Search results for: Lasso

Authors: Georgiana Onicescu, Yuqian Shen

Abstract:

Due to the complex nature of geo-referenced data, multicollinearity of the risk factors in public health spatial studies is a commonly encountered issue, which leads to low parameter estimation accuracy because it inflates the variance in the regression analysis. To address this issue, we proposed a two-stage variable selection method by extending the least absolute shrinkage and selection operator (Lasso) to the Bayesian spatial setting, investigating the impact of risk factors to health outcomes. Specifically, in stage I, we performed the variable selection using Bayesian Lasso and several other variable selection approaches. Then, in stage II, we performed the model selection with only the selected variables from stage I and compared again the methods. To evaluate the performance of the two-stage variable selection methods, we conducted a simulation study with different distributions for the risk factors, using geo-referenced count data as the outcome and Michigan as the research region. We considered the cases when all candidate risk factors are independently normally distributed, or follow a multivariate normal distribution with different correlation levels. Two other Bayesian variable selection methods, Binary indicator, and the combination of Binary indicator and Lasso were considered and compared as alternative methods. The simulation results indicated that the proposed two-stage Bayesian Lasso variable selection method has the best performance for both independent and dependent cases considered. When compared with the one-stage approach, and the other two alternative methods, the two-stage Bayesian Lasso approach provides the highest estimation accuracy in all scenarios considered.

Keywords: Lasso, Bayesian analysis, spatial analysis, variable selection

Procedia PDF Downloads 104

Authors: Mikiyas Adugna

Abstract:

Effort estimation is among the activities carried out in software development processes. An accurate model of estimation leads to project success. The method of agile effort estimation is a complex task because of the dynamic nature of software development. Researchers are still conducting studies on agile effort estimation to enhance prediction accuracy. Due to these reasons, we investigated and proposed a model on LASSO and Elastic Net regression to enhance estimation accuracy. The proposed model has major components: preprocessing, train-test split, training with default parameters, and cross-validation. During the preprocessing phase, the entire dataset is normalized. After normalization, a train-test split is performed on the dataset, setting training at 80% and testing set to 20%. We chose two different phases for training the two algorithms (Elastic Net and LASSO) regression following the train-test-split. In the first phase, the two algorithms are trained using their default parameters and evaluated on the testing data. In the second phase, the grid search technique (the grid is used to search for tuning and select optimum parameters) and 5-fold cross-validation to get the final trained model. Finally, the final trained model is evaluated using the testing set. The experimental work is applied to the agile story point dataset of 21 software projects collected from six firms. The results show that both Elastic Net and LASSO regression outperformed the compared ones. Compared to the proposed algorithms, LASSO regression achieved better predictive performance and has acquired PRED (8%) and PRED (25%) results of 100.0, MMRE of 0.0491, MMER of 0.0551, MdMRE of 0.0593, MdMER of 0.063, and MSE of 0.0007. The result implies LASSO regression algorithm trained model is the most acceptable, and higher estimation performance exists in the literature.

Keywords: agile software development, effort estimation, elastic net regression, LASSO

Procedia PDF Downloads 17

Authors: Songul Cinaroglu

Abstract:

Machine learning regression methods are recommended as an alternative to classical regression methods in the existence of variables which are difficult to model. Data for health expenditure is typically non-normal and have a heavily skewed distribution. This study aims to compare machine learning regression methods by hyperparameter tuning to predict health expenditure per capita. A multiple regression model was conducted and performance results of Lasso Regression, Random Forest Regression and Support Vector Machine Regression recorded when different hyperparameters are assigned. Lambda (λ) value for Lasso Regression, number of trees for Random Forest Regression, epsilon (ε) value for Support Vector Regression was determined as hyperparameters. Study results performed by using 'k' fold cross validation changed from 5 to 50, indicate the difference between machine learning regression results in terms of R², RMSE and MAE values that are statistically significant (p < 0.001). Study results reveal that Random Forest Regression (R² ˃ 0.7500, RMSE ≤ 0.6000 ve MAE ≤ 0.4000) outperforms other machine learning regression methods. It is highly advisable to use machine learning regression methods for modelling health expenditures.

Keywords: machine learning, lasso regression, random forest regression, support vector regression, hyperparameter tuning, health expenditure

Procedia PDF Downloads 187

Authors: Yildiz Stella Dak, Jale Tezcan

Abstract:

Ground motion models that relate a strong motion parameter of interest to a set of predictive seismological variables describing the earthquake source, the propagation path of the seismic wave, and the local site conditions constitute a critical component of seismic hazard analyses. When a sufficient number of strong motion records are available, ground motion relations are developed using statistical analysis of the recorded ground motion data. In regions lacking a sufficient number of recordings, a synthetic database is developed using stochastic, theoretical or hybrid approaches. Regardless of the manner the database was developed, ground motion relations are developed using regression analysis. Development of a ground motion relation is a challenging process which inevitably requires the modeler to make subjective decisions regarding the inclusion criteria of the recordings, the functional form of the model and the set of seismological variables to be included in the model. Because these decisions are critically important to the validity and the applicability of the model, there is a continuous interest on procedures that will facilitate the development of ground motion models. This paper proposes the use of the Least Absolute Shrinkage and Selection Operator (LASSO) in selecting the set predictive seismological variables to be used in developing a ground motion relation. The LASSO can be described as a penalized regression technique with a built-in capability of variable selection. Similar to the ridge regression, the LASSO is based on the idea of shrinking the regression coefficients to reduce the variance of the model. Unlike ridge regression, where the coefficients are shrunk but never set equal to zero, the LASSO sets some of the coefficients exactly to zero, effectively performing variable selection. Given a set of candidate input variables and the output variable of interest, LASSO allows ranking the input variables in terms of their relative importance, thereby facilitating the selection of the set of variables to be included in the model. Because the risk of overfitting increases as the ratio of the number of predictors to the number of recordings increases, selection of a compact set of variables is important in cases where a small number of recordings are available. In addition, identification of a small set of variables can improve the interpretability of the resulting model, especially when there is a large number of candidate predictors. A practical application of the proposed approach is presented, using more than 600 recordings from the National Geospatial-Intelligence Agency (NGA) database, where the effect of a set of seismological predictors on the 5% damped maximum direction spectral acceleration is investigated. The set of candidate predictors considered are Magnitude, Rrup, Vs30. Using LASSO, the relative importance of the candidate predictors has been ranked. Regression models with increasing levels of complexity were constructed using one, two, three, and four best predictors, and the models’ ability to explain the observed variance in the target variable have been compared. The bias-variance trade-off in the context of model selection is discussed.

