Set Of Clusters Research Articles

CIM: A Novel Clustering-based Energy-Efficient Data Imputation Method for Human Activity Recognition

Human activity recognition (HAR) is an important component in a number of health applications, including rehabilitation, Parkinson’s disease, daily activity monitoring, and fitness monitoring. State-of-the-art HAR approaches use multiple sensors on the body to accurately identify activities at runtime. These approaches typically assume that data from all sensors are available for runtime activity recognition. However, data from one or more sensors may be unavailable due to malfunction, energy constraints, or communication challenges between the sensors. Missing data can lead to significant degradation in the accuracy, thus affecting quality of service to users. A common approach for handling missing data is to train classifiers or sensor data recovery algorithms for each combination of missing sensors. However, this results in significant memory and energy overhead on resource-constrained wearable devices. In strong contrast to prior approaches, this paper presents a clustering-based approach (CIM) to impute missing data at runtime. We first define a set of possible clusters and representative data patterns for each sensor in HAR. Then, we create and store a mapping between clusters across sensors. At runtime, when data from a sensor are missing, we utilize the stored mapping table to obtain most likely cluster for the missing sensor. The representative window for the identified cluster is then used as imputation to perform activity classification. We also provide a method to obtain imputation-aware activity prediction sets to handle uncertainty in data when using imputation. Experiments on three HAR datasets show that CIM achieves accuracy within 10% of a baseline without missing data for one missing sensor when providing single activity labels. The accuracy gap drops to less than 1% with imputation-aware classification. Measurements on a low-power processor show that CIM achieves close to 100% energy savings compared to state-of-the-art generative approaches.

Deep Learning Based Analysis of Student Aptitude for Programming at College Freshman Level

Predicting Freshman student’s aptitude for computing is critical for researchers to understand the underlying aptitude for programming. Dataset out of a questionnaire taken from various Senior students in a high school in the city of Kanchipuram, Tamil Nadu, India was used, where the questions related to their social and cultural backgrounds and their experience with computers. Several hypotheses were also generated. The datasets were analyzed using three machine learning algorithms namely, Backpropagation Neural Network (BPN) and Recurrent Neural Network (RNN) (and its variant, Gated Recurrent Network (GNN)) with K-Nearest Neighbor (KNN) used as the classifier. Various models were obtained to validate the underpinning set of hypotheses clusters. The results show that the BPN model achieved a high degree of accuracies on various metrics in predicting Freshman student’s aptitude for computer programming.

Star cluster classification using deep transfer learning with PHANGS-HST

ABSTRACT Currently available star cluster catalogues from the Hubble Space Telescope (HST) imaging of nearby galaxies heavily rely on visual inspection and classification of candidate clusters. The time-consuming nature of this process has limited the production of reliable catalogues and thus also post-observation analysis. To address this problem, deep transfer learning has recently been used to create neural network models that accurately classify star cluster morphologies at production scale for nearby spiral galaxies (D ≲ 20 Mpc). Here, we use HST ultraviolet (UV)–optical imaging of over 20 000 sources in 23 galaxies from the Physics at High Angular resolution in Nearby GalaxieS (PHANGS) survey to train and evaluate two new sets of models: (i) distance-dependent models, based on cluster candidates binned by galaxy distance (9–12, 14–18, and 18–24 Mpc), and (ii) distance-independent models, based on the combined sample of candidates from all galaxies. We find that the overall accuracy of both sets of models is comparable to previous automated star cluster classification studies (∼60–80 per cent) and shows improvement by a factor of 2 in classifying asymmetric and multipeaked clusters from PHANGS-HST. Somewhat surprisingly, while we observe a weak negative correlation between model accuracy and galactic distance, we find that training separate models for the three distance bins does not significantly improve classification accuracy. We also evaluate model accuracy as a function of cluster properties such as brightness, colour, and spectral energy distribution (SED)-fit age. Based on the success of these experiments, our models will provide classifications for the full set of PHANGS-HST candidate clusters (N ∼ 200 000) for public release.

Parametric and nonparametric propensity score estimation in multilevel observational studies.

