Unsupervised Functional Data Clustering Based on Adaptive Weights

In recent years, functional data has been widely used in finance, medicine, biology and other fields. The current clustering analysis can solve the problems in finite-dimensional space, but it is difficult to be directly used for the clustering of functional data. In this paper, we propose a new unsupervised clustering algorithm based on adaptive weights. In the absence of initialization parameter, we use entropy-type penalty terms and fuzzy partition matrix to find the optimal number of clusters. At the same time, we introduce a measure based on adaptive weights to reflect the difference in information content between different clustering metrics. Simulation experiments show that the proposed algorithm has higher purity than some algorithms.

Outlier Detection of Air Quality for Two Indian Urban Cities Using Functional Data Analysis

Human living would be impossible without air quality. Consistent advancements in practically every aspect of contemporary human life have harmed air quality. Everyday industrial, transportation, and home activities turn up dangerous contaminants in our surroundings. This study investigated two years’ worth of air quality and outlier detection data from two Indian cities. Studies on air pollution have used numerous types of methodologies, with various gases being seen as a vector whose components include gas concentration values for each observation per-formed. We use curves to represent the monthly average of daily gas emissions in our technique. The approach, which is based on functional depth, was used to find outliers in the city of Delhi and Kolkata’s gas emissions, and the outcomes were compared to those from the traditional method. In the evaluation and comparison of these models’ performances, the functional approach model studied well.

Spatio-temporal variability of surface chlorophyll a in the Yellow Sea and the East China Sea based on reconstructions of satellite data of 2001-2020

Chlorophyll-a(Chl-a)concentration is a primary indicator for marine environmental monitoring.The spatio-temporal variations of sea surface Chl-a concentration in the Yellow Sea(YS)and the East China Sea(ECS)in 2001-2020 were investigated by reconstructing the MODIS Level 3 products with the data interpolation empirical orthogonal function(DINEOF)method.The reconstructed results by interpolating the combined MODIS daily+8-day datasets were found better than those merely by interpolating daily or 8-day data.Chl-a concentration in the YS and the ECS reached its maximum in spring,with blooms occurring,decreased in summer and autumn,and increased in late autumn and early winter.By performing empirical orthogonal function(EOF)decomposition of the reconstructed data fields and correlation analysis with several potential environmental factors,we found that the sea surface temperature(SST)plays a significant role in the seasonal variation of Chl a,especially during spring and summer.The increase of SST in spring and the upper-layer nutrients mixed up during the last winter might favor the occurrence of spring blooms.The high sea surface temperature(SST)throughout the summer would strengthen the vertical stratification and prevent nutrients supply from deep water,resulting in low surface Chl-a concentrations.The sea surface Chl-a concentration in the YS was found decreased significantly from 2012 to 2020,which was possibly related to the Pacific Decadal Oscillation(PDO).

Source time functions of the Gonghe，China earthquake retrieved from long－period digital waveform data using empirical Green＇s function technique