Keywords: ground motion modeling, least absolute shrinkage and selection operator, penalized regression, variable selection

Procedia PDF Downloads 299

Authors: Fatemeh Valizadeh Gamchi, Edward L. Boone

Abstract:

This research focuses on Lyme disease, a widespread infectious condition in the United States caused by the bacterium Borrelia burgdorferi sensu stricto. It is critical to identify environmental and economic elements that are contributing to the spread of the disease. This study examined data from Virginia to identify a subset of explanatory variables significant for Lyme disease case numbers. To identify relevant variables and avoid overfitting, linear poisson, and regularization regression methods such as a ridge, lasso, and elastic net penalty were employed. Cross-validation was performed to acquire tuning parameters. The methods proposed can automatically identify relevant disease count covariates. The efficacy of the techniques was assessed using four criteria on three simulated datasets. Finally, using the Virginia Department of Health’s Lyme disease data set, the study successfully identified key factors, and the results were consistent with previous studies.

Keywords: lyme disease, Poisson generalized linear model, ridge regression, lasso regression, elastic net regression

Procedia PDF Downloads 87

Authors: Kang-Mo Jung

Abstract:

The use of regularization for statistical methods has become popular. The least absolute shrinkage and selection operator (LASSO) framework has become the standard tool for sparse regression. However, it is well known that the LASSO is sensitive to outliers or leverage points. We consider a new robust estimation which is composed of the weighted loss function of the pairwise difference of residuals and the adaptive penalty function regulating the tuning parameter for each variable. Rank regression is resistant to regression outliers, but not to leverage points. By adopting a weighted loss function, the proposed method is robust to leverage points of the predictor variable. Furthermore, the adaptive penalty function gives us good statistical properties in variable selection such as oracle property and consistency. We develop an efficient algorithm to compute the proposed estimator using basic functions in program R. We used an optimal tuning parameter based on the Bayesian information criterion (BIC). Numerical simulation shows that the proposed estimator is effective for analyzing real data set and contaminated data.

Keywords: adaptive penalty function, robust penalized regression, variable selection, weighted rank regression

Procedia PDF Downloads 429

Authors: Jonathan Iworiso

Abstract:

Forecasting the equity premium out-of-sample is a major concern to researchers in finance and emerging markets. The quest for a superior model that can forecast the equity premium with significant economic gains has resulted in several controversies on the choice of variables and suitable techniques among scholars. This research focuses mainly on the application of Regression Training (RT) techniques to forecast monthly equity premium out-of-sample recursively with an expanding window method. A broad category of sophisticated regression models involving model complexity was employed. The RT models include Ridge, Forward-Backward (FOBA) Ridge, Least Absolute Shrinkage and Selection Operator (LASSO), Relaxed LASSO, Elastic Net, and Least Angle Regression were trained and used to forecast the equity premium out-of-sample. In this study, the empirical investigation of the RT models demonstrates significant evidence of equity premium predictability both statistically and economically relative to the benchmark historical average, delivering significant utility gains. They seek to provide meaningful economic information on mean-variance portfolio investment for investors who are timing the market to earn future gains at minimal risk. Thus, the forecasting models appeared to guarantee an investor in a market setting who optimally reallocates a monthly portfolio between equities and risk-free treasury bills using equity premium forecasts at minimal risk.

Keywords: regression training, out-of-sample forecasts, expanding window, statistical predictability, economic significance, utility gains

Procedia PDF Downloads 70

Authors: Jianan Sun, Ziwen Ye

Abstract:

Multidimensional computerized adaptive testing (MCAT) is a popular research topic in psychometrics. It is important for practitioners to clearly know the item-trait patterns of administered items when a test like MCAT is operated. Item-trait pattern recognition refers to detecting which latent traits in a psychological test are measured by each of the specified items. If the item-trait patterns of the replenished items in MCAT item pool are well detected, the interpretability of the items can be improved, which can further promote the abilities of the examinees who attending the MCAT to be accurately estimated. This research explores to solve the item-trait pattern recognition problem of the replenished items in MCAT item pool from the perspective of statistical variable selection. The popular multidimensional item response theory model, multidimensional two-parameter logistic model, is assumed to fit the response data of MCAT. The proposed method uses the least absolute shrinkage and selection operator (LASSO) to detect item-trait patterns of replenished items based on the essential information of item responses and ability estimates of examinees collected from a designed MCAT procedure. Several advantages of the proposed method are outlined. First, the proposed method does not strictly depend on the relative order between the replenished items and the selected operational items, so it allows the replenished items to be mixed into the operational items in reasonable order such as considering content constraints or other test requirements. Second, the LASSO used in this research improves the interpretability of the multidimensional replenished items in MCAT. Third, the proposed method can exert the advantage of shrinkage method idea for variable selection, so it can help to check item quality and key dimension features of replenished items and saves more costs of time and labors in response data collection than traditional factor analysis method. Moreover, the proposed method makes sure the dimensions of replenished items are recognized to be consistent with the dimensions of operational items in MCAT item pool. Simulation studies are conducted to investigate the performance of the proposed method under different conditions for varying dimensionality of item pool, latent trait correlation, item discrimination, test lengths and item selection criteria in MCAT. Results show that the proposed method can accurately detect the item-trait patterns of the replenished items in the two-dimensional and the three-dimensional item pool. Selecting enough operational items from the item pool consisting of high discriminating items by Bayesian A-optimality in MCAT can improve the recognition accuracy of item-trait patterns of replenished items for the proposed method. The pattern recognition accuracy for the conditions with correlated traits is better than those with independent traits especially for the item pool consisting of comparatively low discriminating items. To sum up, the proposed data-driven method based on the LASSO can accurately and efficiently detect the item-trait patterns of replenished items in MCAT.