There has been growing interest in using nonparametric machine learning approaches for propensity score estimation in order to foster robustness against misspecification of the propensity score model. However, the vast majority of studies focused on single-level data settings, and research on nonparametric propensity score estimation in clustered data settings is scarce. In this article, we extend existing research by describing a general algorithm for incorporating random effects into a machine learning model, which we implemented for generalized boosted modeling (GBM). In a simulation study, we investigated the performance of logistic regression, GBM, and Bayesian additive regression trees for inverse probability of treatment weighting (IPW) when the data are clustered, the treatment exposure mechanism is nonlinear, and unmeasured cluster-level confounding is present. For each approach, we compared fixed and random effects propensity score models to single-level models and evaluated their use in both marginal and clustered IPW. We additionally investigated the performance of the standard Super Learner and the balance Super Learner. The results showed that when there was no unmeasured confounding, logistic regression resulted in moderate bias in both marginal and clustered IPW, whereas the nonparametric approaches were unbiased. In presence of cluster-level confounding, fixed and random effects models greatly reduced bias compared to single-level models in marginal IPW, with fixed effects GBM and fixed effects logistic regression performing best. Finally, clustered IPW was overall preferable to marginal IPW and the balance Super Learner outperformed the standard Super Learner, though neither worked as well as their best candidate model.

Interpretation of x‐ray spectral data using self‐organising maps and hierarchical clustering: A study of Vilhelm Hammershøi's use of painting materials

AbstractThe goal of the Vilhelm Hammershøi Digital Archive project of the National Gallery of Denmark is to understand the Danish painter Vilhelm Hammershøi's painting methods by compiling a comprehensive amount of data on his use of materials and working methods through visual and technical examination of a large number of his paintings, and to make this information available to researchers and the public in an open access digital resource. A clear understanding of the full suite of pigments across the paintings requires determination of which materials comprise the palettes of the ground and paint layers. Scanning electron microscopy/energy‐dispersive x‐ray spectroscopy and macro x‐ray fluorescence spectroscopy were selected as the key analytical techniques due to their ability to yield chemical information at the elemental level. This article presents a method that combines unsupervised machine learning and cluster analysis techniques, to automatically reduce the large x‐ray spectral data to sets of distinct clusters that share similar spectra, making it possible to identify materials more precisely. The proposed method allowed the grouping of materials by chemical composition, which enabled an optimal understanding of the pigments used in the ground layers sampled from a large number of paintings as well as in the paint layer examined at the surface of one selected painting. The method performed well when compared with other well‐established data mining techniques, and it helped reduce the time necessary for the interpretation of the analytical results significantly. Through this approach, a basis for a more nuanced view of Hammershøi's artistic idea and technical development will be generated.

Prediction System for Covid-19 Upcoming Cases Using Ensemble Classification

An epidemic of the novel destructive Coronavirus has been spreading rapidly around the world since 2019 and has caused a great number of deaths. Providing patients with appropriate and most timely care is crucial to combating COVID-19 spread. Testing for the disease must be done quickly and accurately. Therefore, this paper developed an ensemble classification-based country-wise COVID-19 upcoming cases prediction model. This ensemble classification and prediction model shows the upcoming month's Corona virus cases, including newly confirmed cases, recovered cases, and deaths. This analysis is carried out based on these three cases occurring in different countries on sequential dates. The proposed model uses three famous classifiers, namely ANN, Gaussian Process and SVM which have different learning characteristics and architectures at the first stage. In the second stage, they combine their predictions with average calculations. Training and assessment of the proposed model were conducted using 75065 observations comprised of 61 features from John Hopkins University in Maryland. For data preparation, the envisioned work clusters the dataset based on world countries affected by COVID-19 separately. As a result, this set of clusters fetched data once again based on death, newly confirmed, and recovered cases. The experimental result shows the proposed ensemble model provides better performance when compared with previous classification algorithms.

Modelling of similarity characteristics of polycyclic aromatic hydrocarbons (PAHs) in Sungai Perak, Malaysia via rough set theory and principal component analysis (PCA)

This paper presents application of rough set theory and PCA for modelling of similarity characteristics of PAHs from Perak River: Tanjung Belanja Bridge (TBB, l1), Water Treatment Plant Parit (WTPP, l2), Parit Town Discharge (PTD, l3), Water Treatment Plant Senin (WTPS, l4), and Water Treatment Plant Kepayang (WTPK, l5). The clustering involved PAHs as attributes and sampling stations as objects. The l1 and l3 were clustered in Simdegx,y≥0.8. The Simdegx,y≥0.7 was observed in all attributes (except l5) forming two sets of clusters. PCA showed that low molecular weight (LMW) PAHs were prominent with variances of 68.47% (naphthalene, Nap) and 27.63% (carbazole).

