobile apps has greatly changed people’s lives, generating myriads of data and creating new research opportunities. By exploiting mobile apps’ data, existing studies have primarily investigated app users’ usage patterns. Few studies have focused on profiling the geospatial distributions of individual apps’ usage patterns on a large scale. A major challenge for profiling app usage is the heterogeneity and sparsity of individual apps’ data. In this study, we propose a multi-level mixture of kernel density estimation (mlKDE) model for robust profiling of the geospatial distribution of any given app. A major advantage of this model is that it can leverage aggregate information from groups of related apps, and adaptively train the weights for different groups. Using a real-world, large-scale app usage dataset covering more than one thousand apps available on the market, we demonstrate that our model can effectively characterize the distributions of app usage in the real world, and has considerable advantages when compared to the baseline models reported in the literature.

Estimating join selectivities using bandwidth-optimized kernel density models

Accurately predicting the cardinality of intermediate plan operations is an essential part of any modern relational query optimizer. The accuracy of said estimates has a strong and direct impact on the quality of the generated plans, and incorrect estimates can have a negative impact on query performance. One of the biggest challenges in this field is to predict the result size of join operations. Kernel Density Estimation (KDE) is a statistical method to estimate multivariate probability distributions from a data sample. Previously, we introduced a modern, self-tuning selectivity estimator for range scans based on KDE that out-performs state-of-the-art multidimensional histograms and is efficient to evaluate on graphics cards. In this paper, we extend these bandwidth-optimized KDE models to estimate the result size of single and multiple joins. In particular, we propose two approaches: (1) Building a KDE model from a sample drawn from the join result. (2) Efficiently combining the information from base table KDE models. We evaluated our KDE-based join estimators on a variety of synthetic and real-world datasets, demonstrating that they are superior to state-of-the art join estimators based on sketching or sampling.

An Improved Kernel Density Estimation Approach for Moving Objects Detection

Moving object detection based on monitoring video system is often a challenging problem. Specially to moni- tor traffic at both day and night, in different weather and illumination conditions and with changeable background. Kernel Density Estimation (KDE) model is an effective approach to judge background and foreground, however, typical KDE uses fixed parameters, such as bandwidths, threshold, etc. This paper proposes a detection algorithm based on an Im- proved Kernel Density Estimation (IKDE) model. The proper bandwidths, adaptive background sample learning array, and adaptive threshold, and an improved sample updating method for sample learning array are discussed as the funda- mentals of the IKDE model. Furthermore, an algorithm for restraining light field disturbance at night in video scene is proposed. Video image series are evaluated through the algorithm, and moving object detection is conducted in three dif- ferent scenes. Results show that the algorithm can help to achieve a promising high accuracy and robustness for detecting moving objects.

Retina Verification System Based on Biometric Graph Matching

This paper presents an automatic retina verification framework based on the biometric graph matching (BGM) algorithm. The retinal vasculature is extracted using a family of matched filters in the frequency domain and morphological operators. Then, retinal templates are defined as formal spatial graphs derived from the retinal vasculature. The BGM algorithm, a noisy graph matching algorithm, robust to translation, non-linear distortion, and small rotations, is used to compare retinal templates. The BGM algorithm uses graph topology to define three distance measures between a pair of graphs, two of which are new. A support vector machine (SVM) classifier is used to distinguish between genuine and imposter comparisons. Using single as well as multiple graph measures, the classifier achieves complete separation on a training set of images from the VARIA database (60% of the data), equaling the state-of-the-art for retina verification. Because the available data set is small, kernel density estimation (KDE) of the genuine and imposter score distributions of the training set are used to measure performance of the BGM algorithm. In the one dimensional case, the KDE model is validated with the testing set. A 0 EER on testing shows that the KDE model is a good fit for the empirical distribution. For the multiple graph measures, a novel combination of the SVM boundary and the KDE model is used to obtain a fair comparison with the KDE model for the single measure. A clear benefit in using multiple graph measures over a single measure to distinguish genuine and imposter comparisons is demonstrated by a drop in theoretical error of between 60% and more than two orders of magnitude.

Time Adaptive Conditional Kernel Density Estimation for Wind Power Forecasting

This paper reports the application of a new kernel density estimation model based on the Nadaraya-Watson estimator, for the problem of wind power uncertainty forecasting. The new model is described, including the use of kernels specific to the wind power problem. A novel time-adaptive approach is presented. The quality of the new model is benchmarked against a splines quantile regression model currently in use in the industry. The case studies refer to two distinct wind farms in the United States and show that the new model produces better results, evaluated with suitable quality metrics such as calibration, sharpness, and skill score.