Keywords: item-trait pattern recognition, least absolute shrinkage and selection operator, multidimensional computerized adaptive testing, variable selection

Procedia PDF Downloads 92

Authors: ZhenlinJu, Christopher P. Vellano, RehanAkbani, Yiling Lu, Gordon B. Mills

Abstract:

The epithelial–mesenchymal transition (EMT) is a process by which epithelial cells acquire mesenchymal characteristics, such as profound disruption of cell-cell junctions, loss of apical-basolateral polarity, and extensive reorganization of the actin cytoskeleton to induce cell motility and invasion. A hallmark of EMT is its capacity to promote metastasis, which is due in part to activation of several transcription factors and subsequent downregulation of E-cadherin. Unfortunately, current approaches have yet to uncover robust protein marker sets that can classify tumors as possessing strong EMT signatures. In this study, we utilize reverse phase protein array (RPPA) data and consensus clustering methods to successfully classify a subset of cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC) tumors into an EMT protein signaling group (EMT group). The overall survival (OS) of patients in the EMT group is significantly worse than those in the other Hormone and PI3K/AKT signaling groups. In addition to a shrinkage and selection method for linear regression (LASSO), we applied training/test set and Monte Carlo resampling approaches to identify a set of protein markers that predicts the EMT status of CESC tumors. We fit a logistic model to these protein markers and developed a classifier, which was fixed in the training set and validated in the testing set. The classifier robustly predicted the EMT status of the testing set with an area under the curve (AUC) of 0.975 by Receiver Operating Characteristic (ROC) analysis. This method not only identifies a core set of proteins underlying an EMT signature in cervical cancer patients, but also provides a tool to examine protein predictors that drive molecular subtypes in other diseases.

Keywords: consensus clustering, TCGA CESC, Silhouette, Monte Carlo LASSO

Procedia PDF Downloads 434

Authors: Oleg Vlasovets, Fabian Schaipp, Christian L. Mueller

Abstract:

we conduct a network analysis and predict the sex of 1000 participants based on ”connectome” - pairwise Pearson’s correlation across 436 brain parcels. We solve the non-smooth convex optimization problem, known under the name of Graphical Lasso, where the solution includes a low-rank component. With this solution and machine learning model for a sex prediction, we explain the brain parcels-sex connectivity patterns.

Keywords: network analysis, neuroscience, machine learning, optimization

Procedia PDF Downloads 109

Authors: Rauan Sakenov, Peter Bukovics, Peter Gaszler, Veronika Tokacs-Kollar, Beata Bugyi

Abstract:

FH2 (formin homology-2) domains of several proteins, collectively known as formins, including DAAM, DAAM1 and mDia1, promote G-actin nucleation and elongation. FH2 domains of these formins exist as oligomers. Chain dimerization by ring structure formation serves as a structural basis for actin polymerization function of FH2 domain. Proper single chain configuration and specific interactions between its various regions are necessary for individual chains to form a dimer functional in G-actin nucleation and elongation. FH1 and WH2 domain-containing formins were shown to behave as intrinsically disordered proteins. Thus, the aim of this research was to study structural dynamics of FH2 domain of DAAM. To investigate structural features of FH2 domain of DAAM, molecular dynamics simulation of chain A of FH2 domain of DAAM solvated in water box in 50 mM NaCl was conducted at temperatures from 293.15 to 353.15K, with VMD 1.9.2, NAMD 2.14 and Amber Tools 21 using 2z6e and 1v9d PDB structures of DAAM was obtained on I-TASSER webserver. Calcium and ATP bound G-actin 3hbt PDB structure was used as a reference protein with well-described structural dynamics of denaturation. Topology and parameter information of CHARMM 2012 additive all-atom force fields for proteins, carbohydrate derivatives, water and ions were used in NAMD 2.14 and ff19SB force field for proteins in Amber Tools 21. The systems were energy minimized for the first 1000 steps, equilibrated and produced in NPT ensemble for 1ns using stochastic Langevin dynamics and the particle mesh Ewald method. Our root-mean square deviation (RMSD) analysis of molecular dynamics of chain A of FH2 domains of DAAM revealed similar insignificant changes of total molecular average RMSD values of FH2 domain of these formins at temperatures from 293.15 to 353.15K. In contrast, total molecular average RMSD values of G-actin showed considerable increase at 328K, which corresponds to the denaturation of G-actin molecule at this temperature and its transition from native, ordered, to denatured, disordered, state which is well-described in the literature. RMSD values of lasso and tail regions of chain A of FH2 domain of DAAM exhibited higher than total molecular average RMSD at temperatures from 293.15 to 353.15K. These regions are functional in intra- and interchain interactions and contain highly conserved tryptophan residues of lasso region, highly conserved GNYMN sequence of post region and amino acids of the shell of hydrophobic pocket of the salt bridge between Arg171 and Asp321, which are important for structural stability and ordered state of FH2 domain of DAAM and its functions in FH2 domain dimerization. In conclusion, higher than total molecular average RMSD values of lasso and post regions of chain A of FH2 domain of DAAM may explain disordered state of FH2 domain of DAAM at temperatures from 293.15 to 353.15K. Finally, absence of marked transition, in terms of significant changes in average molecular RMSD values between native and denatured states of FH2 domain of DAAM at temperatures from 293.15 to 353.15K, can make it possible to attribute these formins to the group of intrinsically disordered proteins rather than to the group of intrinsically ordered proteins such as G-actin.

