Unsupervised Data Research Articles

Risk and Pattern Analysis of Pakistani Crime Data Using Unsupervised Learning Techniques

To reduce crime rates, there is a need to understand and analyse emerging patterns of criminal activities. This study examines the occurrence patterns of crimes using the crime dataset of Lahore, a metropolitan city in Pakistan. The main aim is to facilitate crime investigation and future risk analysis using visualization and unsupervised data mining techniques including clustering and association rule mining. The visualization of data helps to uncover trends present in the crime dataset. The K-modes clustering algorithm is used to perform the exploratory analysis and risk identification of similar criminal activities that can happen in a particular location. The Apriori algorithm is applied to mine frequent patterns of criminal activities that can happen on a particular day, time, and location in the future. The data were acquired from paper-based records of three police stationsin the Urdu language. The data were then translated into English and digitized for automatic analysis. The result helped identify similar crime-related activities that can happen in a particular location, the risk of potential criminal activities occurring on a specific day, time, and place in the future, and frequent crime patterns of different crime types. The proposed work can help the police department to detect crime events and situations and reduce crime incidents in the early stages by providing insights into criminal activity patterns.

A Recruitment Big Data Approach to interplay of the Target Drugs

Abstract The various model that has been used to predict, datamining, and information retrieval are useful to use through the traditional database, due to big data the prediction should derive in a different role that conduct the hidden structure data based on a stability scale to allow discovering accrue unsupervised drug data. Especially, the drug data must be understandable to analysts. Following this approach, conduct the stability drug data through computation methods are quality measurements, preprocess data, k-mean cluster, and decision tree. This approach seeks to identify the data by two dimensions (vertically and horizontally), which extrapolations, compilation, and interpretation values of the dataset while considering individual attributes. A comparison with clusters defines the set for features using balance value by K-mean algorithm to determine the k clusters that consider the set of features based on two values 0 and 1, which given the discernible between dependent and independent class target, and pinpoint the relationship among them. Keywords: Big Data, Discretize, k-mean cluster Stability, Target drug

Machine-Learning Analysis of Serum Proteomics in Neuropathic Pain after Nerve Injury in Breast Cancer Surgery Points at Chemokine Signaling via SIRT2 Regulation.

Background: Persistent postsurgical neuropathic pain (PPSNP) can occur after intraoperative damage to somatosensory nerves, with a prevalence of 29–57% in breast cancer surgery. Proteomics is an active research field in neuropathic pain and the first results support its utility for establishing diagnoses or finding therapy strategies. Methods: 57 women (30 non-PPSNP/27 PPSNP) who had experienced a surgeon-verified intercostobrachial nerve injury during breast cancer surgery, were examined for patterns in 74 serum proteomic markers that allowed discrimination between subgroups with or without PPSNP. Serum samples were obtained both before and after surgery. Results: Unsupervised data analyses, including principal component analysis and self-organizing maps of artificial neurons, revealed patterns that supported a data structure consistent with pain-related subgroup (non-PPSPN vs. PPSNP) separation. Subsequent supervised machine learning-based analyses revealed 19 proteins (CD244, SIRT2, CCL28, CXCL9, CCL20, CCL3, IL.10RA, MCP.1, TRAIL, CCL25, IL10, uPA, CCL4, DNER, STAMPB, CCL23, CST5, CCL11, FGF.23) that were informative for subgroup separation. In cross-validated training and testing of six different machine-learned algorithms, subgroup assignment was significantly better than chance, whereas this was not possible when training the algorithms with randomly permuted data or with the protein markers not selected. In particular, sirtuin 2 emerged as a key protein, presenting both before and after breast cancer treatments in the PPSNP compared with the non-PPSNP subgroup. Conclusions: The identified proteins play important roles in immune processes such as cell migration, chemotaxis, and cytokine-signaling. They also have considerable overlap with currently known targets of approved or investigational drugs. Taken together, several lines of unsupervised and supervised analyses pointed to structures in serum proteomics data, obtained before and after breast cancer surgery, that relate to neuroinflammatory processes associated with the development of neuropathic pain after an intraoperative nerve lesion.