Two Medoid-Based Algorithms for Clustering Sets

This paper proposes two algorithms for clustering data, which are variable-sized sets of elementary items. An example of such data occurs in the analysis of a medical diagnosis, where the goal is to detect human subjects who share common diseases to possibly predict future illnesses from previous medical history. The first proposed algorithm is based on K-medoids and the second algorithm extends the random swap algorithm, which has proven to be capable of efficient and careful clustering; both algorithms depend on a distance function among data objects (sets), which can use application-sensitive weights or priorities. The proposed distance function makes it possible to exploit several seeding methods that can improve clustering accuracy. A key factor in the two algorithms is their parallel implementation in Java, based on functional programming using streams and lambda expressions. The use of parallelism smooths out the O(N2) computational cost behind K-medoids and clustering indexes such as the Silhouette index and allows for the handling of non-trivial datasets. This paper applies the algorithms to several benchmark case studies of sets and demonstrates how accurate and time-efficient clustering solutions can be achieved.

Two Satellite Gene Clusters Enhance Ergot Alkaloid Biosynthesis Capacity of Aspergillus leporis.

Ergot alkaloids are fungal specialized metabolites that are important in agriculture and serve as sources of several pharmaceuticals. Aspergillus leporis is a soil saprotroph that possesses two ergot alkaloid biosynthetic gene clusters encoding lysergic acid amide production. We identified two additional, partial biosynthetic gene clusters within the A. leporis genome containing some of the ergot alkaloid synthesis (eas) genes required to make two groups of clavine ergot alkaloids, fumigaclavines and rugulovasines. Clavines possess unique biological properties compared to lysergic acid derivatives. Bioinformatic analyses indicated the fumigaclavine cluster contained functional copies of easA, easG, easD, easM, and easN. Genes resembling easQ and easH, which are required for rugulovasine production, were identified in a separate gene cluster. The pathways encoded by these partial, or satellite, clusters would require intermediates from the previously described lysergic acid amide pathway to synthesize a product. Chemical analyses of A. leporis cultures revealed the presence of fumigaclavine A. However, rugulovasine was only detected in a single sample, prompting a heterologous expression approach to confirm functionality of easQ and easH. An easA knockout strain of Metarhizium brunneum, which accumulates the rugulovasine precursor chanoclavine-I aldehyde, was chosen as expression host. Strains of M. brunneum expressing easQ and easH from A. leporis accumulated rugulovasine as demonstrated through mass spectrometry analysis. These data indicate that A. leporis is exceptional among fungi in having the capacity to synthesize products from three branches of the ergot alkaloid pathway and for utilizing an unusual satellite cluster approach to achieve that outcome. IMPORTANCE Ergot alkaloids are chemicals produced by several species of fungi and are notable for their impacts on agriculture and medicine. The ability to make ergot alkaloids is typically encoded by a clustered set of genes that are physically adjacent on a chromosome. Different ergot alkaloid classes are formed via branching of a complex pathway that begins with a core set of the same five genes. Most ergot alkaloid-producing fungi have a single cluster of genes that is complete, or self-sufficient, and produce ergot alkaloids from one or occasionally two branches from that single cluster. Our data show that Aspergillus leporis is exceptional in having the genetic capacity to make products from three pathway branches. Moreover, it uses a satellite cluster approach, in which gene products of partial clusters rely on supplementation with a chemical intermediate produced via another gene cluster, to diversify its biosynthetic potential without duplicating all the steps.

Cultural determinants of the gap between self-estimated navigation ability and wayfinding performance: evidence from 46 countries

Cognitive abilities can vary widely. Some people excel in certain skills, others struggle. However, not all those who describe themselves as gifted are. One possible influence on self-estimates is the surrounding culture. Some cultures may amplify self-assurance and others cultivate humility. Past research has shown that people in different countries can be grouped into a set of consistent cultural clusters with similar values and tendencies, such as attitudes to masculinity or individualism. Here we explored whether such cultural dimensions might relate to the extent to which populations in 46 countries overestimate or underestimate their cognitive abilities in the domain of spatial navigation. Using the Sea Hero Quest navigation test and a large sample (N = 383,187) we found cultural clusters of countries tend to be similar in how they self-rate ability relative to their actual performance. Across the world population sampled, higher self-ratings were associated with better performance. However, at the national level, higher self-ratings as a nation were not associated with better performance as a nation. Germanic and Near East countries were found to be most overconfident in their abilities and Nordic countries to be most under-confident in their abilities. Gender stereotypes may play a role in mediating this pattern, with larger national positive attitudes to male stereotyped roles (Hofstede's masculinity dimension) associated with a greater overconfidence in performance at the national level. We also replicate, with higher precision than prior studies, evidence that older men tend to overestimate their navigation skill more than other groups. These findings give insight into how culture and demographics may impact self-estimates of our abilities.

