Anamai Na-Udom
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Anamai Na-Udom has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Jaratsri Rungrattanaubol
Naresuan University
6 shared publications Jaratsri Rungrattanaubol
Naresuan University
2 shared publications- Ratchaneewan Paisanwarakiat
Department of Mathematics, Faculty of Science, Naresuan University, Phitsanulok, Thailand
2 shared publications - Anong Kiddee
Division of Microbiology and Parasitology, School of Medical Sciences, University of Phayao, Phayao, 56000, Thailand; Unit of Excellence on Research and Application of Natural Products for Health and Well-Being, University of Phayao, Phayao, 56000, Thailand
1 shared publication - Antony Harfield
Department of Computer Science and IT,Faculty of Science, Naresuan University, Phitsanulok, Thailand
1 shared publication Kanit Assawatheptawee
Department of Microbiology and Parasitology, Faculty of Medical Science, Naresuan University, Phitsanulok, Thailand
1 shared publication
Latest Publications
- 1 April 2026
- Published by Pleiades Publishing Ltd in Lobachevskii Journal of Mathematics
- Vol. 47 (4) , 1783-1794
Exploring the Impact of Multivariate Imputation Techniques on the Performance of Classification Models
Missing data poses a substantial challenge in data-driven classification, as predictive performance is highly sensitive to the strategy employed to address missingness. This study investigates the effects of six imputation techniques for constructing classification models: Mean Imputation (MI), Deterministic Linear Regression Imputation (DLR), Expectation Maximization (EM), Random Forest Imputation (RFI), K Nearest Neighbors Imputation with weighted averaging (KNNI), and a proposed hybrid Random Forest-Expectation Maximization method (RFIEM). Experiments are conducted on six numerical datasets subjected to multivariate Missing Completely at Random (MCAR) mechanisms, with missingness levels ranging from 10\(\%\) to 50\(\%\) independently applied across input variables. A stratified ten-fold cross-validation framework is used to evaluate four classification algorithms: Logistic Regression (LR), Random Forest (RF), Artificial Neural Network (ANN), and Support Vector Machine (SVM). The average predictive accuracy of models is employed as a metric for evaluating and comparing the performance of different imputation techniques. Results demonstrate that EM, KNNI, and RFIEM consistently achieve superior performance for small datasets, particularly in combination with SVM and LR. For medium-sized datasets, RFI and RFIEM show robust performance, especially when paired with RF and LR. In large-scale datasets, RFI and RFIEM clearly outperform other imputation approaches when used with the RF classifier. These findings highlight the importance of selecting imputation methods that reflect dataset scale and classifier characteristics, and they provide practical guidance for improving predictive performance in the presence of missing data.- 1 April 2026
- Published by Pleiades Publishing Ltd in Lobachevskii Journal of Mathematics
- Vol. 47 (4) , 1783-1794
Exploring the Impact of Multivariate Imputation Techniques on the Performance of Classification Models
Missing data poses a substantial challenge in data-driven classification, as predictive performance is highly sensitive to the strategy employed to address missingness. This study investigates the effects of six imputation techniques for constructing classification models: Mean Imputation (MI), Deterministic Linear Regression Imputation (DLR), Expectation Maximization (EM), Random Forest Imputation (RFI), K Nearest Neighbors Imputation with weighted averaging (KNNI), and a proposed hybrid Random Forest-Expectation Maximization method (RFIEM). Experiments are conducted on six numerical datasets subjected to multivariate Missing Completely at Random (MCAR) mechanisms, with missingness levels ranging from 10\(\%\) to 50\(\%\) independently applied across input variables. A stratified ten-fold cross-validation framework is used to evaluate four classification algorithms: Logistic Regression (LR), Random Forest (RF), Artificial Neural Network (ANN), and Support Vector Machine (SVM). The average predictive accuracy of models is employed as a metric for evaluating and comparing the performance of different imputation techniques. Results demonstrate that EM, KNNI, and RFIEM consistently achieve superior performance for small datasets, particularly in combination with SVM and LR. For medium-sized datasets, RFI and RFIEM show robust performance, especially when paired with RF and LR. In large-scale datasets, RFI and RFIEM clearly outperform other imputation approaches when used with the RF classifier. These findings highlight the importance of selecting imputation methods that reflect dataset scale and classifier characteristics, and they provide practical guidance for improving predictive performance in the presence of missing data.- 3 April 2022
- Published by Springer Nature
- p. 88-98
Combining Logistic Regression Analysis with Data Mining Techniques to Predict Diabetes
- 3 April 2022
- Published by Springer Nature
- p. 88-98
Combining Logistic Regression Analysis with Data Mining Techniques to Predict Diabetes
- 1 October 2019
- Published by SAGE Publications in Microbial Drug Resistance
- Vol. 25 (8) , 1182-1190
Risk Factors for Extended-Spectrum β-Lactamase-Producing Enterobacteriaceae Carriage in Patients Admitted to Intensive Care Unit in a Tertiary Care Hospital in Thailand
Extended-spectrum β-lactamase-producing Enterobacteriaceae (ESBL-PE) are important causes of serious infections in intensive care unit (ICU). This study aimed to investigate the risk factors for intestinal carriage of ESBL-PE among patients admitted to ICU, subsequent ESBL-PE infections, and outcomes of these patients. This study prospectively collected rectal swabs from 215 ICU patients in Northern Thailand and ESBL-PE were isolated. A high prevalence of ESBL-PE carriage (134/215, 62.3%) at ICU admission was observed, with Escherichia coli representing the predominant organism (67.5%) followed by Klebsiella pneumoniae (19.4%). Multivariate logistic regression analysis identified chronic renal disease as the independent risk factor for ESBL-PE carriage ( p = 0.009; adjusted odds ratio = 4.369; 95% confidence interval = 1.455–13.119). Among colonized patients, 2.2% (3/134) developed ESBL-PE infections during ICU stay. Phylogenetic analysis of E. coli ( n = 108) showed that the predominant group was group A (38.0%), followed by groups B1 (17.6%), D (15.7%), B2 (14.8%), C (7.4%), and F (6.5%). Multilocus sequence typing analysis of the pathogenic groups B2, D, and F revealed 11 different sequence types (STs), with ST131 ( n = 13) as the most prevalent, followed by ST648 ( n = 5), ST38 ( n = 4), ST393 ( n = 3), and ST1193 ( n = 3). These results are of concern since ESBL-PE may be a prerequisite for endogenous infections and potentially disseminate within the hospital. This is the first study describing ESBL-PE carriage among patients at ICU admission and subsequent ESBL-PE infections in Thailand.