Keywords: FH2 domain, DAAM, formins, molecular modelling, computational biophysics

Procedia PDF Downloads 102

Authors: Kourosh Modarresi

Abstract:

The Netflix problem has brought the topic of “Recommendation Systems” into the mainstream of computer science, mathematics, and statistics. Though much progress has been made, the available algorithms do not obtain satisfactory results. The success of these algorithms is rarely above 5%. This work is based on the belief that the main challenge is to come up with “scalable personalization” models. This paper uses an adaptive regularization of inverse singular value decomposition (SVD) that applies adaptive penalization on the singular vectors. The results show far better matching for recommender systems when compared to the ones from the state of the art models in the industry.

Keywords: convex optimization, LASSO, regression, recommender systems, singular value decomposition, low rank approximation

Procedia PDF Downloads 416

Authors: Stylianos Kampakis

Abstract:

This paper presents a new algorithm for neural networks called “Input Dropout with Aggressive Re-weighting” (IDAR) aimed specifically at datasets with many useless features. IDAR combines two techniques (dropout of input neurons and aggressive re weighting) in order to eliminate the influence of noisy features. The technique can be seen as a generalization of dropout. The algorithm is tested on two different benchmark data sets: a noisy version of the iris dataset and the MADELON data set. Its performance is compared against three other popular techniques for dealing with useless features: L2 regularization, LASSO and random forests. The results demonstrate that IDAR can be an effective technique for handling data sets with many useless features.

Keywords: neural networks, feature selection, regularization, aggressive reweighting

Procedia PDF Downloads 423

Authors: Dylan Santos De Pinho, Nabil Ouerhani

Abstract:

Energy consumption of machine-tools is becoming critical for machine-tool builders and end-users because of economic, ecological and legislation-related reasons. Many machine-tool builders are seeking for solutions that allow the reduction of energy consumption of machine-tools while preserving the same productivity rate and the same quality of machined parts. In this paper, we present the first results of a project conducted jointly by academic and industrial partners to reduce the energy consumption of a Swiss-Type lathe. We employ genetic algorithms to find optimal machining parameters – the set of parameters that lead to the best trade-off between energy consumption, part quality and tool lifetime. Three main machining process parameters are considered in our optimization technique, namely depth of cut, spindle rotation speed and material feed rate. These machining process parameters have been identified as the most influential ones in the configuration of the Swiss-type machining process. A state-of-the-art multi-objective genetic algorithm has been used. The algorithm combines three fitness functions, which are objective functions that permit to evaluate a set of parameters against the three objectives: energy consumption, quality of the machined parts, and tool lifetime. In this paper, we focus on the investigation of the fitness function related to energy consumption. Four different energy consumption related fitness functions have been investigated and compared. The first fitness function refers to the Kienzle cutting force model. The second fitness function uses the Material Removal Rate (RMM) as an indicator of energy consumption. The two other fitness functions are non-deterministic, learning-based functions. One fitness function uses a simple Neural Network to learn the relation between the process parameters and the energy consumption from experimental data. Another fitness function uses Lasso regression to determine the same relation. The goal is, then, to find out which fitness functions predict best the energy consumption of a Swiss-Type machining process for the given set of machining process parameters. Once determined, these functions may be used for optimization purposes – determine the optimal machining process parameters leading to minimum energy consumption. The performance of the four fitness functions has been evaluated. The Tornos DT13 Swiss-Type Lathe has been used to carry out the experiments. A mechanical part including various Swiss-Type machining operations has been selected for the experiments. The evaluation process starts with generating a set of CNC (Computer Numerical Control) programs for machining the part at hand. Each CNC program considers a different set of machining process parameters. During the machining process, the power consumption of the spindle is measured. All collected data are assigned to the appropriate CNC program and thus to the set of machining process parameters. The evaluation approach consists in calculating the correlation between the normalized measured power consumption and the normalized power consumption prediction for each of the four fitness functions. The evaluation shows that the Lasso and Neural Network fitness functions have the highest correlation coefficient with 97%. The fitness function “Material Removal Rate” (MRR) has a correlation coefficient of 90%, whereas the Kienzle-based fitness function has a correlation coefficient of 80%.

Keywords: adaptive machining, genetic algorithms, smart manufacturing, parameters optimization

Procedia PDF Downloads 117

Authors: A. Bere, H. G. Sithuba, K. Kyei, C. Sigauke

Abstract:

We employ a discrete logit survival model to investigate the risk factors for early alcohol intake among students at two tertiary institutions in Thohoyandou, South Africa. Data were collected from a sample of 744 students using a self-administered questionnaire. Significant covariates were arrived at through a regularization algorithm implemented using the glmmLasso package. The tuning parameter was determined using a five-fold cross-validation algorithm. The baseline hazard was modelled as a smooth function of time through the use of spline functions. The results show that the hazard of initial alcohol intake peaks at the age of about 16 years and that at any given time, being of a male gender, prior use of other drugs, having drinking peers, having experienced negative life events and physical abuse are associated with a higher risk of alcohol intake debut.

Keywords: cross-validation, discrete hazard model, LASSO, smooth baseline hazard

Procedia PDF Downloads 156

Authors: Hexiong Liu

Abstract:

Powder metallurgy is the optimal method for the consolidation and preparation of W(Mo) alloys, which exhibit excellent application prospects at high temperatures. The properties of W(Mo) alloys are closely related to the sintered density. However, controlling the sintered density and porosity of these alloys is still challenging. In the past, the regulation methods mainly focused on time-consuming and costly trial-and-error experiments. In this study, the sintering data for more than a dozen W(Mo) alloys constituted a small-scale dataset, including both solid and liquid phases of sintering. Furthermore, simple descriptors were used to predict the sintered density of W(Mo) alloys based on the descriptor selection strategy and machine learning method (ML), where the ML algorithm included the least absolute shrinkage and selection operator (Lasso) regression, k-nearest neighbor (k-NN), random forest (RF), and multi-layer perceptron (MLP). The results showed that the interpretable descriptors extracted by our proposed selection strategy and the MLP neural network achieved a high prediction accuracy (R>0.950). By further predicting the sintered density of W(Mo) alloys using different sintering processes, the error between the predicted and experimental values was less than 0.063, confirming the application potential of the model.