Redshift evolution of the baryon and gas fraction in simulated groups and clusters of galaxies

We study the redshift evolution of the baryon budget in a large set of galaxy clusters from the Magneticum suite of smoothed particle hydrodynamical cosmological simulations. At high redshifts (z ≳ 1), we obtain ‘closed-box’ (i.e. baryon mass fraction fbar = Ωbar/Ωtot) systems independently of the mass of the systems on radii greater than 3R500, c, whereas at lower redshifts, only the most massive halos can be considered closed box. Furthermore, in the innermost regions (r < R500, c), the baryon fraction shows a general decrease with redshift, and for less massive objects we observe a much more prominent decrease than for massive halos (fbar × Ωtot/Ωbar = Ybar decreases by ∼4% from z ∼ 2.8 to z ∼ 0.2 for massive systems and by ∼15% for less massive objects in the same redshift range). The gas depletion parameter Ygas = fgas/(Ωbar/Ωtot) shows a steeper and highly scattered radial distribution in the central regions (0.5R500, c ≤ r ≤ 2R500, c) of less massive halos with respect to massive objects at all redshifts, while on larger radii (r ≥ 2R500, c) the gas fraction distributions are independent of the masses or the redshifts. We divide the gas content of halos into the hot and cold phases. The hot, X-ray-observable component of the gas accurately traces the total amount of gas at low redshifts (e.g., for z ∼ 0.2 at R500, c, in the most massive subsample, that is, 4.6 × 1014 ≤ M500, c/M⊙ ≤ 7.5 × 1014 and least massive subsample, that is, 6.0 × 1013 ≤ M500, c/M⊙ ≤ 1.9 × 1014, we obtain Ygas ∼ 0.75 and 0.67, Yhot ∼ 0.73 and 0.64, and Ycold ∼ 0.02 and 0.02, respectively). On the other hand, at higher redshifts, the cold component provides a non-negligible contribution to the total amount of baryons in our simulated systems, especially in less massive objects (e.g., for z ∼ 2.8 at R500, c, in the most massive subsample, that is, 2.5 × 1013 ≤ M500, c/M⊙ ≤ 5.0 × 1013 and least massive subsample, that is, 5.8 × 1012 ≤ M500, c/M⊙ ≤ 9.7 × 1012, we obtain Ygas ∼ 0.63 and 0.64, Yhot ∼ 0.50 and 0.45, and Ycold ∼ 0.13 and 0.18, respectively). Moreover, the behaviour of the baryonic, entire-gas, and hot-gas-phase depletion parameters as functions of radius, mass, and redshift are described by some functional forms for which we provide the best-fit parametrization. The evolution of metallicity and stellar mass in halos suggests that the early (z > 2) enrichment process is dominant, while more recent star-formation processes make negligible contributions to the enrichment of the gas metallicity. In addition, active galactic nuclei (AGN) play an important role in the evolution of the baryon content of galaxy clusters. Therefore, we investigate possible correlations between the time evolution of AGN feedback and the depletion parameters in our numerical simulations. Interestingly, we demonstrate that the energy injected by the AGN activity shows a particularly strong positive correlation with Ybar, Ycold, and Ystar and a negative correlation with Yhot and ZTot. On the other hand, Ygas shows a less prominent level of negative correlation, a result which is highly dependent on the mass of the halos. These trends are consistent with previous theoretical and numerical works, meaning that our results, combined with findings derived from current and future X-ray observations, represent possible proxies with which to test the AGN feedback models used in different suites of numerical simulations.

Clusterome: A Comprehensive Data Set of Atmospheric Molecular Clusters for Machine Learning Applications.

Formation and growth of atmospheric molecular clusters into aerosol particles impact the global climate and contribute to the high uncertainty in modern climate models. Cluster formation is usually studied using quantum chemical methods, which quickly becomes computationally expensive when system sizes grow. In this work, we present a large database of ∼250k atmospheric relevant cluster structures, which can be applied for developing machine learning (ML) models. The database is used to train the ML model kernel ridge regression (KRR) with the FCHL19 representation. We test the ability of the model to extrapolate from smaller clusters to larger clusters, between different molecules, between equilibrium structures and out-of-equilibrium structures, and the transferability onto systems with new interactions. We show that KRR models can extrapolate to larger sizes and transfer acid and base interactions with mean absolute errors below 1 kcal/mol. We suggest introducing an iterative ML step in configurational sampling processes, which can reduce the computational expense. Such an approach would allow us to study significantly more cluster systems at higher accuracy than previously possible and thereby allow us to cover a much larger part of relevant atmospheric compounds.

