Search results for: empirical bayesian kriging regression prediction

Authors: Christian Neunzig, Simon Fahle, Jürgen Schulz, Matthias Möller, Bernd Kuhlenkötter

Abstract:

Digitalisation in production technology is a driver for the application of machine learning methods. Through the application of predictive quality, the great potential for saving necessary quality control can be exploited through the data-based prediction of product quality and states. However, the serial use of machine learning applications is often prevented by various problems. Fluctuations occur in real production data sets, which are reflected in trends and systematic shifts over time. To counteract these problems, data preprocessing includes rule-based data cleaning, the application of dimensionality reduction techniques, and the identification of comparable data subsets to extract stable features. Successful process control of the target variables aims to centre the measured values around a mean and minimise variance. Competitive leaders claim to have mastered their processes. As a result, much of the real data has a relatively low variance. For the training of prediction models, the highest possible generalisability is required, which is at least made more difficult by this data availability. The implementation of a machine learning application can be interpreted as a production process. The CRoss Industry Standard Process for Data Mining (CRISP-DM) is a process model with six phases that describes the life cycle of data science. As in any process, the costs to eliminate errors increase significantly with each advancing process phase. For the quality prediction of hydraulic test steps of directional control valves, the question arises in the initial phase whether a regression or a classification is more suitable. In the context of this work, the initial phase of the CRISP-DM, the business understanding, is critically compared for the use case at Bosch Rexroth with regard to regression and classification. The use of cross-process production data along the value chain of hydraulic valves is a promising approach to predict the quality characteristics of workpieces. Suitable methods for leakage volume flow regression and classification for inspection decision are applied. Impressively, classification is clearly superior to regression and achieves promising accuracies.

Keywords: classification, CRISP-DM, machine learning, predictive quality, regression

Procedia PDF Downloads 117

Authors: Deniz Peksen

Abstract:

This thesis aims to predict trend movement of closing price of stock and to maximize portfolio by utilizing the predictions. In this context, the study aims to define a stock portfolio strategy from models created by using Logistic Regression, Gradient Boosting and Random Forest. Recently, predicting the trend of stock price has gained a significance role in making buy and sell decisions and generating returns with investment strategies formed by machine learning basis decisions. There are plenty of studies in the literature on the prediction of stock prices in capital markets using machine learning methods but most of them focus on closing prices instead of the direction of price trend. Our study differs from literature in terms of target definition. Ours is a classification problem which is focusing on the market trend in next 20 trading days. To predict trend direction, fourteen years of data were used for training. Following three years were used for validation. Finally, last three years were used for testing. Training data are between 2002-06-18 and 2016-12-30 Validation data are between 2017-01-02 and 2019-12-31 Testing data are between 2020-01-02 and 2022-03-17 We determine Hold Stock Portfolio, Best Stock Portfolio and USD-TRY Exchange rate as benchmarks which we should outperform. We compared our machine learning basis portfolio return on test data with return of Hold Stock Portfolio, Best Stock Portfolio and USD-TRY Exchange rate. We assessed our model performance with the help of roc-auc score and lift charts. We use logistic regression, Gradient Boosting and Random Forest with grid search approach to fine-tune hyper-parameters. As a result of the empirical study, the existence of uptrend and downtrend of five stocks could not be predicted by the models. When we use these predictions to define buy and sell decisions in order to generate model-based-portfolio, model-based-portfolio fails in test dataset. It was found that Model-based buy and sell decisions generated a stock portfolio strategy whose returns can not outperform non-model portfolio strategies on test dataset. We found that any effort for predicting the trend which is formulated on stock price is a challenge. We found same results as Random Walk Theory claims which says that stock price or price changes are unpredictable. Our model iterations failed on test dataset. Although, we built up several good models on validation dataset, we failed on test dataset. We implemented Random Forest, Gradient Boosting and Logistic Regression. We discovered that complex models did not provide advantage or additional performance while comparing them with Logistic Regression. More complexity did not lead us to reach better performance. Using a complex model is not an answer to figure out the stock-related prediction problem. Our approach was to predict the trend instead of the price. This approach converted our problem into classification. However, this label approach does not lead us to solve the stock prediction problem and deny or refute the accuracy of the Random Walk Theory for the stock price.

