Data Transformation Methods Research Articles

Long-Term Monitoring of Vegetation Dynamics in the Rhodopi Mountain Range National Park-Greece

While deforestation is a major environmental issue in the tropics, with thousands of hectares converted to agricultural land every year, in Europe the opposite trend is observed, with land abandonment in mountainous and semi-mountainous areas allowing the afforestation of former agricultural and pastoral land. This trend allows semi-natural ecosystems to recover after a prolonged period of exploitation and often over-exploitation, but it may also lead to significant loss of landscape heterogeneity with potentially detrimental effects on biodiversity. The current study aims to monitor changes in the vegetation coverage across a period of 35 years (between 1984 and 2019) in the Rhodopi Mountains range National Park in northern Greece. A time series of LANDSAT TM (16 images), LANDSAT ETM + (1 image) and LANDSAT 8 OLI/TIRS (4 images) were employed. One data transformation method was applied (TCT), and five vegetation indices (NDVI, NDWI, SAVI, EVI2 and BSI) were calculated to capture the land cover transition during the study period. The obtained results and all used indices suggest that over the study period there was a continuous trend of vegetation cover increasing, with open areas decreasing. The observed trend was further confirmed using Object Oriented Image Analysis on two pairs of images sensed in 1984 and 2019, respectively. The results suggest that almost 22.000 ha of open habitats have been lost to broadleaved and conifer woodlands, while the former also appear to be advancing into conifer-covered areas. This trend has led to significant loss of landscape heterogeneity and to a broadleaf-dominated landscape. The results are discussed in relation to their driving forces, the potential implications on biodiversity and the risk of wildfires in the near future.

Prediction of wax deposition thickness on pipe wall by improved GM (1,1) model based on data transformation method

The improved GM (1,1) models for predicting wax deposition thickness on the pipe wall under different deposition times were established by introducing the idea of square root transformation and logarithmic transformation into the modeling step of GM (1,1), the prediction accuracy of the established models and the traditional GM (1,1) modelwere compared and analyzed, and the influences of the change of constant C on the fitting and prediction accuracy of the model were discussed. The calculation results of different examples show that: After the square root transformation (improved model I) and logarithmic transformation (improved model II), the maximum relative errors between the fitting results of the improved models and the actual values are lower than those of the traditional model. The prediction results of the improved model are in good agreement with the actual values, and the prediction accuracy of the improved model is higher than that of the traditional model. The change of C value has an important impact on the prediction accuracy of the improved model, the increase of C value does not mean that the prediction accuracy of the improved model is higher, the reasonable C value can make the model have higher prediction accuracy.

Micro-/Nano-Texturing onto Plasma-Nitrided Tool Surface by Laser Printing for CNC Imprinting and Piercing.

A new data transformation method for micro-manufacturing using a topological model for a micro-/nano-texture was proposed for a surface-decorated product. Femtosecond laser printing was utilized to form the micro-/nano-textures into the hardened thick layer of dies by plasma nitriding. At first, the plasma-nitrided AISI316L flat substrate was laser-printed as a punch to imprint the tailored nano-textures onto the AA1060 aluminum plate for its surface decoration with topological emblems. Second, the plasma-nitrided SKD11 cylindrical punch was laser-trimmed to form the nanostructures on its side surface. This nano-texture was imprinted onto the hole surface concurrently with piercing a circular hole into electrical steel sheet. The fully burnished surface had a shiny, metallic quality due to the nano-texturing. The plasma nitriding, the laser printing and the CNC (computer numerical control) imprinting provided a way of transforming the tailored textures on the metallic product.