A Comparative Analysis of Data Mining Methods and Hierarchical Linear Modeling Using PISA 2018 Data

A Comparative Analysis of Data Mining Methods and Hierarchical Linear Modeling Using PISA 2018 Data

A local universe model for constrained simulations

ABSTRACT The aim of cosmological simulations is to reproduce the properties of the observed Universe, serving as tools to test structure and galaxy formation models. Constrained simulations of our local cosmological region up to a few hundred $h^{-1}\, \mbox{Mpc}$ , the Local Universe, are designed to reproduce the actual cosmic web of structures as observed. A question that often arises is how to judge the quality of constrained simulations against the observations of the Local Universe. Here we introduce the Local Universe model (LUM), a new methodology, whereby many constrained simulations can be judged and the ‘best’ initial conditions can be identified. By characterizing the Local Universe as a set of rich clusters, the model identifies haloes that serve as simulated counterparts to the observed clusters. Their merit is determined against a null hypothesis, the probability that such a counterpart could be identified in a random, unconstrained simulation. This model is applied to 100 constrained simulations using the Cosmicflows-3 data. Cluster counterparts are found for all constrained simulations, their distribution of separation from the true observed cluster position and their mass distribution are investigated. Lastly, the ‘best’ constrained simulation is selected using the LUM and discussed in more detail.

Real-Time Anomaly Detection with Subspace Periodic Clustering Approach

Finding real-time anomalies in any network system is recognized as one of the most challenging studies in the field of information security. It has so many applications, such as IoT and Stock Markets. In any IoT system, the data generated is real-time and temporal in nature. Due to the extreme exposure to the Internet and interconnectivity of the devices, such systems often face problems such as fraud, anomalies, intrusions, etc. Discovering anomalies in such a domain can be interesting. Clustering and rough set theory have been tried in many cases. Considering the time stamp associated with the data, time-dependent patterns including periodic clusters can be generated, which could be helpful for the efficient detection of anomalies by providing a more in-depth analysis of the system. Another issue related to the aforesaid data is its high dimensionality. In this paper, all the issues related to anomaly detection are addressed, and a clustering-based approach is proposed for finding real-time anomalies. The method employs rough set theory, a dynamic k-means clustering algorithm, and an interval superimposition approach for finding periodic, partially periodic, and fuzzy periodic clusters in the subspace of the dataset. The data instances are thought to be anomalous if they either belong to sparse clusters or do not belong to any clusters. The efficacy of the method can be assessed by means of both time-complexity analysis and comparative studies with existing clustering-based anomaly detection algorithms on a synthetic and a real-life dataset. It can be found experimentally that our method outperforms others and runs in cubic time.

Intra- and Inter-Day Reliability of Inertial Loads with Cluster Sets When Performed during a Quarter Squat on a Flywheel Device.

The aims of this study were to (i) estimate the intra- and inter-day reliability of mean concentric (CON) and eccentric (ECC) power at different inertial loads during a flywheel quarter-squat using a cluster set approach and (ii) to determine the acute effect of internal and external attentional focus on mean power when performing the flywheel quarter squat. Twelve collegiate field sport male athletes (age 22.4 ± 3.2 years, weight 81.4 ± 10.3 kg, height 1.81 ± 0.06 m) attended four cluster set testing sessions separated by 7 days. Sessions consisted of 4 sets of 15 repetitions using 4 inertial loads (0.025, 0.050, 0.075, and 0.100 kg·m2). A cluster block consisted of 5 repetitions, including "momentum repetitions" (4 × 5 + 5 + 5). Mean power (MP), CON power, ECC power, and ECC overload were recorded for both internal and external attentional focus groups. The external instructional group attained familiarization after two flywheel sessions (ES = 0.03-0.15) with little volatility between performance measures (CV% = 3.39-9.22). The internal instructional group showed large differences in MP output from session 2 to session 3 for all loads (ES = 0.59-1.25). In conclusion, the flywheel cluster set approach is a reliable training modality for maintaining MP output during all repetitions.