An earthquake of Ms= 6, 9 occurred at the Gonghe, Qinghai Province, China on April 26, 1990. Three larger aftershocks took place at the same region, Ms= 5. 0 on May 7, 1990, Ms= 6. 0 on Jan. 3, 1994 and Ms= 5. 7on Feb. 16, 1994. The long-period recordings of the main shock from China Digital Seismograph Network (CDSN) are deconvolved for the source time functions by the correspondent0 recordings of the three aftershocks asempirical Green's functions (EGFs). No matter which aftershock is taken as EGF, the relative source time functions (RSTFs) Obtained are nearly identical. The RSTFs suggest the Ms= 6. 9 event consists of at least two subevents with approximately equal size whose occurrence times are about 30 s apart, the first one has a duration of 12 s and a rise time of about 5 s, and the second one has a duration of 17 s and a rise time of about & s. COmParing the RSTFs obtained from P- and SH-phases respectively, we notice that those from SH-phases are a slightly more complex than those from p-phases, implying other finer subevents exist during the process of the main shock. It is interesting that the results from the EGF deconvolution of long-Period way form data are in good agreement with the results from the moment tensor inversion and from the EGF deconvolution of broadband waveform data. Additionally, the two larger aftershocks are deconvolved for their RSTFs. The deconvolution results show that the processes of the Ms= 6. 0 event on Jan. 3, 1994 and the Ms= 5. 7 event on Feb. 16,1994 are quite simple, both RSTFs are single impulses.The RSTFs of the Ms= 6. 9 main shock obtained from different stations are noticed to be azimuthally dependent, whose shapes are a slightly different with different stations. However, the RSTFs of the two smaller aftershocks are not azimuthally dependent. The integrations of RSTFs over the processes are quite close to each other, i. e., the scalar seismic moments estimated from different stations are in good agreement. Finally the scalar seismic moments of the three aftershocks are compared. The relative scalar seismic moment Of the three aftershocks deduced from the relative scalar seismic moments of the Ms=6. 9 main shock are very close to those inverted directly from the EGF deconvolution. The relative scalar seismic moment of the Ms =6. 9 main shock calculated using the three aftershocks as EGF are 22 (the Ms= 6. 0 aftershock being EGF), 26 (the Ms= 5. 7 aftershock being EGF) and 66 (the Ms= 5. 5 aftershock being EGF), respectively. Deducingfrom those results, the relative scalar sesimic moments of the Ms= 6. 0 to the Ms= 5. 7 events, the Ms= 6. 0 tothe Ms= 5. 5 events and the Ms= 5. 7 to the Ms= 5. 5 events are 1. 18, 3. 00 and 2. 54, respectively. The correspondent relative scalar seismic moments calculated directly from the waveform recordings are 1. 15, 3. 43, and 3. 05.

HERMITE—BIRKHOFF INTERPOLATION OF SCATTERED DATA BY RADIAL BASIS FUNCTIONS

For Hermite-Birkhoff interpolation of scattered multidumensional data by radial basis function (?),existence and characterization theorems and a variational principle are proved. Examples include (?)(r)=r^b,Duchon's thin-plate splines,Hardy's multiquadrics,and inverse multiquadrics.

Surface Reconstruction of 3D Scattered Data with Radial Basis Functions

We use Radial Basis Functions （RBFs） to reconstruct smooth surfaces from 3D scattered data. An object＇s surface is defined implicitly as the zero set of an RBF fitted to the given surface data. We propose improvements on the methods of surface reconstruction with radial basis functions. A sparse approximation set of scattered data is constructed by reducing the number of interpolating points on the surface. We present an adaptive method for finding the off-surface normal points. The order of the equation decreases greatly as the number of the off-surface constraints reduces gradually. Experimental results are provided to illustrate that the proposed method is robust and may draw beautiful graphics.

A Method for Data Classification Based on Discernibility Matrix and Discernibility Function

A method for data classification will influence the efficiency of classification. Attributes reduction based on discernibility matrix and discernibility function in rough sets can use in data classification, so we put forward a method for data classification. Namely, firstly, we use discernibility matrix and discernibility function to delete superfluous attributes in formation system and get a necessary attribute set. Secondly, we delete superfluous attribute values and get decision rules. Finally, we classify data by means of decision rules. The experiments show that data classification using this method is simpler in the structure, and can improve the efficiency of classification.

Weighted Multi-sensor Data Level Fusion Method of Vibration Signal Based on Correlation Function

