You are currently viewing a new version of our website. To view the old version click .
MDPIMDPI
Search all articles
Life
Download PDF
  • Article
  • Open Access

25 October 2024

Development of an AI-Based Suicide Ideation Prediction Model for People with Disabilities

Department of Public Health, Korea University College of Medicine, Seoul 02841, Republic of Korea
Life2024, 14(11), 1372;https://doi.org/10.3390/life14111372
This article belongs to the Section Life Sciences
Version Notes
Order Reprints

Abstract

South Korea has one of the highest suicide rates among countries in the Organisation for Economic Co-Operation and Development, and the suicide rate among people with disabilities is more than twice that of the general population. This study aimed to develop an artificial intelligence-based suicide ideation prediction model for people with disabilities in order to provide a proactive approach for managing high-risk groups and offer evidence for establishing suicide prevention policies. The support vector machine, adaptive boost (AdaBoost), and bidirectional long short-term memory (Bi-LSTM) models were used in this study. Data from the Disability and Life Dynamics Panel for 2018–2021 were used. The performance of the models was evaluated based on the accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). All the prediction models demonstrated excellent performance, with AUC > 0.80 (0.83–0.87). The best-performing models were AdaBoost (0.87) for accuracy, Bi-LSTM (0.90) for sensitivity, and AdaBoost (0.90) for specificity. This study is the first to develop an artificial intelligence-based suicide ideation prediction model for disabled people and is significant in that it suggests ways to pre-emptively manage groups at high risk for suicide, providing evidence for the establishment of suicide prevention policies.

1. Introduction

Suicide is the major cause of death worldwide, with more than 700,000 people dying by suicide yearly [1]. Therefore, reducing the suicide mortality rate by one-third by 2030 is one of the targets and indicators of the United Nations Sustainable Development Goals (UN SDGs) [1,2] and the World Health Organization’s (WHO) Comprehensive Mental Health Action Plan 2013–2030 [1]. In 2020, South Korea recorded the highest suicide rate among OECD countries [3]. In 2021, the suicide rate was 26.0 per 100,000 people, with an average of 36.6 suicides per day and a total of 13,352 suicides annually, with increases of 0.3, 0.4, and 157, respectively, compared with the previous year [4]. Korea’s suicide rate sharply increased during the 1998 financial crisis and then further increased during the credit card crisis in 2003 and the global financial crisis in 2009, reaching a peak in 2011 (31.7 per 100,000 people). It then decreased until 2017 but slightly increased in 2021, compared with the previous year [4]. Suicide was the leading cause of death among individuals in their 10s to 30s in 2021 and the second-leading cause of death in individuals in their 40s and 50s [4]. Therefore, to prevent suicide, Korea established the ‘Basic Plan for Suicide Prevention Measures’ in 2004 and enacted the ‘Law for Suicide Prevention and Creating a Culture of Respect for Life’ in 2011. The central government, local governments, and the private sector have been working together to promote various suicide prevention projects [4].
The number of registered people with disabilities in South Korea has continuously increased over the past 20 years; it reached 2,641,896 in 2023, accounting for 5.1% of the total population [5]. People with disabilities face difficulties leading independent lives because of their disabilities, which often limit their participation in social activities, including employment. This leads to economic difficulties, which is a major risk factor for suicide. Additionally, the continuous reliance on medical institutions imposes a further economic burden, heightening the likelihood of experiencing psychological distress such as depression and anxiety [6]. In the 2021 Mental Health Screening for People with Disabilities, significant percentages of people were suffering from depression, a major risk factor for suicide, with people in their 20s, 30s, 40s, 50s, 60s, and 70s having depression rates of 23.1%, 28.2%, 29.9%, 23.5%, 19.9%, and 14.2%, respectively [7].
Suicide is also a major cause of death in people with disabilities, ranking as the third leading cause of death among those in their 20s and second among those in their 30s as of 2022 [7]. Moreover, the crude death rate for people with disabilities in 2022 was 53.1 per 100,000 people [8], which was more than double the crude death rate of 25.2% in the general population [8]. Therefore, continued efforts at the national level, including legal and institutional reforms, are necessary to prevent suicide in people with disabilities.
The WHO recommends that each country establishes and promotes government-led suicide prevention strategies, as follows: (1) restriction of access to suicide means; (2) interaction with the media for responsible suicide reporting; (3) promotion of adolescents’ social–emotional life skills; and (4) early identification, evaluation, management, and follow-up of individuals affected by suicidal behaviour [1]. Suicide ideation often leads to suicide; hence, it has been used as an important primary indicator for predicting suicide [9] and as evidence for establishing policies for suicide prevention. Therefore, various studies have been conducted to identify the factors that influence suicide ideation among people with disabilities in South Korea. A review of key prior studies showed that many studies did not differentiate between the types of disabilities [9,10,11,12,13,14,15,16,17,18,19], while others focused on specific disabilities, such as kidney disorders [20], mental disorders [21,22], or physical and brain disabilities [6,23,24,25,26,27]. However, most previous studies used the results of a single survey, causing a wide variation in the reported effects of population and sociological factors, disability and health factors, socioeconomic conditions, and psychological and environmental influences on the suicide ideation of people with disabilities. Consequently, the factors influencing suicide ideation among people with disabilities have remained unclear. Therefore, further research is needed to identify the determinants of suicide ideation among individuals with disabilities.
Recently, there has been an increasing trend in the use of big data and artificial intelligence (AI) in suicide prevention strategies. In the United States, the Durkheim Project was implemented to prevent suicide among specific groups (such as veterans) by integrating public data platforms, such as social media, with healthcare databases and using machine learning to monitor real-time linguistic and behavioural patterns statistically associated with suicide [28,29]. In New York, the Office of Mental Health has utilised various databases, including the Psychiatric Services and Clinical Knowledge Enhancement System (PSYCKES), to develop data-driven suicide prevention strategies [30,31]. Additionally, Australia’s Black Dog Institute has launched the LifeSpan Project, which uses big data to identify regions with high suicide rates and limit access to means of suicide by establishing suicide prevention infrastructure in those areas [28]. Such attempts have continued in Korea, and in February 2023, using a suicide prevention system based on AI technology, Tongyeong saved a woman in her 20s who tried to die by suicide [32]. Furthermore, in May 2023, the Presidential Suicide Crisis Overcome Special Committee emphasised the need for a scientific and precise suicide prevention policy by developing a “suicide prediction model” that could identify high-risk groups early, including the potential for integrating AI technologies [33].
Therefore, this study aimed to develop an AI-based suicide ideation prediction model using a Disability and Life Dynamics Panel that is representative of people with disabilities in South Korea. This study sought to propose proactive management strategies for high-risk groups and provide evidence for the development of suicide prevention policies for people with disabilities. For this purpose, support vector machine (SVM) and adaptive boost (AdaBoost) of machine learning, and bidirectional long short-term memory (Bi-LSTM) of deep learning were selected as study models.