- 1 October 2019
- Published by SAGE Publications in Microbial Drug Resistance
- Vol. 25 (8) , 1182-1190
Risk Factors for Extended-Spectrum β-Lactamase-Producing Enterobacteriaceae Carriage in Patients Admitted to Intensive Care Unit in a Tertiary Care Hospital in Thailand
Extended-spectrum β-lactamase-producing Enterobacteriaceae (ESBL-PE) are important causes of serious infections in intensive care unit (ICU). This study aimed to investigate the risk factors for intestinal carriage of ESBL-PE among patients admitted to ICU, subsequent ESBL-PE infections, and outcomes of these patients. This study prospectively collected rectal swabs from 215 ICU patients in Northern Thailand and ESBL-PE were isolated. A high prevalence of ESBL-PE carriage (134/215, 62.3%) at ICU admission was observed, with Escherichia coli representing the predominant organism (67.5%) followed by Klebsiella pneumoniae (19.4%). Multivariate logistic regression analysis identified chronic renal disease as the independent risk factor for ESBL-PE carriage ( p = 0.009; adjusted odds ratio = 4.369; 95% confidence interval = 1.455–13.119). Among colonized patients, 2.2% (3/134) developed ESBL-PE infections during ICU stay. Phylogenetic analysis of E. coli ( n = 108) showed that the predominant group was group A (38.0%), followed by groups B1 (17.6%), D (15.7%), B2 (14.8%), C (7.4%), and F (6.5%). Multilocus sequence typing analysis of the pathogenic groups B2, D, and F revealed 11 different sequence types (STs), with ST131 ( n = 13) as the most prevalent, followed by ST648 ( n = 5), ST38 ( n = 4), ST393 ( n = 3), and ST1193 ( n = 3). These results are of concern since ESBL-PE may be a prerequisite for endogenous infections and potentially disseminate within the hospital. This is the first study describing ESBL-PE carriage among patients at ICU admission and subsequent ESBL-PE infections in Thailand.
- 20 June 2017
- Published by Springer Nature
- p. 3-12
Applying Stochastic Evolutionary Algorithm for Correlation Control in Monte Carlo Simulation
- 20 June 2017
- Published by Springer Nature
- p. 3-12
Applying Stochastic Evolutionary Algorithm for Correlation Control in Monte Carlo Simulation
- 1 January 2016
- Published by AIP Publishing in AIP Conference Proceedings
- Vol. 1775 (1) , 030040
The relation between optimality criteria and prediction accuracy for computer simulated experiments
Currently advancement in computational power leads to study of physical processes through deterministic computer simulated experiments (CSE). Normally CSEs are time consuming and computationally expensive to run. Hence a selection of proper design to run CSE is very important. The space filling designs, like Latin hypercube design (LHD) which aim to spread the design points over the region of interest are commonly used to generate the output of CSE. The optimal design for a given problem is obtained using search algorithms with a pre-specified optimality criterion. Searching for the optimal design is computational burden especially when the dimension of the problem is large. This paper aims to investigate the relation between the most commonly used optimality criterion namely ɸ p criterion and prediction accuracy of the best designs obtained from search algorithms. The results show that prediction capability would not be affected by different values of criteria as long as we terminate search algorithms within 20% of the threshold distance from the global minimum of ɸ p criterion for small dimensions of problem while a shorter threshold distance, as low as 10% from the global minimum is required for larger dimensions of problem.