As the differences of sensor＇s precision and some random factors are difficult to control,the actual measurement signals are far from the target signals that affect the reliability and precision of rotating machinery fault diagnosis.The traditional signal processing methods,such as classical inference and weighted averaging algorithm usually lack dynamic adaptability that is easy for trends to cause the faults to be misjudged or left out.To enhance the measuring veracity and precision of vibration signal in rotary machine multi-sensor vibration signal fault diagnosis,a novel data level fusion approach is presented on the basis of correlation function analysis to fast determine the weighted value of multi-sensor vibration signals.The approach doesn＇t require knowing the prior information about sensors,and the weighted value of sensors can be confirmed depending on the correlation measure of real-time data tested in the data level fusion process.It gives greater weighted value to the greater correlation measure of sensor signals,and vice versa.The approach can effectively suppress large errors and even can still fuse data in the case of sensor failures because it takes full advantage of sensor＇s own-information to determine the weighted value.Moreover,it has good performance of anti-jamming due to the correlation measures between noise and effective signals are usually small.Through the simulation of typical signal collected from multi-sensors,the comparative analysis of dynamic adaptability and fault tolerance between the proposed approach and traditional weighted averaging approach is taken.Finally,the rotor dynamics and integrated fault simulator is taken as an example to verify the feasibility and advantages of the proposed approach,it is shown that the multi-sensor data level fusion based on correlation function weighted approach is better than the traditional weighted average approach with respect to fusion precision and dynamic adaptability.Meantime,the approach is adaptable and easy to use,can be applied to other areas of vibration measurement.

Functional Analysis of Chemometric Data

The objective of this paper is to present a review of different calibration and classification methods for functional data in the context of chemometric applications. In chemometric, it is usual to measure certain parameters in terms of a set of spectrometric curves that are observed in a finite set of points (functional data). Although the predictor variable is clearly functional, this problem is usually solved by using multivariate calibration techniques that consider it as a finite set of variables associated with the observed points (wavelengths or times). But these explicative variables are highly correlated and it is therefore more informative to reconstruct first the true functional form of the predictor curves. Although it has been published in several articles related to the implementation of functional data analysis techniques in chemometric, their power to solve real problems is not yet well known. Because of this the extension of multivariate calibration techniques (linear regression, principal component regression and partial least squares) and classification methods (linear discriminant analysis and logistic regression) to the functional domain and some relevant chemometric applications are reviewed in this paper.

A KERNEL-TYPE ESTIMATOR OF A QUANTILE FUNCTION UNDER RANDOMLY TRUNCATED DATA

A kernel-type estimator of the quantile function Q（p） = inf{t：F（t） ≥ p}, 0 ≤ p ≤ 1, is proposed based on the kernel smoother when the data are subjected to random truncation. The Bahadur-type representations of the kernel smooth estimator are established, and from Bahadur representations the authors can show that this estimator is strongly consistent, asymptotically normal, and weakly convergent.

CONSERVATIVE ESTIMATING FUNCTION IN THE NONLINEAR REGRESSION MODEL WITH AGGREGATED DATA

The purpose of this paper is to study the theory of conservative estimating functions in nonlinear regression model with aggregated data. In this model, a quasi-score function with aggregated data is defined. When this function happens to be conservative, it is projection of the true score function onto a class of estimation functions. By constructing, the potential function for the projected score with aggregated data is obtained, which have some properties of log-likelihood function.

Application of Multiple Sensor Data Fusion for the Analysis of Human Dynamic Behavior in Space: Assessment and Evaluation of Mobility-Related Functional Impairments

The authors have applied a systems analysis approach to describe the musculoskeletal system as consisting of a stack of superimposed kinematic hier-archical segments in which each lower segment tends to transfer its motion to the other superimposed segments. This segmental chain enables the derivation of both conscious perception and sensory control of action in space. This applied systems analysis approach involves the measurements of the complex motor behavior in order to elucidate the fusion of multiple sensor data for the reliable and efficient acquisition of the kinetic, kinematics and electromyographic data of the human spatial behavior. The acquired kinematic and related kinetic signals represent attributive features of the internal recon-struction of the physical links between the superimposed body segments. In-deed, this reconstruction of the physical links was established as a result of the fusion of the multiple sensor data. Furthermore, this acquired kinematics, kinetics and electromyographic data provided detailed means to record, annotate, process, transmit, and display pertinent information derived from the musculoskeletal system to quantify and differentiate between subjects with mobility-related disabilities and able-bodied subjects, and enabled an inference into the active neural processes underlying balance reactions. To gain insight into the basis for this long-term dependence, the authors have applied the fusion of multiple sensor data to investigate the effects of Cerebral Palsy, Multiple Sclerosis and Diabetic Neuropathy conditions, on biomechanical/neurophysiological changes that may alter the ability of the human loco-motor system to generate ambulation, balance and posture.