2. Materials and Methods

2.1. Study Population and Ethical Considerations

This study selected the Disability and Life Dynamics Panel (Approval No. 438001) [34] as the analysis data. These include individuals with disabilities and their household members who registered their disabilities with the Ministry of Health and Welfare between 1 January 2015 and 31 December 2017, in accordance with Article 32 of the Act on Welfare of Persons with Disabilities. The Life Dynamics Panel was launched in 2018 to provide foundational data for formulating welfare policies for people with disabilities by understanding the process of disability acceptance and the changes experienced in social relationships. The survey covered the following areas: (1) acceptance of and changes in disability; (2) health and medical care; (3) independence; and (4) social participation [34].
The data for this study were obtained from the Korea Disabled People’s Development Institute, and 19,141 responses from the 2018 (1st wave) to 2021 (4th wave) surveys, which included answers to the dependent variable, suicide ideation, were used.
Ethical approval was not required because this study utilised existing survey data. Therefore, an exemption from review was granted by the Korea University Institutional Review Board (KUIRB-2023-0413-01) on 1 December 2023.

2.2. Variables and Categories

The dependent variable was suicide ideation. Through an analysis of 13 previous studies on factors influencing suicide ideation among people with disabilities [6,9,10,11,12,13,14,16,20,21,23,24,25], 39 independent variables were selected (Table 1).
Table 1. Variables and categories.

2.3. Statistical Analysis

AI encompasses machine and deep learning and is often defined as research aimed at automating tasks that humans perform intelligently [42]. Machine learning primarily focuses on predicting the outcomes for new data by learning from existing data [43], whereas deep learning, a subset of machine learning, excels at learning meaningful representations through sequential layers, making it particularly effective at learning representations from data [42].
AI encompasses different learning methods, each with differing learning effects based on data suitability and on the algorithm being used. As such, this study attempted to compare the prediction results of the models based on three algorithms. For this study, the AI models selected were SVM, AdaBoost from the field of machine learning, and Bi-LSTM from deep learning. Data pre-processing and analysis were conducted using Python version 3.12.4. During the pre-processing stage, continuous variables were standardised by adjusting their mean and variance to zero and one, respectively [44], whereas target encoding was applied to categorical variables by replacing individual categories with their mean values [45]. The dataset was split into 99.5% for training data and 0.5% for testing data.
Given that only 8.48% (n = 1623) of the participants (n = 19,141) reported experiencing suicide ideation, there was a class imbalance in the data. To address this imbalance, oversampling was performed on the training data using the synthetic minority oversampling technique, nominally continuous (SMOTE-NC), which is suitable for datasets containing both categorical and continuous variables [46]. For model interpretation, the permutation feature importance (PFI) approach was applied, which measures the change in model performance when the values of the variables are shuffled or permuted [47]. The performance of the models was evaluated using accuracy, sensitivity, specificity, and the area under the receiver operating characteristic (ROC) curve (AUC) [48].