- 1 January 2016
- Published by AIP Publishing in AIP Conference Proceedings
- Vol. 1775 (1) , 030040
The relation between optimality criteria and prediction accuracy for computer simulated experiments
Currently advancement in computational power leads to study of physical processes through deterministic computer simulated experiments (CSE). Normally CSEs are time consuming and computationally expensive to run. Hence a selection of proper design to run CSE is very important. The space filling designs, like Latin hypercube design (LHD) which aim to spread the design points over the region of interest are commonly used to generate the output of CSE. The optimal design for a given problem is obtained using search algorithms with a pre-specified optimality criterion. Searching for the optimal design is computational burden especially when the dimension of the problem is large. This paper aims to investigate the relation between the most commonly used optimality criterion namely ɸ p criterion and prediction accuracy of the best designs obtained from search algorithms. The results show that prediction capability would not be affected by different values of criteria as long as we terminate search algorithms within 20% of the threshold distance from the global minimum of ɸ p criterion for small dimensions of problem while a shorter threshold distance, as low as 10% from the global minimum is required for larger dimensions of problem.
- 1 January 2016
- Published by AIP Publishing in AIP Conference Proceedings
- Vol. 1775 (1) , 030054
An efficient rice yield predictive model in lower northern Thailand
This paper aims to investigate the factors influencing the rice yield by considering two main factors; environment and farmer’s practice. By using two different datasets, the first dataset is from Thai Meteorological Department corresponding to the farmer’s practice factor. The other is from 2nd office of Agriculture Economics representing the environment factor. The combining of these two datasets was done by applying geographic information technique. The construction of the efficient rice yield predictive model is developed based on three designed models, one with input variables from farmer’s practice only, the other two models with input variables from both factors, one with sum of rainfall and temperature and the other with their average. The input variables from farmer’s practice are rice type, method of cultivation, cultivation period, quantity of fertilizer and seed, and irrigation area, while the input variables from environment are rainfall and temperature during cultivation period. By applying Artificial Neural Network (ANN) to construct the model, the most efficient model was formed with the input variables from both factors with the value of root mean square error (RMSE) of 165.0289 and 181.9852 for training set and test set, respectively.
- 1 January 2016
- Published by AIP Publishing in AIP Conference Proceedings
- Vol. 1775 (1) , 030054
An efficient rice yield predictive model in lower northern Thailand
This paper aims to investigate the factors influencing the rice yield by considering two main factors; environment and farmer’s practice. By using two different datasets, the first dataset is from Thai Meteorological Department corresponding to the farmer’s practice factor. The other is from 2nd office of Agriculture Economics representing the environment factor. The combining of these two datasets was done by applying geographic information technique. The construction of the efficient rice yield predictive model is developed based on three designed models, one with input variables from farmer’s practice only, the other two models with input variables from both factors, one with sum of rainfall and temperature and the other with their average. The input variables from farmer’s practice are rice type, method of cultivation, cultivation period, quantity of fertilizer and seed, and irrigation area, while the input variables from environment are rainfall and temperature during cultivation period. By applying Artificial Neural Network (ANN) to construct the model, the most efficient model was formed with the input variables from both factors with the value of root mean square error (RMSE) of 165.0289 and 181.9852 for training set and test set, respectively.
- 18 February 2015
- Published by Springer Nature
- p. 745-752
A Comparison of Artificial Neural Network and Regression Model for Predicting the Rice Production in Lower Northern Thailand
- 18 February 2015
- Published by Springer Nature
- p. 745-752
A Comparison of Artificial Neural Network and Regression Model for Predicting the Rice Production in Lower Northern Thailand
- Open Access
- 1 January 1970
- Published by ECTI in ECTI Transactions on Computer and Information Technology (ECTI-CIT)
- Vol. 6 (2) , 176-185
Empirical Modelling for exploring the factors contributing to disability severity from road traffic accidents in Thailand
This paper introduces a computer-based model for predicting the severity of injuries in road traffic accidents. Using accident data from surveys at hospitals in Thailand, standard data mining techniques were applied to train and test a multilayer perceptron neural network. The resulting neural network specification was loaded into an interactive environment called EDEN that enables further exploration of the computer-based model. Although the model can be used for the classification of accident data in terms of injury severity (in a similar way to other data mining tools), the EDEN tool enables deeper exploration of the underlying factors that might affect injury severity in road traffic accidents. The aim of this paper is to describe the development of the computer-based model and to demonstrate the potential of EDEN as an interactive tool for knowledge discovery.
- 1 January 1970
- Published by ECTI in ECTI Transactions on Computer and Information Technology (ECTI-CIT)
- Vol. 6 (2) , 176-185
Open AccessEmpirical Modelling for exploring the factors contributing to disability severity from road traffic accidents in Thailand
This paper introduces a computer-based model for predicting the severity of injuries in road traffic accidents. Using accident data from surveys at hospitals in Thailand, standard data mining techniques were applied to train and test a multilayer perceptron neural network. The resulting neural network specification was loaded into an interactive environment called EDEN that enables further exploration of the computer-based model. Although the model can be used for the classification of accident data in terms of injury severity (in a similar way to other data mining tools), the EDEN tool enables deeper exploration of the underlying factors that might affect injury severity in road traffic accidents. The aim of this paper is to describe the development of the computer-based model and to demonstrate the potential of EDEN as an interactive tool for knowledge discovery.