3-D CNN-Based Multichannel Contrastive Learning for Alzheimer’s Disease Automatic Diagnosis

Alzheimer’s disease (AD) is a common progressive neurodegenerative disease in the elderly. Mild cognitive impairment (MCI) is the symptomatic predementia stage of AD. Accurately distinguishing AD and MCI patients from normal people is the first step of the disease diagnosis. Several studies have demonstrated the potential of deep learning in the automatic diagnosis of AD and MCI using T1-weighted magnetic resonance imaging (MRI) images. In this article, we proposed an automatic classification method of AD versus normal control (NC) and MCI versus NC based on MRI images. This method used the 3-D convolutional neural network and took the whole 3-D MRI image as the input, which can obtain image information to the greatest extent. In addition, the multichannel contrastive learning strategy based on multiple data transformation methods (e.g., add noise) can combine the supervised classification loss with the unsupervised contrastive loss, which can further improve the classification accuracy and generalization ability of the network. To verify the effectiveness of our method, a large number of experiments were implemented on the ADNI dataset. The results show that our method can achieve excellent performance in accurate diagnosis of AD and MCI; the multichannel contrastive learning strategy can greatly improve the classification accuracy (AD versus NC: 4.19%; MCI versus NC: 4.57%) and generalization ability of the network.

Abstract 12379: Distal Waveform Patterns in the Fontan Circulation Are Related to Pulmonary, Liver and Cardiac Function

Introduction: The lack of a sub-pulmonary ventricle in patients with a Fontan circulation leads to irreversible circulatory decline. Our group has determined that single-site waveform analysis correlates to clinical features of Fontan patients. The goal of this study was to determine if multi-site waveform analysis contributes to detecting early cardiovascular decline. Methods: Patients with a Fontan circulation that underwent cardiac MRI were included in this retrospective study. Hemodynamic flow curves were measured in the ascending aorta (AAo N=94), superior vena cava (SVC N=82), inferior vena cava (IVC N=91) and left pulmonary artery (LPA N=85). Flow curves were interpolated to equal temporal phases and were used to create three data matrices of proximal and distal waveforms. Principal component analysis (PCA), a data transformation method used to obtain a wieldy number of ranked components (the PCs) representing the variance of the input data, was performed on each matrix. PCs were correlated to clinical measures of current patient status. Results: PCs that accounted for more than 20% of the variance in a data matrix are reported (Pearson’s r, p ) in Table 1. Conclusions: These results indicate that increased liver enzyme levels, with a possible link to Fontan associated liver disease, are related to increased diastolic SVC flow and increased volume in the SVC and IVC throughout the cardiac cycle. An increased diastolic SVC flow profile relates to worsening pulmonary function and oxygen consumption during exercise. A decreased ventricular-vascular coupling ratio is related to increased IVC volume and low oxygen saturations may be indicative of increased LPA volume. Little variations were apparent in the AAo waveforms, suggesting cardiovascular decline may appear distally prior to proximal presentation. We have, for the first time, evaluated multi-site flow patterns and correlated them to functional status in one of the largest Fontan patient cohorts to date.

Comparison and Analysis of the Influence of Different Data Transformation Methods on the Fault Identification of Flexible DC Transmission Lines by Convolutional Neural Network

In the fault classification and identification of flexible DC transmission lines, it is inevitable to use the voltage and current characteristics of the transmission line. All kinds of data transformation methods can highlight the hidden characteristics of the original fault electrical quantity. Various artificial intelligence algorithms can further reduce the difficulty of transmission line fault classification. For such fault classification methods, this paper first builds a four-terminal flexible direct current transmission system model on PSCAD/EMTDC platform and obtains data by simulating different faults of transmission lines. Then, empirical mode decomposition (EMD), wavelet transform (WT), fast Fourier transform (FFT), and variational mode decomposition (VMD) are performed on the obtained data, respectively. Finally, the transformed data and original data are used as inputs to classify by convolutional neural network (CNN). The influence of one data transformation method and different combinations of two data transformation methods on CNN classification results is explored. The simulation results show that when only one data transformation method is used, CNN has the best classification effect for the data after VMD transformation. The classification accuracy and recall rate are both increased from 96.9% and 96.3% without data transformation to 99.88%. When VMD and FFT are combined, CNN classification results’ accuracy and recall rate are further improved to 99.96%.