2.3.1. Confusion Matrix

The definition of the four parameters used to evaluate model performance can be explained using a confusion matrix (Table 2) [48].
Table 2. Confusion matrix.
■
True positive: A case in which a model correctly predicts a positive value when the actual value is positive.
■
True negative: A case in which a model correctly predicts a negative when the actual value is negative.
■
False positive: A case in which a model incorrectly predicts a positive value when the actual value is negative.
■
False negative: A case in which a model incorrectly predicts a negative value when the actual value is positive.

2.3.2. Accuracy

Accuracy is an indicator that represents the proportion of correct predictions from the total data [48]. Accuracy refers to the accuracy of the model’s predictions for the entire dataset, including positive and negative instances of suicide ideation.
Accuracy = TP + TN TP + FP + TN + FN

2.3.3. Sensitivity (True Positive Rate or Recall)

Sensitivity is an indicator that represents the proportion of actual positive data that the model correctly predicts as positive [48]. In this study, sensitivity refers to the proportion of individuals with suicide ideation correctly identified by the model.
Sensitivity = TP TP + FN

2.3.4. Specificity (True Negative Rate)

Specificity is an indicator representing the proportion of actual negative data that the model correctly predicts as negative [48]. Specificity refers to the proportion of individuals without suicide ideation that the model correctly identified as not having suicide ideation.
Specificity = TN TN + FP

2.3.5. AUC

The ROC curve is a graph that plots sensitivity on the y-axis and 1-specificity on the x-axis, illustrating the trade-off between true positive and false positive rates at various threshold levels [48]. The AUC represents the area under this curve and is used to evaluate the overall classification performance of the model [48].

3. Results

3.1. Descriptive Statistics

The results of the descriptive statistics analysis of the participants of this study are shown in Table 3.
Table 3. Results of descriptive statistics.
The study participants (n = 19,141) comprised 54.7% male and 45.3% female (Figure 1). The age distribution was as follows: 4.3% in their 10s, 6.2% in their 20s, 5.9% in their 30s, 11.6% in their 40s, 27.6% in their 50s, 35.4% in their 60s, and 9% in their 70s, with the largest group being those in their 60s. Regarding marital status, 23%were unmarried, 52% married, and 25% divorced/separated/bereaved. In terms of disability type, 17.4% had physical disability, 15.6% had brain lesion disability, 13.3% had visual impairment, 12.5% had hearing impairment, 4.3% had speech impairment, 0.5% had facial disability, 10.6% had kidney failure, 2.1% had heart failure, 3.6% had liver failure, 2.7% had respiratory failure, 2.5% had stomatal urinary tract disorder, 2.1% had epilepsy, 5.8% had intellectual disability, 1% had autism, and 6%had a mental disorder (Figure 2). Physical disabilities were the most prevalent type.
Figure 1. Distribution of sex.
Figure 2. Distribution of type of disability.

3.2. SVM

SVM is a traditional machine learning technique that can be applied to both classification and regression problems, offering high generalisation performance [49,50]. It is frequently used because of its simplicity, computational efficiency, and ability to train with small datasets [51,52,53]. To solve classification problems, such as the one in this study, SVM aims to find an optimal boundary between two different categories, where the optimal boundary is a line or surface that separates the training data into regions corresponding to each category [42].
For the SVM model, the hyperparameters selected were cost and linear kernel. Through optimisation, we found that when the cost was set to 8.24, the SVM achieved an accuracy of 0.82, sensitivity of 0.66, specificity of 0.83, and AUC of 0.83. Additionally, after applying the PFI approach, the most important variables were employment type, employment status, job satisfaction, age, depression, marital satisfaction, experience using social welfare facilities, leisure activity satisfaction, satisfaction with the use of social welfare facilities, and severity of disability (Table 4).
Table 4. Permutation importance of SVM.

3.3. AdaBoost

AdaBoost, an ensemble learning technique, is a boosting method first designed to improve binary classification performance [48,54,55]. AdaBoost learns from weak decision trees, known as stumps, and combines them to form an ensemble [55]. Weak classifiers perform slightly better than random estimates, and AdaBoost amplifies their functionality to create a strong classifier [48]. AdaBoost can be applied to various classification problems [55], and although it is not specifically designed for datasets with class imbalance, it helps to address such problems by assigning different weights to each class [48].
The hyperparameters of AdaBoost were selected as N_estimators and learning rate. After optimisation, the results were as follows: N_estimators = 133, learning rate = 0.38, accuracy = 0.87, sensitivity = 0.62, specificity = 0.90, and AUC = 0.87. Additionally, after applying the PFI approach, the most important variables were age, average number of exercise days per week, smoking status, satisfaction with residential environment, area size, number of household members, severity of disability, satisfaction with welfare services, duration of disability, and leisure activity satisfaction (Table 5).
Table 5. Permutation importance of AdaBoost.