Unsupervised Multi-Omics Data Integration Methods: A Comprehensive Review

Through the developments of Omics technologies and dissemination of large-scale datasets, such as those from The Cancer Genome Atlas, Alzheimer’s Disease Neuroimaging Initiative, and Genotype-Tissue Expression, it is becoming increasingly possible to study complex biological processes and disease mechanisms more holistically. However, to obtain a comprehensive view of these complex systems, it is crucial to integrate data across various Omics modalities, and also leverage external knowledge available in biological databases. This review aims to provide an overview of multi-Omics data integration methods with different statistical approaches, focusing on unsupervised learning tasks, including disease onset prediction, biomarker discovery, disease subtyping, module discovery, and network/pathway analysis. We also briefly review feature selection methods, multi-Omics data sets, and resources/tools that constitute critical components for carrying out the integration.

A Contrastive Predictive Coding-Based Classification Framework for Healthcare Sensor Data.

Supervised learning technologies have been used in medical-data classification to improve diagnosis efficiency and reduce human diagnosis errors. A large amount of manually annotated data are required for the fully supervised learning process. However, annotating data information will consume a large amount of manpower and resources. Self-supervised learning has great advantages in solving this problem. Self-supervised learning mainly uses pretext tasks to mine its own supervised information from large-scale unsupervised data. And this constructed supervised information is used to train the network to learn valuable representations for downstream tasks. This study designs a general and efficient model for the diagnosis and classification of medical sensor data based on contrastive predictive coding (CPC) in self-supervised learning, called TCC, which consists of two steps. The first step is to design a pretext task based on the idea of CPC, which aims to extract effective features between different categories using its encoder. The second step designs a downstream classification task with lower time and space complexity to perform a supervised type of training using the features extracted by the encoder of the pretext task. Finally, to demonstrate the performance of the proposed framework in this paper, we compare the proposed framework with recent state-of-the-art works. Experiments comparing the proposed framework with supervised learning are also set up under the condition of different proportions of labeled data.

Data analytics and Bayesian Optimised Extreme Gradient Boosting approach to estimate cut-offs from wireline logs for net reservoir and pay classification

Accurate net pay classification is essential in hydrocarbon resource volumetric calculation. However, there is no universal methodology developed for its evaluation hence the existence of many incongruent views on its application since it is data-driven and differs for each reservoir. This research incorporates machine learning and data analytics in predicting net pay, intending to reduce uncertainties associated with the net-pay classification. Log analysis was performed to determine the cut-offs for sonic, neutron, density, and gamma-ray using unsupervised learning and data analytics. The log cut-offs were calculated with petrophysical properties; shale volume, water saturation, permeability, and porosity. A Bayesian Optimised Extreme Gradient Boosting (Bayes Opt-XGBoost) model was applied to predict the petrophysical properties using five wireline logs. The model’s performance and a computational function in classifying net reservoir resulted in an accuracy of 0.93, a combined precision of 0.94, a combined recall of 0.92, and a combined F1-score of 0.93. The model and methodology were deployed on a new well for validation. The classification of net reservoir zones via the proposed data analytics method, Bayes Opt-XGBoost predicted petrophysical properties, and computational function code matched mobility drawdown test data for the well. These results indicate that the developed methodology and machine learning model can work for other reservoirs since the additional computational function code can be manipulated for any data-driven estimated cut-offs. This developed approach can determine net reservoir and net pay zones in any sandstone reservoir.

Enhancing Seismic Data Interpretation through Unsupervised Machine Learning and Data Analytics for Improved Reservoir Characterization

Enhancing Seismic Data Interpretation through Unsupervised Machine Learning and Data Analytics for Improved Reservoir Characterization