Prediction of Cardiovascular Disease on Different Parameters Using Machine Learning

The most common serious diseases affecting human health are cardiovascular diseases (CVDs). Early diagnosis can prevent or mitigate CVDs, which can reduce the rate of death. It's a promising approach to identify risk factors using machine learning models. We wish to propose a model with different methods to effectively predict heart disease. We have employed effective data collection, data pre-processing and data transformation methods for the precise information of our training model to make our proposed model a success. A combined dataset has been used (Cleveland, Long Beach VA, Switzerland, Hungarian and Stat log). The appropriate function is selected using AASSO (Advanced Absolute Shrinkage and Selection Operator techniques) and AASSO techniques. Appropriate features are selected. New hybrids are developed with integration of the traditional bagging and boosting methods, such as Decision Tree Bagger Method (DTBM), the Random Forest Bagging Method (RFBM), the K-Nearest Neighbour Bagging method (KNNBM), the AdaBoost Boosting Method (ABBM), and the GBBM. Our machine learning algorithms, along with Negative Predictive Value (NGR, false positive rates), and false negative flow rates, also were implemented to calculate accuracy of our model, sensitivity (SEN), error rate, accuracy of the model (FRE) and the F1 score (F1) (FNR). The results are shown for comparisons separately. Based on the result analysis, our proposed model produced the highest precision, Accuracy using RFBM and relief selection methods (99.05 percent).

Detection of features in the series of electric field strength measurements using data transformation and machine learning

Data transformation is a common information processing procedure that can detect features hidden in the data that are not visible in their original form. The article shows some examples of such processing to detect features in the data of long-term observations of the intensity of the surface atmospheric electric field. This uses the data of the EFM 550 field sensor installed in the city of Nalchik. The results are visualized using the Matplotlib Python library to visualize the results of such data processing. The selected data transformation methods are also well suited for machine learning. To demonstrate the possibilities of finding features, several methods of teaching without a teacher are used. It is noted that most of the identified features in the series of measurements of the surface atmospheric electric field are associated with local meteorological phenomena.

Sleep pattern inference using IoT sonar monitoring and machine learning with Kennard-stone balance algorithm

A new paradigm of IoT monitoring using sonar sensors and microphones is studied as a contactless alternative to the traditional devices for sleep disorders in this paper. A pair of sonar sensors are used to measure the distance from the bottom and side of the body to the bedside respectively, and the basic posture of the human body during sleep can be inferred by using machine learning. When a person sleeps still, two streams of sonar signals from two adjacent sides remain unchanged. Any movement would disrupt the stationary sonar streams. Hence it could be detected as change of posture from lying still. Different sonar patterns could be recognized as specific sleeping postures by some non-linear machine learning model. This simple and novel solution could potentially be used as an alternative or supplementary to video analytics which can feedback to the user about their sleeping pattern. A new data transformation method namely Kennard-stone Balance (KSB) algorithm is also proposed for simplifying the data streams and enhancing the accuracy of the machine learning model. Simulation results show the feasibility of this sonar method, and KSB is able to improve the pattern recognition performance.

Advanced Filter Based Machine Learning Models on Clinical Databases for Outlier Detection

Feature selection approaches are used to improve the efficiency of the clinical databases in the machine learning classification. Since, most of the conventional feature selection and classification approaches are difficult to handle high dimensionality for pattern evaluation. Also these models are difficult to filter noise on different heterogeneous features. In this work, a hybrid data transformation and outlier detection methods are developed on the clinical databases to improve the classification accuracy. Experimental results show that the present model has better accuracy in evaluating the accuracy than the conventional models on clinical databases.

Single-Cell Transcriptome Analysis of Embryonic Erythro-Myeloid Progenitor Cells Reveals Lineage Heterogeneity