3.4. Bi-LSTM

LSTM has been proposed to address the shortcomings of RNNs [42,56] and solve the vanishing gradient problem of RNNs by injecting past information at later stages [42]. LSTM networks are designed to transfer important information across multiple future steps and typically include three gates: input, output, and forget [57]. Bi-LSTM enhances the performance of traditional LSTM by adding a backward memory block (or cell) to the LSTM memory block, thereby facilitating end-to-end learning in a deep learning model [58].
The hyperparameters for Bi-LSTM were batch size, hidden size, and iterations (Iter). After optimisation, the results were as follows: batch size = 128, hidden size = 5, iteration = 40, accuracy = 0.67, sensitivity = 0.90, specificity = 0.65, and AUC = 0.87. Additionally, the PFI approach revealed that the most important variables were age, job satisfaction, marital satisfaction, leisure activity satisfaction, severity of disability, disability-related daily life restrictions, satisfaction with welfare services, area size, number of meals a day, and satisfaction with the use of social welfare facilities(Table 6).
Table 6. Permutation importance of Bi-LSTM.

4. Discussion

This study developed an AI-based suicide ideation prediction model for people with disabilities to provide strategies for the proactive management of high-risk groups by national and medical institutions. In addition, this study selected three models (SVM, AdaBoost, and Bi-LSTM) based on different algorithms to compare their prediction results. Our results showed that the AUCs for the models were 0.83 (SVM), 0.87 (AdaBoost), and 0.87 (Bi-LSTM) for SVM, AdaBoost, and Bi-LSTM, respectively. Notably, all the AUC values exceeded 0.80, indicating excellent performance. Generally, an AUC > 0.80 is considered outstanding [59,60]. Considering the performance of each indicator, the model with the highest accuracy, which represents the accuracy of the results predicted by the model in the overall data, was AdaBoost (0.87). Bi-LSTM had the highest sensitivity (0.90), which represents the rate of correctly predicting suicide ideation. AdaBoost also had the highest specificity (0.90), which represents the percentage of those who did not have suicide ideation. If sensitivity, the rate of correctly identifying true positives, is the most important criterion, Bi-LSTM would be the most appropriate model. In addition, the prediction model developed in this study can be used to identify high-risk groups for suicide ideation and to develop a system that can intensively manage these groups. The variables influencing suicide ideation for disabled people, as presented in this study, can be used to establish suicide prevention policies for people with disabilities or treatment methods aimed at improving their mental health.
This study used the PFI approach to identify the importance of the variables for each model. The top ten variables common across all three models included age, severity of disability, and leisure activity satisfaction, which are closely related to physical functioning. In addition, the variables commonly included in the top ten in two of the three models were marital satisfaction, satisfaction with welfare services, area size, job satisfaction, and satisfaction with the use of social welfare facilities, which are closely related to daily life. This means that in the three algorithms, these variables commonly acted as important variables to predict the suicide ideation of people with disabilities. These findings differ from those of previous studies that identified depression and pessimistic self-perception as major risk factors for suicide or suicide ideation [61,62,63,64,65], likely because of differences in the study population or methodology.
Our review of prior studies revealed only four studies that included people with disabilities as part of their research population on AI-based suicide ideation prediction [59,61,66,67], highlighting the limitations of previous research in applying their findings to the disabled population. As people with disabilities have a higher suicide rate than the general population, the need to identify the difference between the disabled and the general population has become increasingly important. However, this distinction has not yet been achieved. Therefore, this study is the first to develop a model for predicting suicide ideation for people with disabilities. It presents the risk factors for suicide ideation among people with disabilities that are distinct from those of the general population, which can be used as basic data for establishing suicide prevention policies for people with disabilities.
Suicide is a global issue. Taiwan enacted the Suicide Prevention and Elimination Act in 2019 owing to the increasing suicide rate [68], while Germany revised the Children and Youth Support Act enacted in 1991 to the Child and Youth Reinforcement Act in 2021 [69]. In addition, the French Minister of Education announced mental health promotion measures in May 2023 to provide education to detect student vulnerability and urge all middle and high school students to include suicide prevention national phone numbers in their communication letters [69]. To date, countries have made efforts to lower suicide rates. In addition, to reduce the increasing suicide rate, Korea has revised its laws to prevent suicide and create a culture of respect for life, implementing suicide prevention projects that manage the use of carbon monoxide, pesticides, and bridges, which are used as suicide means [70]. Additionally, in April, the Korea Suicide Prevention Association issued an emergency statement calling for counter measures against suicide [71], which showed the urgency to develop innovative measures to reduce the suicide rate in Korea.
The current study used machine learning models, including SVM and AdaBoost, as well as the deep learning model Bi-LSTM, and analysed 19,141 data points from the 2018–2021 Disability and Life Dynamics Panel. Notably, machine and deep learning models generally perform better with larger datasets; thus, the 19,141 data points used in this study may be insufficient. Furthermore, this study selected 39 independent variables based on previous research [6,9,10,11,12,13,14,16,20,21,23,24,25]; however, there may be other important variables that can predict suicide ideation among people with disabilities. Therefore, future research on this topic may consider exploiting more variables, data, and additional models other than SVM, AdaBoost, and Bi-LSTM. In addition, because this study was conducted based on Korean disabled data, the results of this study may not be applicable to countries other than Korea. Furthermore, this study did not present the results based on the type of disability due to insufficient data for each type of disability. Therefore, future studies may consider presenting results based on the type of disability. This study utilised the results of a previous survey; hence, it is difficult to rule out the possibility that some information was omitted. Additionally, this study did not account for participants’ internal pain levels and whether they were receiving treatment for mental illness.