FWDGAN-based data augmentation for tomato leaf disease identification

There are few publicly available large-scale datasets for tomato leaf disease detection using deep learning, the workload associated with manual data collection is high, and conventional supervised data augmentation methods (such as image rotation, flipping, and shifting) are challenging to attain satisfactory results with. While Deep Convolutional Generative Adversarial Networks (DCGAN) is a popular unsupervised data augmentation method based on deep learning, the data augmentation method based on DCGAN may have several drawbacks, including poor image quality and excessive computing resource consumption. Given this, a method based on Fast WDBlock based GAN (FWDGAN) was proposed in this paper. For the network’s generator, a wide and deep feature extraction block (WDBlock) with a two-path strategy was designed, combining the extracted depth feature based on ResNet and the extracted global feature based on InceptionV1. By incorporating WDBlock into the generator, the quality of the generated tomato leaf disease images was improved. For the network’s discriminator, the Depthwise separable convolution Discriminator (DSC-Discriminator) that significantly reduced the model's parameters without impairing the network's performance was constructed. Finally, the SeLU activation function was used selectively to improve the training stability of the network. Comparative experiments demonstrated that FWDGAN could generate higher-quality data, with FID scores of 193.998, 264.704, 260.594, and 161.436 for healthy tomato leaf image, Leaf Mold, Septoria leaf spot, and Yellow Leaf Curl Virus generated by FWDGAN, respectively. Furthermore, the total number of parameters in FWDGAN was approximately one-third less than that of DCGAN.

Unsupervised IMU-based evaluation of at-home exercise programmes: a feasibility study

BackgroundThe benefits to be obtained from home-based physical therapy programmes are dependent on the proper execution of physiotherapy exercises during unsupervised treatment. Wearable sensors and appropriate movement-related metrics may be used to determine at-home exercise performance and compliance to a physical therapy program.MethodsA total of thirty healthy volunteers (mean age of 31 years) had their movements captured using wearable inertial measurement units (IMUs), after video recordings of five different exercises with varying levels of complexity were demonstrated to them. Participants were then given wearable sensors to enable a second unsupervised data capture at home. Movement performance between the participants’ recordings was assessed with metrics of movement smoothness, intensity, consistency and control.ResultsIn general, subjects executed all exercises similarly when recording at home and as compared with their performance in the lab. However, participants executed all movements faster compared to the physiotherapist’s demonstrations, indicating the need of a wearable system with user feedback that will set the pace of movement.ConclusionIn light of the Covid-19 pandemic and the imperative transition towards remote consultation and tele-rehabilitation, this work aims to promote new tools and methods for the assessment of adherence to home-based physical therapy programmes. The studied IMU-derived features have shown adequate sensitivity to evaluate home-based programmes in an unsupervised manner. Cost-effective wearables, such as the one presented in this study, can support therapeutic exercises that ought to be performed with appropriate speed, intensity, smoothness and range of motion.

Multi-mode data augmentation and fault diagnosis of rotating machinery using modified ACGAN designed with new framework

As one of the representative unsupervised data augmentation methods, generative adversarial networks (GANs) have the potential to solve the problem of insufficient samples in fault diagnosis of rotating machinery. However, the existing unsupervised GANs are usually incapable of simultaneously generating multi-mode fault samples and have some shortcomings such as mode collapse and gradient vanishing. To overcome these deficiencies, a supervised model called modified auxiliary classifier GAN (MACGAN) designed with new framework is proposed in this paper. Firstly, a new ACGAN framework is developed by adding an independent classifier to improve the compatibility between the classification and discrimination. Secondly, the Wasserstein distance is introduced in the new loss functions to overcome mode collapse and gradient vanishing. Finally, to achieve stable training, a spectral normalization is used to replace the weight clipping to constrain the weight parameters of discriminator. The proposed method is applied to fault diagnosis of bearing and gear. Compared with the existing GANs, the proposed method can more efficiently generate multi-mode fault samples with higher qualities, which can be used to assist the training of deep learning-based fault diagnosis models with high accuracy and good stability.

Empowering those who seek to end hunger through collaboration and data science innovation

Feeding America Eastern Wisconsin (FAEW) distributed 81% more food to community members in need during the COVID-19 pandemic than in the prior year. Though systems were in place to manage food receipt and distribution data, social distancing requirements and technological barriers revealed inefficiencies in how these data were being utilized. In pursuit of data-driven decision making and in the context of a global pandemic, FAEW partnered with Marquette University data scientists through an industry supported grant. Applying newly learned skills in Business Intelligence, students have produced detailed reports of data cleanliness in FAEW’s source systems to improve underlying data quality and better support analytic efforts. Additionally, students have synchronized Human Centered Design Thinking and Visual Analytics to produce an interactive application that is being used to optimize inventory management, storage availability, and product distribution. Finally, students are utilizing Business Analytics techniques such as supervised and unsupervised data mining to provide new insights about food receipt and distribution patterns that will have a sustainable impact on FAEW operations. This unique partnership is providing experiential learning opportunities for students, tangible data science solutions that FAEW will use to ensure best practices, and real-world solutions to collaboratively end hunger in our communities.