The paradigm that multipotential hematopoietic stem cells (HSC) differentiate through a hierarchical series of increasingly specified progenitors has recently been challenged by single cell transcriptome analyses, which suggest greater heterogeneity in progenitor populations and hierarchy than previously thought. These studies are complicated by the wide variety of hematopoietic progenitors at different stages of maturation concurrently present in the adult bone marrow. In the early mammalian embryo, unique hematopoietic progenitors arise prior to, and independently of, HSC. These include erythro-myeloid progenitors (EMP) that arise from hemogenic endothelium in the yolk sac and migrate to the fetal liver to jump-start definitive hematopoiesis before HSC emerge. In vitro, EMP can generate definitive erythroid, megakaryocyte and myeloid cells, although not B-cells, and include at least some single cells with both erythroid and myeloid potential (McGrath et al. 2015 Cell Rep. 11:1892). We have previously shown that day 9.5 (E9.5) murine EMP have a unique and uniform (kit+CD41+CD16/32+) immunophenotype in the yolk sac and circulation, which is maintained as they colonize the E10.5 liver. We performed single-cell transcriptome analysis of this apparently poised, embryonic-specific multi-potential hematopoietic progenitor population, asking whether it contains an underlying heterogeneity or intrinsic lineage hierarchy. High quality data, as assessed by transcriptomic complexity, read depth, and alignment distributions, were generated for 133 single-cell EMP (scEMP) derived from E9.5 mouse embryos. Use both of supervised and of semi-supervised analytical approaches uncovered transcriptome heterogeneity that allowed clustering of scEMP by expression patterns. Based on expression of known lineage-directing transcription factors and cell type enrichment analysis, clusters of cells were identified that could be annotated as multipotential hematopoietic, megakaryocyte/erythroid, myeloid, and a more specific macrophage lineage potential. A novel binary reduction data integration and transformation methods was developed and utilized to further refine and characterize single-cell classifications by facilitating cross platform comparisons with single cell RNA-seq data obtained from EMP progeny in E11.5 liver and with published gene expression datasets derived from adult hematopoietic progenitors and fetal macrophage populations. Predictions of lineage potential based on our analytical approaches were tested by sorting cells based on immunophenotypic markers differentially expressed in specific clusters and analyzing their potential using single cell colony-forming assays. These functional studies support the concept that immunophenotypically homogeneous EMP contain subpopulations of multipotential progenitors differentiating into erythro-megakaryocyte, myeloid, and macrophage-specific lineage fates. DisclosuresNo relevant conflicts of interest to declare.

Clustering using kernel entropy principal component analysis and variable kernel estimator

Clustering as unsupervised learning method is the mission of dividing data objects into clusters with common characteristics. In the present paper, we introduce an enhanced technique of the existing EPCA data transformation method. Incorporating the kernel function into the EPCA, the input space can be mapped implicitly into a high-dimensional of feature space. Then, the Shannon’s entropy estimated via the inertia provided by the contribution of every mapped object in data is the key measure to determine the optimal extracted features space. Our proposed method performs very well the clustering algorithm of the fast search of clusters’ centers based on the local densities’ computing. Experimental results disclose that the approach is feasible and efficient on the performance query.

Measuring efficiency in a general production possibility set allowing for negative data: An extension and a focus on returns to scale

Data envelopment analysis (DEA) is a technique used to measure the relative efficiency of a set of production units that applies multiple inputs to produce multiple outputs. In its original settings, the data is required to be nonnegative. To allow for negative data, several methods have been proposed. While these methods have merits, they also have weaknesses and limitations. This paper generalizes the construction of the production possibility set from production units with nonnegative observations to those with real values. Given the signs of the aggregate target and aggregate observed outputs of the production units to be evaluated, different models are developed to calculate the efficiencies under both variable and constant returns to scale technologies, and an additive model is used to identify the signs of the aggregate target and aggregate observed outputs. Since the efficiencies are calculated from the original observations without transformation or manipulation, the proposed method does not have the drawbacks of the existing methods. A case of the Detroit National Bank shows that the results obtained from the proposed method are more representative and reliable as compared to those obtained from a data transformation method.

Prediction of Student Dropout in Malaysian’s Private Higher Education Institute using Data Mining Application

Student dropout issue is a major concern among the academics and management of the university. The higher rate of student dropout impacted the university reputation such as reducing student enrollment, affecting the revenue of the university, financial losses for the country, and increase the existence of a social problem among the students. In this study, 2 popular classifiers were utilized to predict the student dropout namely decision tree and logistic regression model respectively. Several sets of experimental setting were employed which include three set of data partitioning - along with different types of decision tree and regression model. As for the logistic regression model, different data imputation and transformation method was tested to ensure that the model built is valid. A total of 7706 student data extracted from one of the private universities in Malaysia database (between year 2018-2019) to assess the capability of the classifier. The classifier performance is evaluated using machine learning performance measure of accuracy and misclassification rate. The result indicates that, decision tree - chi-square (2 branches) achieved slightly better classification performance of 89.49% on 80/20 data partitioning. The chosen model also identified the most important variable for accurate prediction of student dropout. Application of this model has the potential to accurately predict at risk student and to reduce student dropout rates.