5. Conclusions

This study is the first to develop an AI-based suicide ideation prediction model for people with disabilities, using 19,141 data points from the ‘Disability and Life Dynamics Panel’ survey, which is representative of the disabled population in South Korea. The results showed that all three models (SVM, AdaBoost, and Bi-LSTM) achieved an AUC > 0.80. Additionally, this study employed the PFI approach to calculate the importance of variables in each model, identifying the factors influencing suicide ideation among people with disabilities that differ from those affecting the general population. The findings of this study provide significant evidence that can be used for the proactive management of high-risk groups and the formulation of suicide prevention policies for people with disabilities.

Funding

This research received no external funding.

Institutional Review Board Statement

Ethical approval was not required because this study utilised existing survey data. Therefore, an exemption from review was granted by the Korea University Institutional Review Board (KUIRB-2023-0413-01) on 1 December 2023.

Data Availability Statement

No new data were created or analyzed in this study.

Acknowledgments

This study is based on the 1st to 4th data of the Disability and Life Dynamics Panel of the Korea Disabled Development Institute.

Conflicts of Interest

The author declares no conflicts of interest.

References

  1. Mental Health, Brain Health and Substance Use (MSD). LIVE LIFE: An Implementation Guide for Suicide Prevention in Countries; World Health Organization: Geneva, Switzerland, 2021. [Google Scholar]
  2. UN Statistics. Global indicator framework for the sustainable development goals and targets of the 2030 agenda for sustainable development. In Developmental Science and Sustainable Development Goals for Children and Youth; Springer: Berlin/Heidelberg, Germany, 2019; p. 439. [Google Scholar]
  3. Min, K.A.; Jo, J.H. Safety Report 2023; Statistics Research Institute: Dae Jeon, Republic of Korea, 2024.
  4. Ministry of Health and Welfare. White Paper in Suicide Prevention 2023; Korea Foundation for Suicide Prevention: Seoul, Republic of Korea, 2023.
  5. Ministry of Health & Welfare. Available online: https://www.mohw.go.kr/board.es?mid=a10503000000&bid=0027&list_no=1481120&act=view (accessed on 18 October 2024).
  6. Huh, J.H. Factors Related to Suicidal Ideation in Physical Disabilities. Master’s Thesis, Yonsei University, Seoul, Republic of Korea, 2020. [Google Scholar]
  7. National Rehabilitation Center. Report on the Health Care Project for Persons with Disabilities; National Rehabilitation Center: Seoul, Republic of Korea, 2024.
  8. Korean Statistical Information Service. Available online: https://kosis.kr/index/index.do (accessed on 18 October 2024).
  9. Kim, G.H.; Shin, E.K. A Study on the Factors Influencing Suicidal Ideation among Middle-Aged and Elderly People with Disabilities. Ind. Promot. Res. 2022, 7, 43–58. [Google Scholar]
  10. Kim, Y.S.; Nam, Y.H. Analysis of factors related to suicidal ideation among people with and without disabilities in elderly living alone. J. Korean Soc. Sch. Community Health Educ. 2021, 22, 69–81. [Google Scholar] [CrossRef]
  11. Won, S.J.; Kim, H.M. Perceived discrimination and suicidal ideation of Korean adults with disability: Examining the mediating role of depressive symptoms. Korean J. Health Educ. Promot. 2019, 36, 65–76. [Google Scholar] [CrossRef]
  12. Hwang, S.H.; Im, W.G. Influence of depression and disabled Person’s acceptance of his/her condition on suicidal ideation of persons with disabilities due to accident. J. Rehabil. Res. 2012, 16, 244–268. [Google Scholar]
  13. Im, W.G.; Hwang, S.H. Investigation of the causes of suicidal ideation of persons with severe disabilities living in S-gu-Focusing on their daily activities, experience of committing suicide, and depression. Korean J. Stress Res. 2014, 22, 35–42. [Google Scholar] [CrossRef][Green Version]