Modelling personalised car-following behaviour: a memory-based deep reinforcement learning approach

To adapt to human-driving habits, this study develops a personalised car-following model via a memory-based deep reinforcement learning approach. Specifically, Twin Delayed Deep Deterministic Policy Gradients (TD3) is integrated with a long short-term memory (LSTM) (abbreviated as LSTM-TD3). Using the NGSIM dataset, unsupervised learning-based clustering and data feature analyses are performed. The driving characteristics related to safety, efficiency and comfort are extracted for different driving styles, i.e. aggressive, common and conservative. Then, reward functions are constructed for different driving styles by incorporating their driving characteristics. By resorting to the TD3 policy within a recurrent actor–critic framework, LSTM-TD3 optimises the car-following behaviour via trial-and-error interactions according to the reward functions. Results show that compared with LSTM-DDPG and DDPG, LSTM-TD3 reproduces personalised car-following behaviour with desirable convergence speed and reward. It reveals that LSTM-TD3 can reflect the essential difference in safety, efficiency and comfort requirements among different driving styles.

Direct-laser writing for subnanometer focusing and single-molecule imaging

Two-photon direct laser writing is an additive fabrication process that utilizes two-photon absorption of tightly focused femtosecond laser pulses to implement spatially controlled polymerization of a liquid-phase photoresist. Two-photon direct laser writing is capable of nanofabricating arbitrary three-dimensional structures with nanometer accuracy. Here, we explore direct laser writing for high-resolution optical microscopy by fabricating unique 3D optical fiducials for single-molecule tracking and 3D single-molecule localization microscopy. By having control over the position and three-dimensional architecture of the fiducials, we improve axial discrimination and demonstrate isotropic subnanometer 3D focusing (<0.8 nm) over tens of micrometers using a standard inverted microscope. We perform 3D single-molecule acquisitions over cellular volumes, unsupervised data acquisition and live-cell single-particle tracking with nanometer accuracy.

Deep learning approaches for imaging fluorescence correlation spectroscopy parameter estimation with limited data sets

Deep Learning approaches using Convolutional Neural Networks (CNNs) are successfully applied in many disciplines and especially in imaging applications. CNNs provide advantages in terms of data processing time, required data size to reach a particular precision in parameter estimation, as well as the possibility of unsupervised data evaluation. Here we propose two different approaches to estimate diffusion coefficients from Imaging Fluorescence Correlation Spectroscopy (Imaging FCS) experiments and compare their performance with standard non-linear least squares (NLS) fits.

Unsupervised Data Mining and Effect of Fast Rehabilitation Nursing Intervention in Fracture Surgery.

At present, the most commonly used surgical treatment for fractures caused by external force injury is clinical, and unsupervised data mining is more advantageous in the face of the unknown format of perioperative network data. Therefore, this research aims to explore the application effect of unsupervised data mining in the concept of rapid rehabilitation nursing intervention after fracture surgery. 80 patients who underwent fracture surgery in the Department of Orthopedics of XXX Hospital were determined as the subjects, who were rolled into a research group (group R, 40 cases) and a control group (group C, 40 cases) by drawing lots. An unsupervised data mining algorithm based on unsupervised data mining for support vector machines (VDMSVMs) was proposed and applied to two groups of patients undergoing perioperative fracture surgery with the rapid rehabilitation nursing intervention and basic routine nursing. The results showed that the number of important features selected by the VDMSVM algorithm (5) was obviously more than that of the compressed edge fragment sampling (CEFS) algorithm (1) and the multicorrelation forward searching (MCFS) algorithm (2) (P < 0.05). The number of noise features screened by the VDMSVM algorithm (3) was much less in contrast to that of the CEFS algorithm and the MCFS algorithm, which was 8 and 10, respectively (P < 0.05). The Visual Analogue Scale (VAS) scores of the fracture site at the 4th, 8th, 12th, and 16th hour after surgery in group R were all lower than the scores in group C (P < 0.05). The length of hospital stay (LoHS) in group R was greatly shorter than that in group C (P < 0.05). After different nursing methods, the World Health Organization Quality of Life (WHOQOL-BREF) score of patients in group R (89.64 points) was greatly higher than the score in group C (61.45 points) (P < 0.05). The nursing satisfaction score of group R was 92.35 ± 3.65 points, and that in group C was 2.14 ± 1.25 points, respectively (P < 0.05). The test results verified the effectiveness of the feature selection of the VDMSVM algorithm. The rapid rehabilitation nursing intervention was conductive to reducing the postoperative pain of fracture patients, shortening the LoHS of patients, improving the quality of life (QOL) of fracture surgery patients, and increasing the patient's satisfaction with nursing.