Keywords: sintered density, machine learning, interpretable descriptors, W(Mo) alloy

Procedia PDF Downloads 46

Authors: Libo Jiang, Huan Li, Rongling Wu

Abstract:

Ordinary differentiation equations (ODE) have proven to be powerful for reconstructing precise and informative gene regulatory networks (GRNs) from dynamic gene expression data. However, joint modeling and analysis of all genes, essential for the systematical characterization of genetic interactions, are challenging due to high dimensionality and a complex pattern of genetic regulation including activation, repression, and antitermination. Here, we address these challenges by unifying variable selection and game theory through ODE. Each gene within a GRN is co-expressed with its partner genes in a way like a game of multiple players, each of which tends to choose an optimal strategy to maximize its “fitness” across the whole network. Based on this unifying theory, we designed and conducted a real experiment to infer salt tolerance-related GRNs for Euphrates poplar, a hero tree that can grow in the saline desert. The pattern and magnitude of interactions between several hub genes within these GRNs were found to determine the capacity of Euphrates poplar to resist to saline stress.

Keywords: gene regulatory network, ordinary differential equation, game theory, LASSO, saline resistance

Procedia PDF Downloads 609

Authors: Sarah Beaver, Renee Bryce

Abstract:

Audiograms detect hearing impairment, but missing values pose problems. This work explores imputations in an attempt to improve accuracy. This work implements Linear Regression, Lasso, Linear Support Vector Regression, Bayesian Ridge, K Nearest Neighbors (KNN), and Random Forest machine learning techniques to impute audiogram frequencies ranging from 125Hz to 8000Hz. The data contains patients who had or were candidates for cochlear implants. Accuracy is compared across two different Nested Cross-Validation k values. Over 4000 audiograms were used from 800 unique patients. Additionally, training on data combines and compares left and right ear audiograms versus single ear side audiograms. The accuracy achieved using Root Mean Square Error (RMSE) values for the best models for Random Forest ranges from 4.74 to 6.37. The R\textsuperscript{2} values for the best models for Random Forest ranges from .91 to .96. The accuracy achieved using RMSE values for the best models for KNN ranges from 5.00 to 7.72. The R\textsuperscript{2} values for the best models for KNN ranges from .89 to .95. The best imputation models received R\textsuperscript{2} between .89 to .96 and RMSE values less than 8dB. We also show that the accuracy of classification predictive models performed better with our best imputation models versus constant imputations by a two percent increase.

Keywords: machine learning, audiograms, data imputations, single imputations

Procedia PDF Downloads 48

Authors: Khaled M. Alqahtani

Abstract:

Copy number alterations (CNA) are structural variations in the genome in which some regions exhibit more or less than the normal two chromosomal copies. This genomic CNA profile provides critical information in tumor progression and is, therefore, informative for patients' survival. It is currently a statistical challenge to model patients' survival using their genomic CNA profiles while at the same time identifying regions in the genome that are associated with patients' survival. Some methods have been proposed, including the Cox proportional hazard (PH) model with ridge, lasso, or elastic net penalties. However, these methods do not take the general dependencies between genomic regions into account and produce results that are difficult to interpret. In this paper, we extend the elastic net penalty by introducing an additional penalty that takes into account general dependencies between genomic regions. This model produces smooth parameter estimates while simultaneously performing variable selection via sparse solution. The results indicate that the proposed method shows a better prediction performance than other models in our simulation study while enabling us to investigate regions in the genome that are associated with the patient's survival with sensible interpretation. We illustrate the method using a real dataset from a lung cancer cohort and simulated data.

Keywords: copy number alterations, cox proportional hazard, lung cancer, regression, sparse solution

Procedia PDF Downloads 8

Authors: Hongmeng Su, Luyu Zhao, Yanyan Qian, Hong Fan

Abstract:

Aim: Hepatocellular carcinoma (HCC) is one of the most common malignant tumors that endanger human health seriously. RNA methylation, especially N6-methyladenosine (m⁶A) and 5-methylcytosine (m⁵C), a crucial epigenetic transcriptional regulatory mechanism, plays an important role in tumorigenesis, progression and prognosis. This research aims to systematically evaluate the prognostic value of m⁶A and m⁵C modulators in HCC patients. Methods: Twenty-four modulators of m⁶A and m⁵C were candidates to analyze their expression level and their contribution to predict the prognosis of HCC. Consensus clustering analysis was applied to classify HCC patients. Cox and LASSO regression were used to construct the risk model. According to the risk score, HCC patients were divided into high-risk and low/medium-risk groups. The clinical pathology factors of HCC patients were analyzed by univariate and multivariate Cox regression analysis. Results: The HCC patients were classified into 2 clusters with significant differences in overall survival and clinical characteristics. Nine-gene risk model was constructed including METTL3, VIRMA, YTHDF1, YTHDF2, NOP2, NSUN4, NSUN5, DNMT3A and ALYREF. It was indicated that the risk score could serve as an independent prognostic factor for patients with HCC. Conclusion: This study constructed a Nine-gene risk model by modulators of m⁶A and m⁵C and investigated its effect on the clinical prognosis of HCC. This model may provide important consideration for the therapeutic strategy and prognosis evaluation analysis of patients with HCC.

