Article
Computer Science, Information Systems
P. N. R. L. Chandra Sekhar, T. N. Shankar
Summary: With modern social networking, we can share thoughts and experiences with loved ones anywhere, but photo editing tools provide the opportunity to challenge the audience. When altered images go viral on social media, it can lead to loss of confidence and integrity. Therefore, there is a need for a reliable forensic technique to authenticate such images.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Pratiksha Meshram, Radha Krishna Rambola
Summary: This paper presents a novel approach using a convolutional neural network model to analyze facial images for detecting depression. The model utilizes user-generated data to differentiate between different depressive groups and predicts depression levels based on dynamic textual descriptions and psychiatric illness history. The proposed framework improves facial detection and feature extraction by 2.7% compared to existing frameworks.
Article
Geriatrics & Gerontology
Yang Ya, Lirong Ji, Yujing Jia, Nan Zou, Zhen Jiang, Hongkun Yin, Chengjie Mao, Weifeng Luo, Erlei Wang, Guohua Fan
Summary: This study developed machine learning models for the diagnosis of Parkinson's disease using multiple structural MRI features and validated their performance. The combined model showed favorable diagnostic performance and clinical net benefit, with potential as a non-invasive method for PD diagnosis.
FRONTIERS IN AGING NEUROSCIENCE
(2022)
Article
Computer Science, Information Systems
Sandipan Sahu, Raghvendra Kumar, Hoang Viet Long, Pathan Mohd Shafi
Summary: The Indian movie industry is the largest and most diverse industry in terms of the number of movies produced per year. Only a few movies achieve success, and revenue uncertainties create immense pressure on the industry. This study focuses on predicting the success of upcoming Indian movies through a K-fold Hybrid Deep Ensemble learning Model (KHDEM) that combines Deep Learning models and ensemble learning models. With the implementation of a Logistic Regression classifier, the study achieved 96% accuracy, outperforming all benchmark models. The introduction of derived features improved the accuracy by 17.62%. This study highlights the potential of predictive and prescriptive data analytics in supporting industry decisions.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Zhang Runchi, Xue Liguo, Wang Qin
Summary: This paper proposes a novel ensemble model called logistic-BWE based on logistic regression, which generates multiple training sub datasets using sample balancing algorithm and dynamically calculates the weight of predicted results based on the performance in the validation stage. Empirical results show that the logistic-BWE model has the strongest ability to recognize default samples, best generalization ability, and maintains interpretability.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Information Systems
K. V. T. K. N. Prashanth, Tene Ramakrishnudu
Summary: There is a growing interest in studying how to detect psychological stress from social media platforms like Twitter. This research addresses the issue of sparse data caused by Twitter's character limitation and proposes two solutions to leverage the text content for stress detection at the tweet level. The first solution introduces a new feature called "Sarcasm_Level" to measure the presence of sarcasm in tweets and its influence on stress detection. The second solution is a novel approach that incorporates the content of previous tweets, known as neighborhood tweets, for stress detection. Experimental results demonstrate that the proposed model outperforms other machine learning models in stress detection by incorporating information from neighborhood tweets and incorporating the new feature.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Geosciences, Multidisciplinary
Indrajit Chowdhuri, Subodh Chandra Pal, Rabin Chakrabortty, Sadhan Malik, Biswajit Das, Paramita Roy
Summary: Different techniques were used to predict landslide susceptibility, with the BRT-RF model showing exceptional predictive rates. The research findings have practical implications for planning and management to minimize landslide impacts.
Article
Biology
Hongwu Lv, Ke Yan, Yichen Guo, Quan Zou, Abd El-Latif Hesham, Bin Liu
Summary: In this paper, a novel AMP prediction method called AMPpred-EL is proposed, which utilizes ensemble learning strategy combined with LightGBM and logistic regression. The experimental results demonstrate that AMPpred-EL outperforms several state-of-the-art methods and improves efficiency performance.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Mathematics
Lu Liu, Junheng Gao, Georgia Beasley, Sin-Ho Jung
Summary: This paper discusses the standard approach of using machine learning methods to select features and build prediction models, and the issues with popular methods like LASSO and elastic net. The paper proposes a combination of standard regression methods and stepwise variable selection to overcome these issues and highlights the advantages of this method in terms of statistical significance and prediction accuracy compared to LASSO and elastic net.
Article
Public, Environmental & Occupational Health
Md. Abdul Wadood, Md. Rezaul Karim, Sheikh Md. Abu Hena Mostafa Alim, Md. Masud Rana, Md. Golam Hossain
Summary: The prevalence of depression is significantly high among married adults in Rajshahi City of Bangladesh, with females being the most vulnerable group. Multiple marriages, hard work, poor relationship with spouse, chronic medical comorbidity, and 7-12 years duration of conjugal life are the main factors associated with depression.
Article
Computer Science, Artificial Intelligence
Yilei Wang, Yiting Zhang, Xiujuan Zhang, Hai Liang, Guangshun Li, Xiaoying Wang
Summary: This paper analyzes real-time data of COVID-19 through visualization, and establishes a logistic growth model and a multiple-feature epidemic model to predict the development trend of the pandemic. The simulation results show that the predicted epidemic trend aligns with the actual trend, indicating the good performance of the models in predicting COVID-19.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2022)
Article
Engineering, Environmental
Lea Friedli, David Ginsbourger, Jonas Bhend
Summary: This study introduces a statistical postprocessing method for ensemble precipitation predictions, which leverages topographical covariates to improve the calibration of high-resolution ensemble forecasts. The approach is found to enhance model performance without the need for local historical data, as confirmed by a case study across Switzerland.
STOCHASTIC ENVIRONMENTAL RESEARCH AND RISK ASSESSMENT
(2021)
Article
Materials Science, Multidisciplinary
Sehyeon Kim, Insung Hwang, Dong-Yoon Kim, Young-Min Kim, Munjin Kang, Jiyoung Yu
Summary: An algorithm was proposed for predicting resistance spot weld quality based on quality acceptance criteria, successfully predicting geometrical and physical properties of spot-welded joints. The study utilized four statistical models to predict tensile shear strength, indentation depth, expulsion occurrence, and failure modes, achieving high prediction accuracies.
Article
Chemistry, Multidisciplinary
Jae-Yeong Lee, Ji-Sung Kim
Summary: This study presents a method for determining flood vulnerable areas using hydrological-topographic characteristics, establishing a model for calculating the flood vulnerability of the study area through data analysis. The model can be used to efficiently select target sites for flood prevention facilities and pre-detect areas vulnerable to flooding using real-time rainfall forecasting.
APPLIED SCIENCES-BASEL
(2021)
Article
Medicine, General & Internal
Seo-Eun Cho, Zong Woo Geem, Kyoung-Sae Na
Summary: Depression is a leading cause of disability worldwide, warranting appropriate screening for community dwellers. This study found that participants in the depression group had poorer socioeconomic, health, functional, and biological measures than those in the non-depression group, and utilized synthetic minority oversampling technique and least absolute shrinkage and selection operator for feature reduction and classifier modeling.