Measuring colorant fading within raster regions of printed scanned customer content using a novel unsupervised clustering method

Colorant fading defects in raster region of interest (ROI) are among the most common printing issues in electrophotographic printers. Colorant fading manifests as faint print or customer content and is usually caused by low-level ink/cartridge. This paper presents an accurate method to detect the colorant fading defects and classify the defects based on their severity. There are two modules for this method. The first module uses the SLIC super-pixel method to separate the raster ROI and extract the smooth super-pixels. We design a novel unsupervised clustering algorithm to automatically extract the three or four main colors from the smooth super-pixels. Unsupervised data clustering is an important problem that arises in many image processing applications. We propose a new approach to unsupervised data clustering called Data Inter-Distance MEdiated Clustering (DIDMEC). Our new method is based on analyzing the matrix of Euclidean distances between each pair of points in the data set. Based on three simple properties, we devise an approach that effectively yields the same accuracy as the K-Means algorithm but at a much lower computational cost for a moderate number of sample points. The second module extracts feature vectors for each main color clustered by DIDMEC to classify the colorant fading defects based on their severity.

Multisensor hyperspectral imaging approach for the microchemical analysis of ultramarine blue pigments

This work presents a multisensor hyperspectral approach for the characterization of ultramarine blue, a valuable historical pigment, at the microscopic scale combining the information of four analytical techniques at the elemental and molecular levels. The hyperspectral images collected were combined in a single hypercube, where the pixels of the various spectral components are aligned on top of each other. Selected spectral descriptors have been defined to reduce data dimensionality before applying unsupervised chemometric data analysis approaches. Lazurite, responsible for the blue color of the pigment, was detected as the major mineral phase present in synthetic and good quality pigments. Impurities like pyrite were detected in lower quality samples, although the clear identification of other mineral phases with silicate basis was more difficult. There is no correlation between the spatial distribution of the bands arising in the Raman spectra of natural samples in the region 1200–1850 cm−1 and any of the transition metals or rare earth elements (REE). With this information, the previous hypothesis (based on bulk analysis) attributing these bands to luminescence emissions due to impurities of these elements must be revised. We propose the consideration of CO2 molecules trapped in the cages of the aluminosilicate structure of sodalite-type. Additionally, correlation between certain Raman features and the combined presence of Ca, P, and REE, in particular Nd, was detected for the lowest quality pigment. Our results highlight the usefulness of fusing chemical images obtained via different imaging techniques to obtain relevant information on chemical structure and properties.

Picking out the bad apples: unsupervised biometric data filtering for refined age estimation

Introduction of large training datasets was essential for the recent advancement and success of deep learning methods. Due to the difficulties related to biometric data collection, facial image datasets with biometric trait labels are scarce and usually limited in terms of size and sample diversity. Web-scraping approaches for automatic data collection can produce large amounts of weakly labeled and noisy data. This work is focused on picking out the bad apples from web-scraped facial datasets by automatically removing erroneous samples that impair their usability. The unsupervised facial biometric data filtering method presented in this work greatly reduces label noise levels in web-scraped facial biometric data. Experiments on two large state-of-the-art web-scraped datasets demonstrate the effectiveness of the proposed method with respect to real and apparent age estimation based on five different age estimation methods. Furthermore, we apply the proposed method, together with a newly devised strategy for merging multiple datasets, to data collected from three major web-based data sources (i.e., IMDb, Wikipedia, Google) and derive the new Biometrically Filtered Famous Figure Dataset or B3FD. The proposed dataset, which is made publicly available, enables considerable performance gains for all tested age estimation methods and age estimation tasks. This work highlights the importance of training data quality compared to data quantity and selection of the estimation method.