Keywords: hepatocellular carcinoma, m⁶A, m⁵C, prognosis, RNA methylation

Procedia PDF Downloads 32

Authors: Mohammed Abdullah Alshahrani

Abstract:

This research focuses on the investigation of partial least squares (PLS) methodology to deal with high-dimensional correlated data. Current developments in technology have enabled experiments to produce data that are characterized by, first, the number of variables that far exceeds the number of observations and, second, variables that are substantially correlated between them. These types of data are commonly found in, first, chemometrics, where absorbance levels of chemical samples are recorded across hundreds of wavelengths in a calibration of a near-infrared (NIR) spectrometer. Second, they are also common to be found in genomics where copy number alterations (CNA) are recorded across thousands of genomic regions from cancer patients. In our study, we investigated key areas to address these challenges. Firstly, we tackled the issue of three main PLS algorithms having potentially different interpretations of relevant quantities. We unified these interpretations by identifying scenarios where all three algorithms yield the same estimates. Secondly, we explored the phenomenon of unusual negative shrinkage factors encountered during PLS model fitting. Unlike ridge regression or principal component regression, where shrinkage factors range between zero and one, PLS can exhibit factors greater than one or even negative, hence more aptly termed ‘filter factors’ rather than ‘shrinkage factors’. This characteristic allows PLS to effectively handle high-dimensional data by applying shrinkage to estimates. To our knowledge, there has been no previous meaningful investigation on the negative filter factors (NFF) in PLS. In this research we present a novel result whereby we identify the condition for NFF to happen and investigate characteristics of the data that are associated with NFF to get an insight. Lastly, the main challenge of the application of PLS is in the interpretation of weights associated with the predictors. With hundreds and thousands of predictors, each and every predictor variable has non-zero weight. However, we expect that only some predictor variables are contributing to the association with the outcome variable. We, therefore, resort to the sparse estimation of predictor weights where some weights are zero estimated and the other weights are non-zero. A (standard) lasso estimation has a weakness in dealing with correlated variables as it picks up one variable within a correlation block without knowing the reason. A novel approach is needed to consider the dependencies between predictor variables in estimating the weights. We propose a new method where a new penalty function is introduced in the likelihood function associated with the estimation of weights. The penalty function is a combination of a lasso penalty that imposes sparsity and a penalty based on Cauchy distribution with a smoother matrix to take into account dependencies between genomic regions. The results show that the estimates of the weights are sparse: many weights are zero estimated, and those non-zero estimates are grouped and exhibit smoothness within them. The interpretation of genomic regions becomes easy, and the identification of important regions for each component can be done simultaneously with prediction in a single modeling framework. We investigate the relation between PLS and graphical modeling using the information in the weights to construct the graph with unsuccessful results. High-dimensional data where the number of predictors (p) exceeds the number of observations (n) are widely used in many applications of regression analysis. Ordinary least squares regression (OLS), which is the most well-known method for regression problems, has less performance with high-dimensional and highly- correlated data. Previous studies have shown that there is an association between copy number alterations (CNA) in some key genes and disease phenotypes. Moreover, it is very important in high-dimensional data to classify the samples into groups, such as tumor types, of gene expression data in bioinformatics and biology. However, the standard regression of classification methods will fail in these cases because the predictors matrix is singular and so, cannot be inverted. Hence, regularised methods are needed such as shrinkage methods and dimension reduction methods. One of the most suggested methods in the literature is partial least squares regression (PLS) for linear regression and classification.

Keywords: negative filter factors, partial least square regression, high-dimensional data, biostatistics, bioinformatics

Procedia PDF Downloads 11

Authors: Noora Al Muftah, Reda Rawi, Richard Thompson, Halima Bensmail

Abstract:

Clinical responses to anticancer therapies are often restricted to a subset of patients. In some cases, mutated cancer genes are potent biomarkers of response to targeted agents. There is an urgent need to identify biomarkers that predict which patients with are most likely to respond to treatment. Systematic efforts to correlate tumor mutational data with biologic dependencies may facilitate the translation of somatic mutation catalogs into meaningful biomarkers for patient stratification. To identify genomic features associated with drug sensitivity and uncover new biomarkers of sensitivity and resistance to cancer therapeutics, we have screened and integrated a panel of several hundred cancer cell lines from different databases, mutation, DNA copy number, and gene expression data for hundreds of cell lines with their responses to targeted and cytotoxic therapies with drugs under clinical and preclinical investigation. We found mutated cancer genes were associated with cellular response to most currently available Glioma cancer drugs and some frequently mutated genes were associated with sensitivity to a broad range of therapeutic agents. By linking drug activity to the functional complexity of cancer genomes, systematic pharmacogenomic profiling in cancer cell lines provides a powerful biomarker discovery platform to guide rational cancer therapeutic strategies.

Keywords: cancer, gene network, Lasso, penalized regression, P-values, unbiased estimator

Procedia PDF Downloads 378

Authors: Chen Zhang, Nan Chen

Abstract:

High-dimensional process monitoring becomes increasingly important in many application domains, where usually the process distribution is unknown and much more complicated than the normal distribution, and the between-stream correlation can not be neglected. However, since the process dimension is generally much bigger than the reference sample size, most traditional nonparametric multivariate control charts fail in high-dimensional cases due to the curse of dimensionality. Furthermore, when the process goes out of control, the influenced variables are quite sparse compared with the whole dimension, which increases the detection difficulty. Targeting at these issues, this paper proposes a new nonparametric monitoring scheme for high-dimensional processes. This scheme first projects the high-dimensional process into several subprocesses using random projections for dimension reduction. Then, for every subprocess with the dimension much smaller than the reference sample size, a local nonparametric control chart is constructed based on the spatial rank test to detect changes in this subprocess. Finally, the results of all the local charts are fused together for decision. Furthermore, after an out-of-control (OC) alarm is triggered, a diagnostic framework is proposed. using the square-root LASSO. Numerical studies demonstrate that the chart has satisfactory detection power for sparse OC changes and robust performance for non-normally distributed data, The diagnostic framework is also effective to identify truly changed variables. Finally, a real-data example is presented to demonstrate the application of the proposed method.