Keywords: stock prediction, portfolio optimization, data science, machine learning

Procedia PDF Downloads 53

Authors: B. Sir, M. Podhoranyi, S. Kuchar, T. Kocyan

Abstract:

Rainfall-runoff models play important role in hydrological predictions. However, the model is only one part of the process for creation of flood prediction. The aim of this paper is to show the process of successful prediction for flood event (May 15–May 18 2014). The prediction was performed by rainfall runoff model HEC–HMS, one of the models computed within Floreon+ system. The paper briefly evaluates the results of automatic hydrologic prediction on the river Olše catchment and its gages Český Těšín and Věřňovice.

Keywords: flood, HEC-HMS, prediction, rainfall, runoff

Procedia PDF Downloads 364

Authors: Golnaz Shahtahmassebi, Jose Maria Sarabia

Abstract:

In this talk, we introduce a new class of conjugate prior distributions obtained from conditional specification methodology. We illustrate the application of such distribution in Bayesian change point detection in Poisson processes. We obtain the posterior distribution of model parameters using a general bivariate distribution with gamma conditionals. Simulation from the posterior is readily implemented using a Gibbs sampling algorithm. The Gibbs sampling is implemented even when using conditional densities that are incompatible or only compatible with an improper joint density. The application of such methods will be demonstrated using examples of simulated and real data.

Keywords: change point, bayesian inference, Gibbs sampler, conditional specification, gamma conditional distributions

Procedia PDF Downloads 156

Authors: P. V. Pramila , V. Mahesh

Abstract:

Pulmonary Function Tests are important non-invasive diagnostic tests to assess respiratory impairments and provides quantifiable measures of lung function. Spirometry is the most frequently used measure of lung function and plays an essential role in the diagnosis and management of pulmonary diseases. However, the test requires considerable patient effort and cooperation, markedly related to the age of patients esulting in incomplete data sets. This paper presents, a nonlinear model built using Multivariate adaptive regression splines and Random forest regression model to predict the missing spirometric features. Random forest based feature selection is used to enhance both the generalization capability and the model interpretability. In the present study, flow-volume data are recorded for N= 198 subjects. The ranked order of feature importance index calculated by the random forests model shows that the spirometric features FVC, FEF 25, PEF,FEF 25-75, FEF50, and the demographic parameter height are the important descriptors. A comparison of performance assessment of both models prove that, the prediction ability of MARS with the `top two ranked features namely the FVC and FEF 25 is higher, yielding a model fit of R2= 0.96 and R2= 0.99 for normal and abnormal subjects. The Root Mean Square Error analysis of the RF model and the MARS model also shows that the latter is capable of predicting the missing values of FEV1 with a notably lower error value of 0.0191 (normal subjects) and 0.0106 (abnormal subjects). It is concluded that combining feature selection with a prediction model provides a minimum subset of predominant features to train the model, yielding better prediction performance. This analysis can assist clinicians with a intelligence support system in the medical diagnosis and improvement of clinical care.

Keywords: FEV, multivariate adaptive regression splines pulmonary function test, random forest

Procedia PDF Downloads 276

Authors: Tomas Berhanu Bekele

Abstract:

Traffic congestion is one of the main problems related to transportation in developed as well as developing countries. Traffic control systems are based on the idea of avoiding traffic instabilities and homogenizing traffic flow in such a way that the risk of accidents is minimized and traffic flow is maximized. Lately, Intelligent Transport Systems (ITS) has become an important area of research to solve such road traffic-related issues for making smart decisions. It links people, roads and vehicles together using communication technologies to increase safety and mobility. Moreover, accurate prediction of road traffic is important to manage traffic congestion. The aim of this study is to develop an ANN model for the prediction of traffic flow and to compare the ANN model with the linear regression model of traffic flow predictions. Data extraction was carried out in intervals of 15 minutes from the video player. Video of mixed traffic flow was taken and then counted during office work in order to determine the traffic volume. Vehicles were classified into six categories, namely Car, Motorcycle, Minibus, mid-bus, Bus, and Truck vehicles. The average time taken by each vehicle type to travel the trap length was measured by time displayed on a video screen.