Machine learning of phase transitions in nonlinear polariton lattices

Polaritonic lattices offer a unique testbed for studying nonlinear driven-dissipative physics. They show qualitative changes of their steady state as a function of system parameters, which resemble non-equilibrium phase transitions. Unlike their equilibrium counterparts, these transitions cannot be characterised by conventional statistical physics methods. Here, we study a lattice of square-arranged polariton condensates with nearest-neighbour coupling, and simulate the polarisation (pseudospin) dynamics of the polariton lattice, observing regions with distinct steady-state polarisation patterns. We classify these patterns using machine learning methods and determine the boundaries separating different regions. First, we use unsupervised data mining techniques to sketch the boundaries of phase transitions. We then apply learning by confusion, a neural network-based method for learning labels in a dataset, and extract the polaritonic phase diagram. Our work takes a step towards AI-enabled studies of polaritonic systems.

Abstract PO-140: Unsupervised clustering and data modeling reveals molecular signatures linked a distinct African American enriched cluster with higher probability of death in triple negative breast cancer

Abstract Triple-negative breast cancer is considered the most aggressive of all breast cancers, with the worst prognosis and increased risk in African Americans. This study aims to characterize proteomic, transcriptomic, and genomic signatures using tissue microarrays from resected TNBC tumors and correlate them with clinical data in an integrated approach to query immune mechanisms and predict patient survival. To this end, our cohort consisted of 118 localized, non-metastatic, resected TNBC patients with diverse demographics (African American 23% (n=28), Hispanic/Latino 9% (n=11), White/Caucasian 36% (n=43) and Other 30%(n=36)) with a median follow-up of 85 months. Here, we studied the patient transcriptomic profiles generated using NanoString sequencing. Immune cell subsets within the tumor microenvironment were assessed on tumor microarray slides by Multiplex Immunohistochemistry Consecutive on Single Slide (MICSSS). Protein markers measured include PD-L1, CD3, and CD8, CD20, CD66b, FOXP3, DC-LAMP, TLS, CD68, and CD163. We utilized unsupervised clustering of the patient transcriptomes as a data-driven approach to identify patient clusters defined by co-expressed RNA molecules. Then, we performed differential expression using mixed linear models for both data types (RNA &amp; Protein), to compare these clusters to each other identifying cluster-specific molecular signatures and association to outcomes. This unbiased modeling strategy enabled us to include critical covariates in the analysis such as demographics, neoadjuvant treatment, clinical tumor parameters, and quantify the effect of potential confounders. Finally, these results revealed a cluster of patients with the best survival associated with high expression of genes such as GZMK/A, CCR5, IL10RA, IL2RG, and the highest levels of CD8, PD-L1, CD163, FOXP3, and CD68 protein tissue markers. By contrast, the cluster with the worst outcomes was enriched in African Americans (44% n=8/18, p.value&amp;lt;0.05, compared to an average of 16% (n=8/25) in other clusters). Furthermore, this cluster included increased BRCA1, MYC, KIF2C, BIRC5, and HMGA1 gene expression, with an absence of immune proteomic markers, and it was independent of chemotherapy type, histology scores, tumor status, or lymphatic vessel invasion. Genomic alteration analyses are pending. In conclusion, we are using a data-driven approach to characterize patients with triple-negative breast cancer and patient clusters by integrating proteomic and transcriptomic molecular tumor profiles, and we identified a gene cluster enriched in African Americans associated with worse outcomes and poor immune infiltration. By combining a wider immune tissue characterization using an extended panel of MICSSS markers together with genetic mutational and expression profiling of tumors, we expect to refine and distinguish subsets of high-risk TNBC patients for whom more aggressive tailored treatment regimens may be indicated. Citation Format: Edgar Gonzalez-Kozlova, Clelia Chalumeau, Ilaria LaFace, Charles Shapiro, Guray Akturk, Diane Marie Del Valle, Sacha Gnjatic. Unsupervised clustering and data modeling reveals molecular signatures linked a distinct African American enriched cluster with higher probability of death in triple negative breast cancer [abstract]. In: Proceedings of the AACR Virtual Conference: 14th AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2021 Oct 6-8. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2022;31(1 Suppl):Abstract nr PO-140.