  14. Kang, J.H.; Yoo, E.K. A study on multidimensional factors influencing suicidal ideation in old people with disabilities: Focused on Jeju special self-governing province. Soc. Sci. Rev. 2019, 35, 161–186. [Google Scholar] [CrossRef]
  15. Lee, Y.S.; Kim, H.S. Suicide ideation and suicide attempt among people with disabilities: The effects of household income and experienced discrimination. J. Disabil. Welf. 2016, 33, 5–34. [Google Scholar]
  16. Park, H.S.; Yang, H.T. The influence of disability acceptance of elderly with disability on suicidal ideation: Focusing on mediating effect of social participation. Disabil. Employ. 2015, 25, 229–250. [Google Scholar] [CrossRef]
  17. Jeong, J.S.; Lee, H.K. A study on the factors influencing on the suicidal ideation of the disabled elderly. Ment. Health Soc. Work 2016, 44, 34–63. [Google Scholar]
  18. Heo, S.Y.; Kim, H.M. Suicidal ideation of elderly with disabilities. J. Korean Data Anal. Soc. 2016, 18, 2135–2147. [Google Scholar]
  19. Yoon, M.S.; Kim, S.B. Longitudinal analysis of factors affecting the suicidal ideation of the elderly with disabilities: Focused on comparison between aging with disability and disability with aging. Ment. Health Soc. Work 2020, 48, 84–109. [Google Scholar] [CrossRef]
  20. Lee, E.M.; Hong, N.S.; Lee, S.J. Factors related to suicidal ideation of kidney patients. J. Health Inform. Stat. 2018, 43, 184–190. [Google Scholar] [CrossRef][Green Version]
  21. Song, S.Y.; Lee, Y.P. A study on the factors affecting suicidal ideation in people with mental disabilities. J. Korea Contents Assoc. 2021, 21, 765–775. [Google Scholar]
  22. Seo, S.Y.; Kim, H.S.; Kim, Y.T. Factors influencing suicidal ideation in people with mental disorder. J. Korean Acad. Community Health Nurs. 2013, 24, 245–254. [Google Scholar] [CrossRef]
  23. Lee, Y.M. A study about the factors affecting suicidal ideation of the elderly with stroke. Korean J. Care Manag. 2015, 17, 201–220. [Google Scholar]
  24. Lee, J.H. Relationship Between Social Discrimination and Suicidal Ideation of Middle Aged and Older People with Physical Disabilities: The Mediating Effects of Health-Related Quality of Life. Master’s Thesis, Yonsei University, Seoul, Republic of Korea, 2020. [Google Scholar]
  25. Nam, Y.H.; Han, S.G.; Moon, S.H. A study on the factors influencing the suicidal thoughts of people with spinal cord injury-centered on the effect parameters of the disability acceptance, social participation. J. Rehabil. Res. 2015, 19, 83–107. [Google Scholar] [CrossRef]
  26. Kim, C.J.; Koo, K.M. The effects of physical activities of disabled men with stroke on depression and suicidal ideation. Korean J. Phys. Educ. 2017, 56, 657–664. [Google Scholar] [CrossRef]
  27. Park, E.Y.; Kim, J.H. Factors related to suicidal ideation in stroke patients in South Korea. J. Ment. Health 2016, 25, 109–113. [Google Scholar] [CrossRef]
  28. Lee, J.H. Big data and suicide prevention. Glob. Soc. Sec. Rev. 2018, 4, 148–152. [Google Scholar]
  29. PR Newswire. “The Durkheim Project” Will Analyze Opt-In Data from Veterans′ Social Media and Mobile Content—Seeking Real-Time Predictive Analytics for Suicide Risk. Available online: https://www.prnewswire.com/news-releases/the-durkheim-project-will-analyze-opt-in-data-from-veterans-social-media-and-mobile-content----seeking-real-time-predictive-analytics-for-suicide-risk-213922041.html (accessed on 18 October 2024).
  30. Son, H.I. Utilization of PSYCKES(psychiatric services and clinical knowledge enhancement system) as suicide prevention tool in New York State. Glob. Soc. Sec. Rev. 2019, 10, 114–126. [Google Scholar]
  31. South-Central Behavioral Health Care Collaborative. Understanding PSYCKES: New York State’s Data System for Mental Health. Available online: https://scbhcc.org/understanding-psyckes-new-york-states-data-system-for-mental-health/ (accessed on 18 October 2024).