PENGARUH TRANSFORMASI DATA PADA METODE LEARNING VECTOR QUANTIZATION TERHADAP AKURASI KLASIFIKASI DIAGNOSIS PENYAKIT JANTUNG

Learning Vector Quantization (LVQ) is a type of Artificial Neural Network with a supervised learning process based on competitive learning. Despite the absence of assumptions in LVQ is an advantage, it can be a problem when the predictor variables have big different ranges.This problems can be overcome by equalizing the range of all variables by data transformation so that all variables have relatively same effect. Heart Disease UCI dataset which used in this study is transformed by several transformation methods, such as minmax, decimal scaling, z-score, mean-MAD, sigmoid, and softmax. The result show that the six transformed data can provide better LVQ classification accuracy than the raw data which has 75.99% for training performance accuracy. LVQ classification accuracy with data transformation of minmax, decimal scaling, z-score, mean-MAD, sigmoid, and softmax are 89.16%, 88.22%, 89.7%, 90.1%, 88.17% and 92.18%. Based on the One-way ANOVA test and DMRT post hoc test known that there are significant differences between the results of the classification with data transformations and raw data in 0,05 significant level of α. It is also known that the best data transformation methods are softmax for training and sigmoid for testing. Keywords: heart disease, neural network, learning vector quantization, classification, data transformation

Normality Assessment of Several Quantitative Data Transformation Procedures

Usually, quantitative data standardization and/or normalization procedures requested in biological and as well in biomedical data analysis with the purpose to infer about linear regression relationship between processed variables and/or conditions. Here, we embarked to understand performance of quantitative data transformation systems in terms of reducing data variability as well as assessing data distribution normality by a computational statistic approach. For this purpose, we performed several multivariate descriptive and analytical statistical tests. Even if results shown drastic reduction of data variability by applying presently data transformation procedures, it is noteworthy to underline the relative opposite attitude of Exponential (Expo) data standardization system in that sense. In addition although, results revealed variance homogeneity for data processed by both Maximum and Logarithm data transformation methods, it is noteworthy to underline a relative variance homogeneity with regard data submitted to Box-Cox, Z-score, Minimum-Maximum and Square Root data transformation methods. Further, findings exhibited high aptitude of Square Root, Box-Cox and Logarithm quantitative data standardization methods, in stabilizing processed data variability. Interestingly, results shown high performances of Logarithm and Box-Cox data standardization systems in term of adjusting data normal distribution. In addition, multiple comparison of mean by Turkey contrast test suggested the high performance in term of data normality with regard Box-Cox standardization method. In conclusion, even if our results revealed heterogenic performances of presently processed quantitative data transformation methods, it is noteworthy to underline the high performances of both Box-Cox and Logarithm methods

A needle in a seedstack: an improved method for detection of rare alleles in bulk seed testing through KASP.

Amaranthus palmeri is an aggressive and prolific weed species with major impact on agricultural yield and is a prohibited noxious weed across the Midwest. Morphological identification of A. palmeri from other Amaranthus species is extremely difficult in seeds, which has led to genetic testing for seed identification in commercial seed lots. We created an inexpensive and reliable genetic test based on novel, species-specific, single nucleotide polymorphisms (SNPs) from GBS (Genotyping by Sequencing) data. We report three SNP-based genetic tests for identifying A. palmeri alone or in a mixed pool of Amaranthus spp. Sensitivity ranged from 99.8 to 100%, specificity from 99.59 to 100%. Accuracy for all three tests is > 99.7%. All three are capable of reliably detecting one A. palmeri seed in a pool of 200 Amaranthus spp. seeds. The test was validated across 20 populations of A. palmeri, along with eight other Amaranthus species, the largest and most genetically diverse panel of Amaranthus samples to date. Our work represents a marked improvement over existing commercial assays resulting in an identification assay that is (i) accurate, (ii) robust, (iii) easy to interpret and (iv) applicable to both leaf tissue and pools of up to 200 seeds. Included is a data transformation method for calling of closely grouped competitive fluorescence assays. We also present a comprehensive GBS dataset from the largest geographic panel of Amaranthus populations sequenced. Our approach serves as a model for developing markers for other difficult to identify species. © 2021 Society of Chemical Industry.