Keywords: intelligent transport system (ITS), traffic flow prediction, artificial neural network (ANN), linear regression

Procedia PDF Downloads 27

Authors: Yegnanew A. Shiferaw

Abstract:

Background: commodity price volatility originating from excessive commodity price fluctuation has been a global problem especially after the recent financial crises. Volatility is a measure of risk or uncertainty in financial analysis. It plays a vital role in risk management, portfolio management, and pricing equity. Objectives: the core objective of this paper is to examine the relationship between the prices of agricultural commodities with oil price, gas price, coal price and exchange rate (USD/Rand). In addition, the paper tries to fit an appropriate model that best describes the log return price volatility and estimate Value-at-Risk and expected shortfall. Data and methods: the data used in this study are the daily returns of agricultural commodity prices from 02 January 2007 to 31st October 2016. The data sets consists of the daily returns of agricultural commodity prices namely: white maize, yellow maize, wheat, sunflower, soya, corn, and sorghum. The paper applies the three-state Markov-switching (MS) regression, the standard single-regime GARCH and the two regime Markov-switching GARCH (MS-GARCH) models. Results: to choose the best fit model, the log-likelihood function, Akaike information criterion (AIC), Bayesian information criterion (BIC) and deviance information criterion (DIC) are employed under three distributions for innovations. The results indicate that: (i) the price of agricultural commodities was found to be significantly associated with the price of coal, price of natural gas, price of oil and exchange rate, (ii) for all agricultural commodities except sunflower, k=3 had higher log-likelihood values and lower AIC and BIC values. Thus, the three-state MS regression model outperformed the two-state MS regression model (iii) MS-GARCH(1,1) with generalized error distribution (ged) innovation performs best for white maize and yellow maize; MS-GARCH(1,1) with student-t distribution (std) innovation performs better for sorghum; MS-gjrGARCH(1,1) with ged innovation performs better for wheat, sunflower and soya and MS-GARCH(1,1) with std innovation performs better for corn. In conclusion, this paper provided a practical guide for modelling agricultural commodity prices by MS regression and MS-GARCH processes. This paper can be good as a reference when facing modelling agricultural commodity price problems.

Keywords: commodity prices, MS-GARCH model, MS regression model, South Africa, volatility

Procedia PDF Downloads 174

Authors: Queen Suraajini Rajendran, Sai Hung Cheung

Abstract:

Statistical down scaling models are required to connect the global climate model outputs and the local weather variables for climate change impact prediction. For reliable climate change impact studies, the uncertainty associated with the model including natural variability, uncertainty in the climate model(s), down scaling model, model inadequacy and in the predicted results should be quantified appropriately. In this work, a new approach is developed by the authors for statistical classification, statistical down scaling and uncertainty assessment and is applied to Singapore rainfall. It is a robust Bayesian uncertainty analysis methodology and tools based on coupling dependent modeling error with classification and statistical down scaling models in a way that the dependency among modeling errors will impact the results of both classification and statistical down scaling model calibration and uncertainty analysis for future prediction. Singapore data are considered here and the uncertainty and prediction results are obtained. From the results obtained, directions of research for improvement are briefly presented.

Keywords: statistical downscaling, global climate model, climate change, uncertainty

Procedia PDF Downloads 339

Authors: Tauha Hussain Ali, Shabir Hussain Khahro, Nafees Ahmed Memon

Abstract:

Accurate prediction of project time for planning and bid preparation stage should contain realistic dates. Constructors use their experience to estimate the project duration for the new projects, which is based on intuitions. It has been a constant concern to both researchers and constructors to analyze the accurate prediction of project duration for bid preparation stage. In Pakistan, such study for time cost relationship has been lacked to predict duration performance for the construction projects. This study is an attempt to explore the time cost relationship that would conclude with a mathematical model to predict the time for the drainage rehabilitation projects in the province of Sindh, Pakistan. The data has been collected from National Engineering Services (NESPAK), Pakistan and regression analysis has been carried out for the analysis of results. Significant relationship has been found between time and cost of the construction projects in Sindh and the generated mathematical model can be used by the constructors to predict the project duration for the upcoming projects of same nature. This study also provides the professionals with a requisite knowledge to make decisions regarding project duration, which is significantly important to win the projects at the bid stage.