Evaluating temporally weighted kernel density methods for predicting the next event location in a series

One aspect of tactical crime or terrorism analysis is predicting the location of the next event in a series. The objective of this article is to present a methodology to identify the optimal parameters and to test the performance of temporally weighted kernel density estimation models for predicting the next event in a criminal or terrorist event series. By placing event series in a space–time point pattern framework, the next event prediction models are shown to be based on estimating a conditional spatial density function. We use temporal weights that indicate how much influence past events have toward predicting future event locations, which can also incorporate uncertainty in the event timing. Results of applying this methodology to crime series in Baltimore County, MD, indicate that performance can vary greatly by crime type and little by series length and is fairly robust to choice of bandwidth.

Unsupervised Salient Object Segmentation Based on Kernel Density Estimation and Two-Phase Graph Cut

In this paper, we propose an unsupervised salient object segmentation approach based on kernel density estimation (KDE) and two-phase graph cut. A set of KDE models are first constructed based on the pre-segmentation result of the input image, and then for each pixel, a set of likelihoods to fit all KDE models are calculated accordingly. The color saliency and spatial saliency of each KDE model are then evaluated based on its color distinctiveness and spatial distribution, and the pixel-wise saliency map is generated by integrating likelihood measures of pixels and saliency measures of KDE models. In the first phase of salient object segmentation, the saliency map based graph cut is exploited to obtain an initial segmentation result. In the second phase, the segmentation is further refined based on an iterative seed adjustment method, which efficiently utilizes the information of minimum cut generated using the KDE model based graph cut, and exploits a balancing weight update scheme for convergence of segmentation refinement. Experimental results on a dataset containing 1000 test images with ground truths demonstrate the better segmentation performance of our approach.

A Bibliometric Analysis of Science Cooperation Group Size

The paper analyzes the applicability of Weibull Density Distribution and Kernel Density Estimation in research of scientific productivity and collaboration group size. The datasets from five categories were retrieved and downloaded from ISI, and they are used to verify Kernel Density Estimation curve fits well with the Weibull Density Distribution curve. As a comparatively convenient function, Kernel Density Estimation can be used as a model to study the relationship between scientific productivity and group size. The best group size and reasonable group size are suggested through Kernel Density Estimation model. We assumed that the best group size in scientific cooperation is positive to discipline biology connotation, however, the result didn’t show such correlation obviously.

A model for dynamic object segmentation with kernel density estimation based on gradient features

The dynamic object segmentation in videos taken from a static camera is a basic technique in many vision surveillance applications. In order to suppress fake objects caused by dynamic cast shadows and reflection images, this paper presents a novel segmentation model with the function of cast shadow and reflection image suppression. This model is a kernel density estimation model based on dynamic gradient features. Unlike the conventional kernel density estimation model which can only suppress cast shadows in color videos, this model can also suppress them in intensity videos, and under the circumstance of diffusion it can suppress reflection images effectively. Although this model may cause the increase of the false negative rate, its function of fake object suppression is remarkable. Furthermore, the false negative rate can be reduced with other convenient methods. Some experimental results by real videos are also presented in this paper to demonstrate the effectiveness of this model.

Trends in Income Distribution in Italy: A Non-Parametric 90 and a Semi-Parametric Analysis

Using kernel density estimation and mixture models, household size-adjusted income distributions in Italy are cross-sectionally examined over the period 1987-2002. A non-parametric test is used to assess the number of modes in the distributions. Evidence shows that income tends to cluster around more than one point. This gives good reason to model the shapes by a finite mixture density with an appropriate choice of components which represent homogeneous sub-populations. Effects of social and demographic factors on the probability of households to belong to one of the components of the mixture are identified by a compositional data analysis.

A General Approach to Clustering in Large Databases with Noise

Several clustering algorithms can be applied to clustering in large multimedia databases. The effectiveness and efficiency of the existing algorithms, however, are somewhat limited, since clustering in multimedia databases requires clustering of high-dimensional feature vectors and because multimedia databases often contain large amounts of noise. In this paper, we therefore introduce a new Kernel Density Estimation-based algorithm for clustering in large multimedia databases called DENCLUE (DENsity-based CLUstEring). Kernel Density Estimation (KDE) models the overall point density analytically as the sum of kernel (or influence) functions of the data points. Clusters can then be identified by determining density attractors and clusters of arbitrary shape can be easily described by a simple equation of the overall density function. The advantages of our KDE-based DENCLUE approach are: (1) it has a firm mathematical basis; (2) it has good clustering properties in data sets with large amounts of noise; (3) it allows a compact mathematical description of arbitrarily shaped clusters in high-dimensional data sets; and (4) it is significantly faster than existing algorithms. To demonstrate the effectiveness and efficiency of DENCLUE, we perform a series of experiments on a number of different data sets from CAD and molecular biology. A comparison with k-Means, DBSCAN, and BIRCH shows the superiority of our new algorithm.