Automatically Smoothing the Spectroscopic Data by Cubic B-Spline Basis Functions

In the present paper,a new criterion is derived to obtain the optimum fitting curve while using Cubic B-spline basis functions to remove the statistical noise in the spectroscopic data.In this criterion,firstly,smoothed fitting curves using Cubic B-spline basis functions are selected with the increasing knot number.Then,the best fitting curves are selected according to the value of the minimum residual sum of squares(RSS)of two adjacent fitting curves.In the case of more than one best fitting curves,the authors use Reinsch's first condition to find a better one.The minimum residual sum of squares(RSS)of fitting curve with noisy data is not recommended as the criterion to determine the best fitting curve,because this value decreases to zero as the number of selected channels increases and the minimum value gives no smoothing effect.Compared with Reinsch's method,the derived criterion is simple and enables the smoothing conditions to be determined automatically without any initial input parameter.With the derived criterion,the satisfactory result was obtained for the experimental spectroscopic data to remove the statistical noise using Cubic B-spline basis functions.

Large Scattered Data Fitting Based on Radial Basis Functions

Solving large radial basis function （RBF） interpolation problem with non-customized methods is computationally expensive and the matrices that occur are typically badly conditioned. In order to avoid these difficulties, we present a fitting based on radial basis functions satisfying side conditions by least squares, although compared with interpolation the method loses some accuracy, it reduces the computational cost largely. Since the fitting accuracy and the non-singularity of coefficient matrix in normal equation are relevant to the uniformity of chosen centers of the fitted RBE we present a choice method of uniform centers. Numerical results confirm the fitting efficiency.

STRONG REPRESENTATIONS OF THE SURVIVAL FUNCTION ESTIMATOR ON INCREASING SETS FOR TRUNCATED AND CENSORED DATA

In this paper, based on random left truncated and right censored data, the authors derive strong representations of the cumulative hazard function estimator and the product-limit estimator of the survival function. which are valid up to a given order statistic of the observations. A precise bound for the errors is obtained which only depends on the index of the last order statistic to be included.

区间函数型数据构权方法研究

针对由数据量过大而引起的函数型综合评价信息损失或是计算复杂度过大等问题,文中提出了一种基于区间函数型数据的综合评价方法,并针对区间函数型数据表支持的综合评价问题的特殊性,提出了一种新的确定权重系数的"全局"拉开档次法.相较于函数型综合评价,区间函数型综合评价在呈现连续的函数特征同时,增加了区间化的步骤,能够更好地挖掘数据信息,提升综合评价的效率.最后通过以义乌小商品景气指数为例,分别使用函数型数据形式和区间函数型数据形式下的"全局"拉开档次法对其进行赋权,进而进行评价研究.结果表明该文提出的区间函数型综合评价方法更具优势.

基于非负矩阵分解的函数型聚类算法改进与比较

非负函数型数据可以不等间隔观测,在理论和实践中应用广泛,对其进行聚类可以更好地探索客观规律。文章利用位置积分变换将函数型数据转化为高维向量,再通过非负矩阵分解(NMF)将其转化为低维向量,以此构建函数型聚类算法。针对基于NMF的函数型谱聚类算法,给出了确定聚类个数K的两种方法:一种是根据Laplacian矩阵的特征值确定K;另一种是构建新评价指标,通过搜索确定K。数值实验结果显示:基于位置积分变换和NMF的函数型聚类算法有效,对函数结构要求宽松,但需限制函数取值为正;NMF的秩可通过cophenetic相关系数确定,建议取较小的值,以剔除类的冗余特征。在确定谱聚类的聚类个数K时,建议对降维后的数据进行标准化处理,以缩小样本间的距离变化范围;聚类个数变点图直观有效,再结合特征值差分法确定K很有参考价值,建议阈值取[0.05,0.08];根据吻合度与相似比确定K的方法有效且简单易懂。