Keywords: BTC Model, project time, relationship of time cost, regression

Procedia PDF Downloads 356

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 191

Authors: Ersin Sener, Ibrahim Demir, Hasan Aykut Karaboga, Kadir Kilinc

Abstract:

Bayesian methods, which have very wide range of applications, are implemented to the data obtained from the production of F class fly ash-based geopolymer bricks’ experimental design. In this study, dependent variable is compressive strength, independent variables are treatment type (oven and steam), treatment time, molding time, temperature, water absorbtion ratio and density. The effect of independent variables on compressive strength is investigated. There is no difference among treatment types, but there is a correlation between independent variables. Therefore, hierarchical Bayesian path analysis is applied. In consequence of analysis we specified that treatment time, temperature and density effects on compressive strength is higher, molding time, and water absorbtion ratio is relatively low.

Keywords: experimental design, F class fly ash, geopolymer bricks, hierarchical Bayesian path analysis

Procedia PDF Downloads 361

Authors: Pitsanu Tongkhow, Pichet Jiraprasertwong

Abstract:

This research investigates risk factors for defective products in autoparts factories. Under a Bayesian framework, a generalized linear mixed model (GLMM) in which the dependent variable, the number of defective products, has a Poisson distribution is adopted. Its performance is compared with the Poisson GLM under a Bayesian framework. The factors considered are production process, machines, and workers. The products coded RT50 are observed. The study found that the Poisson GLMM is more appropriate than the Poisson GLM. For the production Process factor, the highest risk of producing defective products is Process 1, for the Machine factor, the highest risk is Machine 5, and for the Worker factor, the highest risk is Worker 6.

Keywords: defective autoparts products, Bayesian framework, generalized linear mixed model (GLMM), risk factors

Procedia PDF Downloads 544

Authors: Chen Li-Ching

Abstract:

The multinomial logistic regression model is used popularly for inferring the relationship of risk factors and disease with multiple categories. This study based on the discrepancy between the nonparametric maximum likelihood estimator and semiparametric maximum likelihood estimator of the cumulative distribution function to propose a Kolmogorov-Smirnov type test statistic to assess adequacy of the multinomial logistic regression model for case-control data. A bootstrap procedure is presented to calculate the critical value of the proposed test statistic. Empirical type I error rates and powers of the test are performed by simulation studies. Some examples will be illustrated the implementation of the test.

Keywords: case-control studies, goodness-of-fit test, Kolmogorov-Smirnov test, multinomial logistic regression

Procedia PDF Downloads 427

Authors: Yu Zhao, Min Zhang, Yuanqian Wang, Qiuyu Yu

Abstract:

The prediction of aesthetics of apparel is helpful for the development of a new type of apparel. This study is to build the quantitative relationship between the aesthetics and its design parameters. In particular, women’s pants have been preliminarily studied. This aforementioned relationship has been carried out by statistical analysis. The contributions of this study include the development of a more personalized apparel design mechanism and the provision of some empirical knowledge for the development of other products in the aspect of aesthetics.

Keywords: aesthetics, crease line, cropped straight leg pants, knee width

Procedia PDF Downloads 161

Authors: D. Mohammadzadeh S., J. Bolouri Bazaz

Abstract:

This paper presents a new prediction model for compression index of fine-grained soils using multi-gene genetic programming (MGGP) technique. The proposed model relates the soil compression index to its liquid limit, plastic limit and void ratio. Several laboratory test results for fine-grained were used to develop the models. Various criteria were considered to check the validity of the model. The parametric and sensitivity analyses were performed and discussed. The MGGP method was found to be very effective for predicting the soil compression index. A comparative study was further performed to prove the superiority of the MGGP model to the existing soft computing and traditional empirical equations.

Keywords: new prediction model, compression index soil, multi-gene genetic programming, MGGP

Procedia PDF Downloads 343

Authors: Amylyn F. Labasano

Abstract:

A significant number of empirical studies are carried out to investigate factors affecting college students’ performance in the academic examination. With a wide-array of literature-and studies-supported findings, this study is limited only on the students’ probability of passing periodical exams which is associated with students’ gender, absences in the class, use of reference book, and hours of study. Binary logistic regression was the technique used in the analysis. The research is based on the students’ record and data collected through survey. The result reveals that gender, use of reference book and hours of study are significant predictors of passing an examination while students’ absenteeism is an insignificant predictor. Females have 45% likelihood of passing the exam than their male classmates. Students who use and read their reference book are 38 times more likely pass the exam than those who do not use and read their reference book. Those who spent more than 3 hours in studying are four (4) times more likely pass the exam than those who spent only 3 hours or less in studying.