  32. Busan. Available online: https://www.busan.com/view/biz/view.php?code=2023022014212105375 (accessed on 18 October 2024).
  33. Newspim. Available online: https://www.newspim.com/news/view/20230511000925 (accessed on 18 October 2024).
  34. Korea Institute for the Disabled. User’s Guide for Disability and Life Dynamics Panel; Korea Institute for the Disabled: Seoul, Republic of Korea, 2023.
  35. Lim, S.H. Influencing factors on suicidal ideation of disabled persons—Using a panel survey on the life of the disabled. Crisisonomy 2022, 18, 123–134. [Google Scholar]
  36. Park, J.K. An Analysis of the Factors Influencing Life Satisfaction of the Employed and Unemployed Persons; Employment Development Institute: Gyeonggi-do, Republic of Korea, 2009. [Google Scholar]
  37. World Health Organization. Depressive Disorder (Depression). Available online: https://www.who.int/news-room/fact-sheets/detail/depression (accessed on 18 October 2024).
  38. Lee, J.Y.; Nam, S.K.; Lee, M.K.; Lee, J.H.; Lee, S.M. Rosenberg’ Self-Esteem Scale: Analysis of Item-Level Validity. Korean J. Couns. Psychother. 2009, 21, 173–189. [Google Scholar]
  39. Choi, J.H. A study on the family stress and coping strategy, family strengths among commuting couples. J. Fam. Better Life 2004, 22, 69–83. [Google Scholar]
  40. Yoo, Y.J. A study on the development of Korean family strengths Scale for strengthening the family. Fam. Relat. 2004, 9, 119–151. [Google Scholar]
  41. Korean Law Information Center. Available online: https://www.law.go.kr/lsSc.do?section=&menuId=1&subMenuId=15&tabMenuId=81&eventGubun=060101&query=%EC%82%AC%ED%9A%8C%EB%B3%B5%EC%A7%80%EC%82%AC%EC%97%85%EB%B2%95#undefined (accessed on 18 October 2024).
  42. Chollet, F. Deep Learning from the Founder of Keras; Park, H.S., Ed.; Gilbut: Seoul, Republic of Korea, 2018. [Google Scholar]
  43. Kim, E.J. Introduction to Artificial Intelligence, Machine Learning, and Deep Learning (Learning by Algorithm); Wiki Books: Gyeonggi-do, Republic of Korea, 2016. [Google Scholar]
  44. Ali, P.J.M.; Faraj, R.H.; Koya, E.; Ali, P.J.M.; Faraj, R.H. Data normalization and standardization: A technical report. Mach. Learn Tech. Rep. 2014, 1, 1–6. [Google Scholar]
  45. Micci-Barreca, D. A preprocessing scheme for high-cardinality categorical attributes in classification and prediction problems. SIGKDD Explor. Newsl. 2001, 3, 27–32. [Google Scholar] [CrossRef]
  46. Chawla, N.V.; Bowyer, K.W.; Hall, L.O.; Kegelmeyer, W.P. SMOTE: Synthetic minority over-sampling technique. J. Artif. Intell. Res. 2022, 16, 321–357. [Google Scholar] [CrossRef]
  47. Altmann, A.; Toloşi, L.; Sander, O.; Lengauer, T. Permutation importance: A corrected feature importance measure. Bioinformatics 2010, 26, 1340–1347. [Google Scholar] [CrossRef]
  48. de Giorgio, A.; Cola, G.; Wang, L. Systematic review of class imbalance problems in manufacturing. J. Manuf. Syst. 2023, 71, 620–644. [Google Scholar] [CrossRef]
  49. Lee, W.S.; Alchanatis, V.; Yang, C.; Hirafuji, M.; Moshou, D.; Li, C. Sensing technologies for precision specialty crop production. Comput. Electron. Agric. 2010, 74, 2–33. [Google Scholar] [CrossRef]
  50. Vapnik, V.N. The Nature of Statistical Learning Theory; Springer Science & Business Media: New York, NY, USA, 1995. [Google Scholar]
  51. Shammi, S.; Sohel, F.; Diepeveen, D.; Zander, S.; Jones, M.G.K. A survey of image-based computational learning techniques for frost detection in plants. Inf. Process Agric. 2023, 10, 164–191. [Google Scholar] [CrossRef]
  52. Cortes, C. Support-vector networks. Mach. Learn. 1995, 20, 273–297. [Google Scholar] [CrossRef]
  53. Cervantes, J.; Garcia-Lamont, F.; Rodríguez-Mazahua, L.; Lopez, A. A comprehensive survey on support vector machine classification: Applications, challenges and trends. Neurocomputing 2020, 408, 189–215. [Google Scholar] [CrossRef]