城市功能区识别研究进展与趋势

随着经济社会的快速发展,城市开发边界迅速由中心向外蔓延,识别城市功能区可为城市建设与规划提供参考依据,且对城市空间和资源合理配置利用具有重要意义。本文在梳理有关城市功能区划分与识别的国内外文献基础上,对城市功能区识别的研究现状进行综述。首先,介绍了用于城市功能区识别的多种数据源,并分析比较其优缺点;然后,总结了城市功能区识别的4类方法,重点分析了深度学习方法在城市功能区识别中的应用,并展开实例分析对比,说明不同数据源和方法对城市功能区识别的有效性;最后,指出了城市功能区划分识别研究领域存在的问题和研究趋势。

基于可见/近红外光谱和函数型线性回归模型的成熟期苹果可溶性固形物含量预测

可溶性固形物含量(SSC)是反映苹果品质和成熟度的重要指标,能够用于苹果品质分析和成熟度预测。以新疆阿克苏冰糖心红富士苹果为研究对象,从果实膨大定形期至完熟期,以3d等间隔周期采摘样本,采集其380~1110nm的可见/近红外光谱,测定其SSC,共552个样品。然后,利用基函数平滑方法将采集的可见/近红外光谱离散数据转化为光谱曲线,即函数型数据,并以可见/近红外光谱曲线、一阶导曲线、二阶导曲线为函数型解释变量,SSC为标量响应变量,分别建立函数型线性回归模型。为了验证和分析模型的性能,根据原始光谱离散数据,经过移动平滑、一阶导和二阶导预处理后,分别建立偏最小二乘回归(PLSR)、核支持向量机(KSVM)、随机森林(RF)、梯度提升树(GBM)和深度神经网络(DeepNN)。结果表明,在建立的18个模型中,针对训练集,PLSR-dNIR模型、KSVM-dNIR模型、RF-dNIR模型、GBM-dNIR模型和Deep NN-d2NIR模型都优于FunLR-NIR模型、FunLR-dNIR模型、FunLR-d2NIR模型,且Deep NN-dNIR模型最优(r_(c)=0.9996,R_(c)^(2)=0.9986,RMSEC=0.0740,RPDC=27.4366);针对测试集,FunLR-NIR模型、FunLR-dNIR模型、FunLR-d2NIR模型均优于其他所有模型,且FunLR-NIR模型最优(r_(v)=0.9534,R_(v)^(2)=0.9077,RMSEV=0.5856,RPDV=3.3017)。综合训练集和测试集的结果来看,核支持向量机模型、随机森林模型、梯度提升树模型和深度神经网络模型容易过拟合,而函数型线性回归模型具有更好的普适性。此外,从三个函数型线性回归模型(FunLR-NIR模型、FunLR-dNIR模型、FunLR-d2NIR模型)的预测效果看,模型均具有良好的鲁棒性和较高的预测精度。试验结果表明,结合可见/近红外光谱技术与函数型数据分析构建的函数型线性回归模型,可成功、有效地实现成熟期苹果的可溶性固形物含量预测。

基于MobileViT的轻量型入侵检测模型研究

为解决入侵检测中数据不平衡对神经网络模型训练的影响和模型参数量高的问题,提出一种基于改进MobileViT的入侵检测模型。首先,使用方差分析提取对检测结果影响较高的特征,将提取后的特征转化为图像型数据,将其输入至MobileViT网络;其次,针对占比较少的攻击流量,采用焦点损失函数自适应地调整攻击流量的损失贡献,使模型更加专注于不平衡的攻击流量;最后,为解决神经元死亡问题,使用GeLU激活函数替换MobileViT网络中MV2的ReLU6激活函数,加快模型收敛速度。实验结果表明,改进的MobileViT模型参数量仅为5.67 MB,与Shufflenet、Mobilenet相比拥有最少的参数量,模型的准确率、召回率以及F_(1)分数分别达到了98.40%、96.49%、95.17%。