Keywords: absences, binary logistic regression, gender, hours of study prediction-causation method, periodical exams, random sampling, reference book

Procedia PDF Downloads 281

Authors: N. H. Adenan, M. S. M. Noorani

Abstract:

River flow prediction is an essential to ensure proper management of water resources can be optimally distribute water to consumers. This study presents an analysis and prediction by using nonlinear prediction method involving monthly river flow data in Tanjung Tualang from 1976 to 2006. Nonlinear prediction method involves the reconstruction of phase space and local linear approximation approach. The phase space reconstruction involves the reconstruction of one-dimensional (the observed 287 months of data) in a multidimensional phase space to reveal the dynamics of the system. Revenue of phase space reconstruction is used to predict the next 72 months. A comparison of prediction performance based on correlation coefficient (CC) and root mean square error (RMSE) have been employed to compare prediction performance for nonlinear prediction method, ARIMA and SVM. Prediction performance comparisons show the prediction results using nonlinear prediction method is better than ARIMA and SVM. Therefore, the result of this study could be used to developed an efficient water management system to optimize the allocation water resources.

Keywords: river flow, nonlinear prediction method, phase space, local linear approximation

Procedia PDF Downloads 386

Authors: H. Bonakdari, I. Ebtehaj

Abstract:

The prediction of scour depth around bridge piers is frequently considered in river engineering. One of the key aspects in efficient and optimum bridge structure design is considered to be scour depth estimation around bridge piers. In this study, scour depth around bridge piers is estimated using two methods, namely the Adaptive Neuro-Fuzzy Inference System (ANFIS) and Artificial Neural Network (ANN). Therefore, the effective parameters in scour depth prediction are determined using the ANN and ANFIS methods via dimensional analysis, and subsequently, the parameters are predicted. In the current study, the methods’ performances are compared with the nonlinear regression (NLR) method. The results show that both methods presented in this study outperform existing methods. Moreover, using the ratio of pier length to flow depth, ratio of median diameter of particles to flow depth, ratio of pier width to flow depth, the Froude number and standard deviation of bed grain size parameters leads to optimal performance in scour depth estimation.

Keywords: adaptive neuro-fuzzy inference system (ANFIS), artificial neural network (ANN), bridge pier, scour depth, nonlinear regression (NLR)

Procedia PDF Downloads 187

Authors: Kiyoumars Roushangar, Saeid Sadaghian

Abstract:

Although accurate sediment load prediction is very important in planning, designing, operating and maintenance of water resources structures, the transport mechanism is complex, and the deterministic transport models are based on simplifying assumptions often lead to large prediction errors. In this research, firstly, two intelligent ANN methods, Radial Basis and General Regression Neural Networks, are adopted to model of total sediment load transport into Madani Dam reservoir (north of Iran) using the measured data and then applicability of the sediment transport methods developed by Engelund and Hansen, Ackers and White, Yang, and Toffaleti for predicting of sediment load discharge are evaluated. Based on comparison of the results, it is found that the GRNN model gives better estimates than the sediment rating curve and mentioned classic methods.

Keywords: sediment transport, dam reservoir, RBF, GRNN, prediction

Procedia PDF Downloads 471

Authors: Suparman Suparman

Abstract:

A white noise in autoregressive (AR) model is often assumed to be normally distributed. In application, the white noise usually do not follows a normal distribution. This paper aims to estimate a parameter of AR model that has a exponential white noise. A Bayesian method is adopted. A prior distribution of the parameter of AR model is selected and then this prior distribution is combined with a likelihood function of data to get a posterior distribution. Based on this posterior distribution, a Bayesian estimator for the parameter of AR model is estimated. Because the order of AR model is considered a parameter, this Bayesian estimator cannot be explicitly calculated. To resolve this problem, a method of reversible jump Markov Chain Monte Carlo (MCMC) is adopted. A result is a estimation of the parameter AR model can be simultaneously calculated.