  54. Freund, Y.; Schapire, R.E. Experiments with a New Boosting Algorithm. In Proceedings of the 13th International Conference on Machine Learning, Bari, Italy, 3–6 July 1996; Volume 96, pp. 148–156. [Google Scholar]
  55. Khan, A.A.; Chaudhari, O.; Chandra, R. A review of ensemble learning and data augmentation models for class imbalanced problems: Combination, implementation and evaluation. Expert Syst. Appl. 2024, 244, 122778. [Google Scholar] [CrossRef]
  56. Hochreiter, S.; Schmidhuber, J. Long short-term memory. Neural Comput. 1997, 9, 1735–1780. [Google Scholar] [CrossRef]
  57. Subasi, A. Practical Machine Learning for Data Analysis Using Python; Mara Conner: Amsterdam, The Netherlands, 2020. [Google Scholar]
  58. Kim, K.J.; Lee, C.W. Prediction of music generation on time series using bi-LSTM model. Smart Media J. 2022, 11, 65–75. [Google Scholar]
  59. Na, K.S.; Geem, Z.W.; Cho, S.E. The development of a suicidal ideation predictive model for community-dwelling elderly aged >55 years. Neuropsychiatr. Dis. Treat. 2022, 18, 163–172. [Google Scholar] [CrossRef]
  60. Hosmer, D.W., Jr.; Lemeshow, S.; Sturdivant, R.X. Assessing the Fit of the Model, 3rd ed.; John Wiley and Sons: Hoboken, NJ, USA, 2013. [Google Scholar]
  61. Lee, J.Y.; Pak, T.Y. Machine learning prediction of suicidal ideation, planning, and attempt among Korean adults: A population-based study. SSM Popul. Health 2022, 19, 101231. [Google Scholar] [CrossRef]
  62. Bhar, S.; Ghahramanlou-Holloway, M.; Brown, G.; Beck, A.T. Self-esteem and suicide ideation in psychiatric outpatients. Suicide Life Threat Behav. 2008, 38, 511–516. [Google Scholar] [CrossRef]
  63. Ducasse, D.; Loas, G.; Dassa, D.; Gramaglia, C.; Zeppegno, P.; Guillaume, S.; Olié, E.; Courtet, P. Anhedonia is associated with suicidal ideation independently of depression: A meta-analysis. Depress. Anxiety 2018, 35, 382–392. [Google Scholar] [CrossRef]
  64. Fredriksen, K.J.; Schoeyen, H.K.; Johannessen, J.O.; Walby, F.A.; Davidson, L.; Schaufel, M.A. Psychotic depression and suicidal behavior. Psychiatry 2017, 80, 17–29. [Google Scholar] [CrossRef] [PubMed]
  65. Raschke, N.; Mohsenpour, A.; Aschentrup, L.; Fischer, F.; Wrona, K.J. Socioeconomic factors associated with suicidal behaviors in South Korea: Systematic review on the current state of evidence. BMC Public Health 2022, 22, 129. [Google Scholar] [CrossRef] [PubMed]
  66. Oh, B.; Yun, J.Y.; Yeo, E.C.; Kim, D.H.; Kim, J.; Cho, B.J. Prediction of suicidal ideation among Korean adults using machine learning: A cross-sectional study. Psychiatry Investig. 2020, 17, 331–340. [Google Scholar] [CrossRef] [PubMed]
  67. Cho, S.E.; Geem, Z.W.; Na, K.S. Development of a suicide prediction model for the elderly using health screening data. Int. J. Environ. Res. Public Health 2021, 18, 10150. [Google Scholar] [CrossRef]
  68. World Laws Information Center. Available online: https://world.moleg.go.kr/web/dta/lgslTrendReadPage.do?&CTS_SEQ=48326&AST_SEQ=300&ETC=9 (accessed on 18 October 2024).
  69. Han, M.J.; Kim, J.L.; Jung, J.D.; Yoon, S.J.; Kim, S.C. Issue Brief on Foreign Lwas; Korea Legislation Research Institute: Sejong, Republic of Korea, 2023; p. 6.
  70. Ministry of Health and Welfare. Available online: https://www.mohw.go.kr/board.es?mid=a10503000000&bid=0027&list_no=1480942&act=view (accessed on 18 October 2024).
  71. Korea Association for Suicide Prevention. Available online: http://suicideprevention.or.kr/04_sub/03_sub.html?no=206&bbs_cmd=view&page=1&key=&keyfield=&category=&bbs_code=Site_BBS_3 (accessed on 18 October 2024).
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Journal Statistics
Multiple requests from the same IP address are counted as one view.
Download PDF
Predicting Prognosis of Early-Stage Mycosis Fungoides with Utilization of Machine Learning
Previous
The Impact of Cannabidiol Treatment on Anxiety Disorders: A Systematic Review of Randomized Controlled Clinical Trials
Next