A novel decline curve regression procedure for analyzing shale gas production

Accurate prediction of gas production is critical for the shale gas development. Compared with the complicated physics-based simulation, the decline curve analysis (DCA) models based on parameter linearization or direct curve regression are much more straightforward and efficient. However, the linearization method might be time consuming in obtaining the parameters, and the direct regression might lead to large errors in production forecast as it assigns equal weight to the historical data. In this paper, we develop a new DCA procedure combining a data transformation method which convert the production data into logarithmic space, and a nonlinear regression algorithm. The new procedure can effectively capture the trend of late-time production history. The efficiency of the new procedure is verified with the production data of 550 gas wells in the Barnett and Marcellus shales. Based on these field data, we also analyze the performance of seven popular DCA models, i.e. Arps, power law exponential (PLE) decline, stretched exponential production decline (SEPD), Duong, Wang , variable decline modified Arps (VDMA) and logistic growth models. The performances of these models are compared and the tuned parameters for each model are provided. It is shown that the accuracy of the production forecast by the above models can be significantly improved by our new regression method, which is beneficial to the prediction and optimization of shale gas production.

Review on Integrating Geospatial Big Datasets and Open Research Issues

Big data and geographic information systems (GIS) are two technologies that have increasingly influenced many areas in the last 10 years and will continue to improve and help solve serious global problems, such as consequences of climate change or global pandemics. A wide spectrum of GIS applications interacts with the continuous growth of geospatial big data sources to drive precise and informed decisions. Geospatial big data integration is designed to accomplish the compatibility of distinct geospatial datasets regardless of their spatial coverage. The large number of geospatial big data sources demand effective data integration for storing and handling such datasets, which will be used for geospatial data analysis and visualization. For instance, risk management datasets related to healthcare and the environment are heterogeneous and disparate. Obtaining a unified view of such geospatial big datasets is complicated and challenging, especially if we consider problems related to healthcare pandemics and environmental disasters. Hence, before we can attempt to predict and mitigate processes occurring in these domains, we must realize that geospatial big data integration is crucial in consolidating datasets. We explore and discuss issues involved in integrating geospatial big datasets in this study. We then classify big data integration processes into three categories, namely, data warehousing, data transformation and integration methods. Furthermore, several research challenges focused on geospatial big data, big earth data, data warehousing, data transformation and linked data are presented. Lastly, open research issues and emerging trends that require in-depth investigations in the near future are highlighted in this study.

BSSA: Binary Salp Swarm Algorithm With Hybrid Data Transformation for Feature Selection

Feature selection is a technique commonly used in Data Mining and Machine Learning. Traditional feature selection methods, when applied to large datasets, generate a large number of feature subsets. Selecting optimal features within this high dimensional data space is time-consuming and negatively affects the system's performance. This paper proposes a new binary Salp Swarm Algorithm (bSSA) for selecting the best feature set from transformed datasets. The proposed feature selection method first transforms the original data-set using Principal Component Analysis (PCA) and fast Independent Component Analysis (fastICA) based hybrid data transformation methods; next, a binary Salp Swarm optimizer is used for finding the best features. The proposed feature selection approach improves accuracy and eliminates the selection of irrelevant features. We validate our technique on fifteen different benchmark data sets. We conduct an extensive study to measure the performance and feature selection accuracy of the proposed technique. The proposed bSSA is compared to Binary Genetic Algorithm (bGA), Binary Binomial Cuckoo Search (bBCS), Binary Grey Wolf Optimizer (bGWO), Binary Competitive Swarm Optimizer (bCSO), and Binary Crow Search Algorithm (bCSA). The proposed method attains a mean accuracy of 95.26% with 7.78% features on PCA-fastICA transformed datasets. The results show that bSSA outperforms the existing methods for the majority of the performance measures.