Keywords: random projection, high-dimensional process control, spatial rank, sequential change detection

Procedia PDF Downloads 273

Authors: Lule Basha, Eralda Gjika

Abstract:

This study is focused on analyzing climate change in Albania and its potential effects on cereal yields. Initially, monthly temperature and rainfalls in Albania were studied for the period 1960-2021. Climacteric variables are important variables when trying to model cereal yield behavior, especially when significant changes in weather conditions are observed. For this purpose, in the second part of the study, linear and nonlinear models explaining cereal yield are constructed for the same period, 1960-2021. The multiple linear regression analysis and lasso regression method are applied to the data between cereal yield and each independent variable: average temperature, average rainfall, fertilizer consumption, arable land, land under cereal production, and nitrous oxide emissions. In our regression model, heteroscedasticity is not observed, data follow a normal distribution, and there is a low correlation between factors, so we do not have the problem of multicollinearity. Machine-learning methods, such as random forest, are used to predict cereal yield responses to climacteric and other variables. Random Forest showed high accuracy compared to the other statistical models in the prediction of cereal yield. We found that changes in average temperature negatively affect cereal yield. The coefficients of fertilizer consumption, arable land, and land under cereal production are positively affecting production. Our results show that the Random Forest method is an effective and versatile machine-learning method for cereal yield prediction compared to the other two methods.

Keywords: cereal yield, climate change, machine learning, multiple regression model, random forest

Procedia PDF Downloads 52

Authors: Nico Rosamilia

Abstract:

The study evaluates the importance of non-financial disclosure practices for regulators, investors, businesses, and markets. It aims to create a sector-specific set of indicators for environmental, social, and governance (ESG) performances alternative to the ratings of the agencies. The existing literature extensively studies the implementation of ESG rating systems. Conversely, this study has a twofold outcome. Firstly, it should generalize incentive systems and governance policies for ESG and sustainable principles. Therefore, it should contribute to the EU Sustainable Finance Disclosure Regulation. Secondly, it concerns the market and the investors by highlighting successful sustainable investing. Indeed, the study contemplates the effect of ESG adoption practices on corporate value. The research explores the asset pricing angle in order to shed light on the fragmented argument on the finance of ESG. Investors may be misguided about the positive or negative effects of ESG on performances. The paper proposes a different method to evaluate ESG performances. By comparing the results of a traditional econometric approach (Lasso) with a machine learning algorithm (Random Forest), the study establishes a set of indicators for ESG performance. Therefore, the research also empirically contributes to the theoretical strands of literature regarding model selection and variable importance in a finance framework. The algorithms will spit out sector-specific indicators. This set of indicators defines an alternative to the compounded scores of ESG rating agencies and avoids the possible offsetting effect of scores. With this approach, the paper defines a sector-specific set of indicators to standardize ESG disclosure. Additionally, it tries to shed light on the absence of a clear understanding of the direction of the ESG effect on corporate value (the problem of endogeneity).

Keywords: ESG ratings, non-financial information, value of firms, sustainable finance

Procedia PDF Downloads 50

Authors: Faiz N. K. Yusufi, Aquil Ahmed, Jamal Ahmad

Abstract:

Diabetes in India is growing at an alarming rate and the complications caused by it need to be controlled. Coronary heart disease (CHD) is one of the complications that will be discussed for prediction in this study. India has the second most number of diabetes patients in the world. To the best of our knowledge, there is no CHD risk score for Indian type 2 diabetes patients. Any form of CHD has been taken as the event of interest. A sample of 750 was determined and randomly collected from the Rajiv Gandhi Centre for Diabetes and Endocrinology, J.N.M.C., A.M.U., Aligarh, India. Collected variables include patients data such as sex, age, height, weight, body mass index (BMI), blood sugar fasting (BSF), post prandial sugar (PP), glycosylated haemoglobin (HbA1c), diastolic blood pressure (DBP), systolic blood pressure (SBP), smoking, alcohol habits, total cholesterol (TC), triglycerides (TG), high density lipoprotein (HDL), low density lipoprotein (LDL), very low density lipoprotein (VLDL), physical activity, duration of diabetes, diet control, history of antihypertensive drug treatment, family history of diabetes, waist circumference, hip circumference, medications, central obesity and history of CHD. Predictive risk scores of CHD events are designed by cox proportional hazard regression. Model calibration and discrimination is assessed from Hosmer Lemeshow and area under receiver operating characteristic (ROC) curve. Overfitting and underfitting of the model is checked by applying regularization techniques and best method is selected between ridge, lasso and elastic net regression. Youden’s index is used to choose the optimal cut off point from the scores. Five year probability of CHD is predicted by both survival function and Markov chain two state model and the better technique is concluded. The risk scores for CHD developed can be calculated by doctors and patients for self-control of diabetes. Furthermore, the five-year probabilities can be implemented as well to forecast and maintain the condition of patients.

Keywords: coronary heart disease, cox proportional hazard regression, ROC curve, type 2 diabetes Mellitus

Procedia PDF Downloads 186

Authors: Faiz N. K. Yusufi, Aquil Ahmed, Jamal Ahmad

Abstract:

