Table 1.

Relevant works on psychological risk analysis using predictive analytics.

References	Reviewed risks	Approach	Inference
Bauer et al. (36)	Bipolar disorder	Paper-based survey	47% elderly people utilized internet and 87% youths exhibit bipolar disorder.
Dhaka and Johari (37)	Mental disorder	Genetic algorithm and MongoDB tool	Storage and processing massive mental risks data on MongoDB database.
Kumar and Bala (38)	Depression	Sentimental analysis and save data on Hadoop	Preprocessing online social media perspective on specific business products.
Furnham (39)	Personality disorder	Hogan “dark side” measure (HDS) concept of dependent personality disorder (DPD)	Most personality risk factors are highly linked to a type of cooperative personality.
Bleidorn and Hopwood (40)	Personality assessment	Prediction models and K-fold validation	Focused on aspects such as organized adaptability and arguments to improve verification of predictive techniques.
Sarraf and Tofighi (41)	Alzheimer's risk	Convolutional neural network	Mental health instances were successfully categorized with 96.86% accuracy rate.
Fiscon et al. (42)	Brain disorders	Decision tree and EEG signals	Decision tree outperforms others in precise risk detection with 90% accuracy and 87% specificity with use of cross validation method.
Chatterjee et al. (4)	Anxiety analysis	Regression and bayesian classifiers	Used a probabilistic technique to validate patients with anxiety levels. It concluded that Bayesian Network showed the best accuracy of 73.33%.
Omurca and Ekinci (43)	Traumatic stress risks	Neural networks and social media optimization	A hybrid system to classify PTSD individuals and allowed feature selection methods to find vital metrics of patients' risks. The accuracy differed between 74 and 79%.
Dabek and Caban (7)	Mental risks	Neural network	Analyzed 89,840 samples and recorded a classification accuracy of a range (73%-95%).
Katsis et al. (44)	Anxiety disorders	Integrated meta classifiers	Proposed a hybrid model with mental health signals for assessing anxiety risks. Accuracy of 77.33, 80.83, and 78.5% was the output with neural network, radial networks, and SVM, respectively.
Saxe et al. (45)	Stress risks	SVM and Lasso regression	Optimal AUC value noted was 79 and 78% with SVM and RF, respectively.
Karstoft et al. (46)	Stress and depression	Hybrid method Feature selection and SVM	Target Information Equivalence Algorithm optimized detection of PTSD when used with support vector machine. The mean AUC was 0.75.