Keywords: autoregressive (AR) model, exponential white Noise, bayesian, reversible jump Markov Chain Monte Carlo (MCMC)

Procedia PDF Downloads 332

Authors: Zeytu G. Asfaw, Serkalem K. Abrha, Demisew G. Degefu

Abstract:

Background: HIV/AIDS remains a major public health problem in Ethiopia and heavily affecting people of productive and reproductive age. We aimed to compare the performance of Parametric Survival Analysis and Bayesian Survival Analysis using simulations and in a real dataset application focused on determining predictors of HIV patient survival. Methods: A Parametric Survival Models - Exponential, Weibull, Log-normal, Log-logistic, Gompertz and Generalized gamma distributions were considered. Simulation study was carried out with two different algorithms that were informative and noninformative priors. A retrospective cohort study was implemented for HIV infected patients under Highly Active Antiretroviral Therapy in Alamata General Hospital, North Ethiopia. Results: A total of 320 HIV patients were included in the study where 52.19% females and 47.81% males. According to Kaplan-Meier survival estimates for the two sex groups, females has shown better survival time in comparison with their male counterparts. The median survival time of HIV patients was 79 months. During the follow-up period 89 (27.81%) deaths and 231 (72.19%) censored individuals registered. The average baseline cluster of differentiation 4 (CD4) cells count for HIV/AIDS patients were 126.01 but after a three-year antiretroviral therapy follow-up the average cluster of differentiation 4 (CD4) cells counts were 305.74, which was quite encouraging. Age, functional status, tuberculosis screen, past opportunistic infection, baseline cluster of differentiation 4 (CD4) cells, World Health Organization clinical stage, sex, marital status, employment status, occupation type, baseline weight were found statistically significant factors for longer survival of HIV patients. The standard error of all covariate in Bayesian log-normal survival model is less than the classical one. Hence, Bayesian survival analysis showed better performance than classical parametric survival analysis, when subjective data analysis was performed by considering expert opinions and historical knowledge about the parameters. Conclusions: Thus, HIV/AIDS patient mortality rate could be reduced through timely antiretroviral therapy with special care on the potential factors. Moreover, Bayesian log-normal survival model was preferable than the classical log-normal survival model for determining predictors of HIV patients survival.

Keywords: antiretroviral therapy (ART), Bayesian analysis, HIV, log-normal, parametric survival models

Procedia PDF Downloads 160

Authors: Yijun Shao, Yan Cheng, Rashmee U. Shah, Charlene R. Weir, Bruce E. Bray, Qing Zeng-Treitler

Abstract:

Deep neural network (DNN) models are being explored in the clinical domain, following the recent success in other domains such as image recognition. For clinical adoption, outcome prediction models require explanation, but due to the multiple non-linear inner transformations, DNN models are viewed by many as a black box. In this study, we developed a deep neural network model for predicting 1-year mortality of patients who underwent major cardio vascular procedures (MCVPs), using temporal image representation of past medical history as input. The dataset was obtained from the electronic medical data warehouse administered by Veteran Affairs Information and Computing Infrastructure (VINCI). We identified 21,355 veterans who had their first MCVP in 2014. Features for prediction included demographics, diagnoses, procedures, medication orders, hospitalizations, and frailty measures extracted from clinical notes. Temporal variables were created based on the patient history data in the 2-year window prior to the index MCVP. A temporal image was created based on these variables for each individual patient. To generate the explanation for the DNN model, we defined a new concept called impact score, based on the presence/value of clinical conditions’ impact on the predicted outcome. Like (log) odds ratio reported by the logistic regression (LR) model, impact scores are continuous variables intended to shed light on the black box model. For comparison, a logistic regression model was fitted on the same dataset. In our cohort, about 6.8% of patients died within one year. The prediction of the DNN model achieved an area under the curve (AUC) of 78.5% while the LR model achieved an AUC of 74.6%. A strong but not perfect correlation was found between the aggregated impact scores and the log odds ratios (Spearman’s rho = 0.74), which helped validate our explanation.