Predicting the risk of diabetic retinopathy (DR) in Indian type 2 diabetes patients is immensely necessary. India, being the second largest country after China in terms of a number of diabetic patients, to the best of our knowledge not a single risk score for complications has ever been investigated. Diabetic retinopathy is a serious complication and is the topmost reason for visual impairment across countries. Any type or form of DR has been taken as the event of interest, be it mild, back, grade I, II, III, and IV DR. A sample was determined and randomly collected from the Rajiv Gandhi Centre for Diabetes and Endocrinology, J.N.M.C., A.M.U., Aligarh, India. Collected variables include patients data such as sex, age, height, weight, body mass index (BMI), blood sugar fasting (BSF), post prandial sugar (PP), glycosylated haemoglobin (HbA1c), diastolic blood pressure (DBP), systolic blood pressure (SBP), smoking, alcohol habits, total cholesterol (TC), triglycerides (TG), high density lipoprotein (HDL), low density lipoprotein (LDL), very low density lipoprotein (VLDL), physical activity, duration of diabetes, diet control, history of antihypertensive drug treatment, family history of diabetes, waist circumference, hip circumference, medications, central obesity and history of DR. Cox proportional hazard regression is used to design risk scores for the prediction of retinopathy. Model calibration and discrimination are assessed from Hosmer Lemeshow and area under receiver operating characteristic curve (ROC). Overfitting and underfitting of the model are checked by applying regularization techniques and best method is selected between ridge, lasso and elastic net regression. Optimal cut off point is chosen by Youden’s index. Five-year probability of DR is predicted by both survival function, and Markov chain two state model and the better technique is concluded. The risk scores developed can be applied by doctors and patients themselves for self evaluation. Furthermore, the five-year probabilities can be applied as well to forecast and maintain the condition of patients. This provides immense benefit in real application of DR prediction in T2DM.

Keywords: Cox proportional hazard regression, diabetic retinopathy, ROC curve, type 2 diabetes mellitus

Procedia PDF Downloads 145

Authors: Xiangtuo Chen, Paul-Henry Cournéde

Abstract:

Crop yield prediction is a paramount issue in agriculture. The main idea of this paper is to find out efficient way to predict the yield of corn based meteorological records. The prediction models used in this paper can be classified into model-driven approaches and data-driven approaches, according to the different modeling methodologies. The model-driven approaches are based on crop mechanistic modeling. They describe crop growth in interaction with their environment as dynamical systems. But the calibration process of the dynamic system comes up with much difficulty, because it turns out to be a multidimensional non-convex optimization problem. An original contribution of this paper is to propose a statistical methodology, Multi-Scenarios Parameters Estimation (MSPE), for the parametrization of potentially complex mechanistic models from a new type of datasets (climatic data, final yield in many situations). It is tested with CORNFLO, a crop model for maize growth. On the other hand, the data-driven approach for yield prediction is free of the complex biophysical process. But it has some strict requirements about the dataset. A second contribution of the paper is the comparison of these model-driven methods with classical data-driven methods. For this purpose, we consider two classes of regression methods, methods derived from linear regression (Ridge and Lasso Regression, Principal Components Regression or Partial Least Squares Regression) and machine learning methods (Random Forest, k-Nearest Neighbor, Artificial Neural Network and SVM regression). The dataset consists of 720 records of corn yield at county scale provided by the United States Department of Agriculture (USDA) and the associated climatic data. A 5-folds cross-validation process and two accuracy metrics: root mean square error of prediction(RMSEP), mean absolute error of prediction(MAEP) were used to evaluate the crop prediction capacity. The results show that among the data-driven approaches, Random Forest is the most robust and generally achieves the best prediction error (MAEP 4.27%). It also outperforms our model-driven approach (MAEP 6.11%). However, the method to calibrate the mechanistic model from dataset easy to access offers several side-perspectives. The mechanistic model can potentially help to underline the stresses suffered by the crop or to identify the biological parameters of interest for breeding purposes. For this reason, an interesting perspective is to combine these two types of approaches.

Keywords: crop yield prediction, crop model, sensitivity analysis, paramater estimation, particle swarm optimization, random forest

Procedia PDF Downloads 204

Authors: Javier E. García-Gallo, Nelson J. Fonseca-Ruiz, John F. Duitama-Munoz

Abstract:

Sepsis is a syndrome that occurs with physiological and biochemical abnormalities induced by severe infection and carries a high mortality and morbidity, therefore the severity of its condition must be interpreted quickly. After patient admission in an intensive care unit (ICU), it is necessary to synthesize the large volume of information that is collected from patients in a value that represents the severity of their condition. Traditional severity of illness scores seeks to be applicable to all patient populations, and usually assess in-hospital mortality. However, the use of machine learning techniques and the data of a population that shares a common characteristic could lead to the development of customized mortality prediction scores with better performance. This study presents the development of a score for the one-year mortality prediction of the patients that are admitted to an ICU with a sepsis diagnosis. 5650 ICU admissions extracted from the MIMICIII database were evaluated, divided into two groups: 70% to develop the score and 30% to validate it. Comorbidities, demographics and clinical information of the first 24 hours after the ICU admission were used to develop a mortality prediction score. LASSO (least absolute shrinkage and selection operator) and SGB (Stochastic Gradient Boosting) variable importance methodologies were used to select the set of variables that make up the developed score; each of this variables was dichotomized and a cut-off point that divides the population into two groups with different mean mortalities was found; if the patient is in the group that presents a higher mortality a one is assigned to the particular variable, otherwise a zero is assigned. These binary variables are used in a logistic regression (LR) model, and its coefficients were rounded to the nearest integer. The resulting integers are the point values that make up the score when multiplied with each binary variables and summed. The one-year mortality probability was estimated using the score as the only variable in a LR model. Predictive power of the score, was evaluated using the 1695 admissions of the validation subset obtaining an area under the receiver operating characteristic curve of 0.7528, which outperforms the results obtained with Sequential Organ Failure Assessment (SOFA), Oxford Acute Severity of Illness Score (OASIS) and Simplified Acute Physiology Score II (SAPSII) scores on the same validation subset. Observed and predicted mortality rates within estimated probabilities deciles were compared graphically and found to be similar, indicating that the risk estimate obtained with the score is close to the observed mortality, it is also observed that the number of events (deaths) is indeed increasing as the outcome go from the decile with the lowest probabilities to the decile with the highest probabilities. Sepsis is a syndrome that carries a high mortality, 43.3% for the patients included in this study; therefore, tools that help clinicians to quickly and accurately predict a worse prognosis are needed. This work demonstrates the importance of customization of mortality prediction scores since the developed score provides better performance than traditional scoring systems.

Keywords: intensive care, logistic regression model, mortality prediction, sepsis, severity of illness, stochastic gradient boosting

Procedia PDF Downloads 181