Achieving feasibility for clustered traveling salesman problems using PQ‐trees

AbstractLet be a hypergraph, where is a set of vertices and is a set of clusters , , such that the clusters in are not necessarily disjoint. This article considers the feasibility clustered traveling salesman problem, denoted by . In the we aim to decide whether a simple path exists that visits each vertex exactly once, such that the vertices of each cluster are visited consecutively. We focus on hypergraphs with no feasible solution path and consider removing vertices from clusters, such that the hypergraph with the new clusters has a feasible solution path for . The algorithm uses a PQ‐tree data structure and runs in linear time.

Clustering retail stores for inventory transshipment

Rapid growth in the world of e-commerce is forcing traditional retailers to rethink their operational efficiency and revenue/cost streams. Retailers are facing increasing pressure to better utilize in-store inventory and reduce end-of-season markdowns. Transshipments between stores can re-balance inventory levels while preventing shortages and backorders. However, the joint inventory and transshipment problem is difficult to optimize in an extensive network, and many retailers lack the capital and managerial bandwidth required to create a system capable of transshipments between any two stores in a network of hundreds. We consider two multi-location newsvendor problems with reactive and proactive transshipments, respectively. We propose clustering stores into transshipment groups while considering both demand correlation and physical distance between locations. Since minimizing demand correlation does not follow the triangle inequality, we introduce two modified clustering algorithms that outperform existing algorithms for distance- and correlation-based clustering. For a given set of clusters, we mathematically model these two problems considering total system profit. In addition, we quantify the performance of distance-based and correlation-based clustering under several inventory ordering and transshipment policies to show the robustness of our results. To verify our methods, we test the proposed methodology using grocery sales data from IRi over three years in Houston and Dallas, TX and liquor licensee purchase information over three years from the Iowa state government. We present correlation, transshipment, and profitability analysis. We show that when demand correlation is present and stable over time, clustering based on demand correlation can maintain high transshipments and profits while decreasing solution complexity.

Mapping hotspot clustering: An approach to study the spatial pattern of quality of living households in India

This study provides insight into the regional disparity of households’ quality of living standards in India through the selective indicators. A household quality of living index (HQI) which is called a composite index was developed by integrating four indices which are housing facility index (HFI-1), basic facility index (BFI-2), financial asset index (FAI-3) and human capital index (HCI-4). These four indices have been developed from 23 census indicators through entropy techniques. The research findings indicate that the central (Chhattisgarh, Madhya Pradesh, Uttar Pradesh), eastern (Odisha, Jharkhand), and northeastern region (Arunanchal Pradesh and Nagaland) have poor living standards, and these are highly influenced by the basic facility index and financial asset index. Further, the indicators like concrete materials, radio, computer, two-wheeler, four-wheeler and drainage systems are totally in the poor category for almost 95% of the districts of India. Accordingly, hotspot GIS maps were generated and these maps explored that the 24.6% (157 districts) of the study region covered by hotspot showing poor quality of living. Nearly 14 states covered by hotspots, in which Bihar has the highest hotspot district (29), followed by Odisha (24), Madhya Pradesh (22) and Jharkhand (20). Additionally, three set of hotspot clusters were created for the developmental purpose: Cluster 1: (Odisha, Chhattisgarh and Madhya Pradesh), cluster 2: (Bihar, Jharkhand and West Bengal) and cluster 3: (Assam and Meghalaya). Where cluster 1 needs immediate attention followed by cluster 2 and 3. The current study results certainly assists the regional and national policy and decision makers to implement the development plan in hotspot clusters to enhance the quality of living.

The fractal or scaling perspective on progressively generated intra-urban clusters from street junctions

ABSTRACT The underlying complexity of urban space can be manifested by its fractal forms and scaling statistics. This paper examines these characteristics at the intra-urban scale through the lens of clustered street junctions (including road ends) in two Chinese metropolitan areas: Beijing and Shenzhen. We derived the cluster sets with Euclidean distance thresholds starting at 100 meters (m) and ending at 1000 m, and outlined each cluster using a concave-hull method to maintain their original irregular shapes. Within each delimited cluster, we examined four urban attributes: gross domestic product, number of street nodes, polygon area, and population. Our analysis revealed that power law distribution applied to almost every cluster set in terms of the four attributes, but varied from one attribute to another or from city to city, represented primarily by fluctuated power law exponents and ht-index values whose profiles along with the cluster growth can effectively characterize the urban structure. Additionally, we computed the spectrum of intra-urban scaling exponents with cluster size increments, contributing new insights into the allometric relationships between urban configuration and function.