Keywords: deep neural network, temporal data, prediction, frailty, logistic regression model

Procedia PDF Downloads 134

Authors: Abdulrahman Sumayli, Saad M. AlShahrani

Abstract:

For the hydrogenation process, knowing the solubility of hydrogen (H2) in hydrocarbons is critical to improve the efficiency of the process. We investigated the H2 solubility computation in four heavy crude oil feedstocks using machine learning techniques. Temperature, pressure, and feedstock type were considered as the inputs to the models, while the hydrogen solubility was the sole response. Specifically, we employed three different models: Support Vector Regression (SVR), Gaussian process regression (GPR), and Bayesian ridge regression (BRR). To achieve the best performance, the hyper-parameters of these models are optimized using the whale optimization algorithm (WOA). We evaluated the models using a dataset of solubility measurements in various feedstocks, and we compared their performance based on several metrics. Our results show that the WOA-SVR model tuned with WOA achieves the best performance overall, with an RMSE of 1.38 × 10− 2 and an R-squared of 0.991. These findings suggest that machine learning techniques can provide accurate predictions of hydrogen solubility in different feedstocks, which could be useful in the development of hydrogen-related technologies. Besides, the solubility of hydrogen in the four heavy oil fractions is estimated in different ranges of temperatures and pressures of 150 ◦C–350 ◦C and 1.2 MPa–10.8 MPa, respectively

Keywords: temperature, pressure variations, machine learning, oil treatment

Procedia PDF Downloads 43

Authors: Eva L. Sanjuán, Jacinto Martín, M. Isabel Parra, Mario M. Pizarro

Abstract:

In Extreme Value Theory, inference estimation for the parameters of the distribution is made employing a small part of the observation values. When block maxima values are taken, many data are discarded. We developed a new Bayesian inference model to seize all the information provided by the data, introducing informative priors and using the relations between baseline and limit parameters. Firstly, we studied the accuracy of the new model for three baseline distributions that lead to a Gumbel extreme distribution: Exponential, Normal and Gumbel. Secondly, we considered mixtures of Normal variables, to simulate practical situations when data do not adjust to pure distributions, because of perturbations (noise).

Keywords: bayesian inference, extreme value theory, Gumbel distribution, highly informative prior

Procedia PDF Downloads 168

Authors: The Danh Phan

Abstract:

House price forecasting is a main topic in the real estate market research. Effective house price prediction models could not only allow home buyers and real estate agents to make better data-driven decisions but may also be beneficial for the property policymaking process. This study investigates the housing market by using machine learning techniques to analyze real historical house sale transactions in Australia. It seeks useful models which could be deployed as an application for house buyers and sellers. Data analytics show a high discrepancy between the house price in the most expensive suburbs and the most affordable suburbs in the city of Melbourne. In addition, experiments demonstrate that the combination of Stepwise and Support Vector Machine (SVM), based on the Mean Squared Error (MSE) measurement, consistently outperforms other models in terms of prediction accuracy.

Keywords: house price prediction, regression trees, neural network, support vector machine, stepwise

Procedia PDF Downloads 190

Authors: Jing Yuan, Hongwei Yang

Abstract:

This paper aims to understand U.S. middle school students’ mathematics achievements by examining relevant student and school-level predictors. Through a variance component analysis, the study first identifies evidence supporting the use of multilevel modeling. Then, a multilevel analysis is performed under Bayesian statistical inference where prior information is incorporated into the modeling process. During the analysis, independent variables are entered sequentially in the order of theoretical importance to create a hierarchy of models. By evaluating each model using Bayesian fit indices, a best-fit and most parsimonious model is selected where Bayesian statistical inference is performed for the purpose of result interpretation and discussion. The primary dataset for Bayesian modeling is derived from the Program for International Student Assessment (PISA) in 2012 with a secondary PISA dataset from 2003 analyzed under the traditional ordinary least squares method to provide the information needed to specify informative priors for a subset of the model parameters. The dependent variable is a composite measure of mathematics literacy, calculated from an exploratory factor analysis of all five PISA 2012 mathematics achievement plausible values for which multiple evidences are found supporting data unidimensionality. The independent variables include demographics variables and content-specific variables: mathematics efficacy, teacher-student ratio, proportion of girls in the school, etc. Finally, the entire analysis is performed using the MCMCpack and MCMCglmm packages in R.

Keywords: Bayesian multilevel modeling, mathematics education, PISA, multilevel

Procedia PDF Downloads 304

Authors: Muhammet Baldan, Emel Timuçin

Abstract:

Generally, absorption and bioavailability increase if solubility increases; therefore, it is crucial to predict them in drug discovery applications. Molecular descriptors and Molecular properties are traditionally used for the prediction of water solubility. There are various key descriptors that are used for this purpose, namely Drogan Descriptors, Morgan Descriptors, Maccs keys, etc., and each has different prediction capabilities with differentiating successes between different data sets. Another source for the prediction of solubility is structural features; they are commonly used for the prediction of solubility. However, there are little to no studies that combine three or more properties or descriptors for prediction to produce a more powerful prediction model. Unlike available models, we used a combination of those features in a random forest machine learning model for improved solubility prediction to better predict and, therefore, contribute to drug discovery systems.

Keywords: solubility, molecular descriptors, machine learning, random forest

Procedia PDF Downloads 24

Authors: B. Elshafei, X. Mao

Abstract:

The demand for renewable energy is significantly increasing, major investments are being supplied to the wind power generation industry as a leading source of clean energy. The wind energy sector is entirely dependable and driven by the prediction of wind speed, which by the nature of wind is very stochastic and widely random. This s0tudy employs deep multi-fidelity Gaussian process regression, used to predict wind speeds for medium term time horizons. Data of the RUNE experiment in the west coast of Denmark were provided by the Technical University of Denmark, which represent the wind speed across the study area from the period between December 2015 and March 2016. The study aims to investigate the effect of pre-processing the data by denoising the signal using empirical wavelet transform (EWT) and engaging the vector components of wind speed to increase the number of input data layers for data fusion using deep multi-fidelity Gaussian process regression (GPR). The outcomes were compared using root mean square error (RMSE) and the results demonstrated a significant increase in the accuracy of predictions which demonstrated that using vector components of the wind speed as additional predictors exhibits more accurate predictions than strategies that ignore them, reflecting the importance of the inclusion of all sub data and pre-processing signals for wind speed forecasting models.

Keywords: data fusion, Gaussian process regression, signal denoise, temporal extrapolation

Procedia PDF Downloads 111

Authors: Farouk Metiri, Halim Zeghdoudi, Mohamed Riad Remita

Abstract:

Credibility theory is an experience rating technique in actuarial science which can be seen as one of quantitative tools that allows the insurers to perform experience rating, that is, to adjust future premiums based on past experiences. It is used usually in automobile insurance, worker's compensation premium, and IBNR (incurred but not reported claims to the insurer) where credibility theory can be used to estimate the claim size amount. In this study, we focused on a popular tool in credibility theory which is the Bayesian premium estimator, considering Lindley distribution as a claim distribution. We derive this estimator under entropy loss which is asymmetric and squared error loss which is a symmetric loss function with informative and non-informative priors. In a purely Bayesian setting, the prior distribution represents the insurer’s prior belief about the insured’s risk level after collection of the insured’s data at the end of the period. However, the explicit form of the Bayesian premium in the case when the prior is not a member of the exponential family could be quite difficult to obtain as it involves a number of integrations which are not analytically solvable. The paper finds a solution to this problem by deriving this estimator using numerical approximation (Lindley approximation) which is one of the suitable approximation methods for solving such problems, it approaches the ratio of the integrals as a whole and produces a single numerical result. Simulation study using Monte Carlo method is then performed to evaluate this estimator and mean squared error technique is made to compare the Bayesian premium estimator under the above loss functions.

Keywords: bayesian estimator, credibility theory, entropy loss, monte carlo simulation

Procedia PDF Downloads 298

Authors: N. Alhazmi

Abstract:

Knowing the thermodynamics properties of hydrocarbons is vital when it comes to analyzing the related chemical reaction outcomes and understanding the reaction process, especially in terms of petrochemical industrial applications, combustions, and catalytic reactions. However, measuring the thermodynamics properties experimentally is time-consuming and costly. In this paper, Gaussian process regression (GPR) has been used to directly predict the main thermodynamic properties - standard enthalpy of formation, standard entropy, and heat capacity -for more than 360 cyclic and non-cyclic alkanes, alkenes, and alkynes. A simple workflow has been proposed that can be applied to directly predict the main properties of any hydrocarbon by knowing its descriptors and chemical structure and can be generalized to predict the main properties of any material. The model was evaluated by calculating the statistical error R², which was more than 0.9794 for all the predicted properties.

Keywords: thermodynamic, Gaussian process regression, hydrocarbons, regression, supervised learning, entropy, enthalpy, heat capacity

Procedia PDF Downloads 190