supervised machine learning

The main purpose of machine learning is a complex process that is carried out by determining the model and training it using a large volume of data. In the past, the main focus in this field was more on improving the structures of models and algorithms, but recently more emphasis has been placed on the quality and quantity of data. This article aims to provide an overview of the problems in data collection and offer a solution for them.

In this study, the challenges faced by researchers in collecting data and evaluating supervised machine-learning models were examined through a review method. Documentation from PubMed, Scopus, Science Direct databases, and Google Scholar search engine from 2001 to 2023 was retrieved. After screening, a total of 17 full articles were reviewed and included in the study.

The findings indicate that researchers in supervised machine learning studies face four challenges in data collection, which are: insufficient number of samples, unrepresentative training data, poor data quality, and irrelevant features, and in model evaluation, they face four challenges: overfitting, lack of generalizability, lack of sufficient data for validation, and mismatched data.

Increasing the sample size, utilizing a random selection algorithm, data cleansing, using the correct statistical test, feature selection, feature extraction, using a simpler model, the K-fold technique, and data processing are among the factors that contribute to achieving a model with better performance.

According to the report released by world health organization (WHO), the ST-segment elevation myocardial infarction- cardiogenic shock (STEMI-CS) is one of the important factors in patient mortality within hospitals (1), (2), (3), (4). CS and its related complications need a huge financial and medical burden. Some researchers stated that high mortality and complication rates of STEMI-CS patients are associated with the lack of effective early preventive treatments. Given the risk of CS and the different risk factors associated with it, accurate clinical risk prediction tools need to be developed to accurately predict the onset of CS. Recently, researchers have been used various machine learning methods to predict the risk of mortality in STEMI-CS patients. Recently, machine learning (ML) methods were developed to establish predictive models to identify the in-hospital mortality risk of STEMI-CS patients. The existing methods achieved encouraging results; however, their performance is not ideal, and more effort is needed to improve the performance. The aim of this study is to present a hybrid machine learning method for predicting the risk of mortality in STEMI-CS patients. Our proposed method combines a powerful swarm intelligence strategy, anti-coronavirus optimization algorithm (ACVO) with support vector machine (SVM) in risk prediction phase. The proposed model is compared with standard support vector machine (SVM), least absolute shrinkage and selection operator (LASSO), and adaptive neuro fuzzy inference system (ANFIS) on a real-world benchmark dataset.

To predict the mortality status of STEMI-CS patients, we proposed the ACVO-SVM algorithm. The proposed method is a hybrid machine learning algorithm that combines the SVM with ACVO algorithm to identify the most effective parameters on the death of patients. The incentive mechanism of using ACVO is to optimally configure the parameters of SVM to improve its prediction performance. The proposed ACO-SVM is also useful in determining the optimal subset of features and treatment strategies that have the greatest impact in predicting the status of STEMI-CS patients. The proposed approach models the problem of predicting the status of patients as an optimization problem. In order to determine the most effective features in predicting the survival or death of STEMI-CS patients, the proposed ACVO-SVM model is trained with different combinations of patient characteristics and adopted treatment strategies. Then the best combination of features that provides the highest performance is considered as the superior combination. To select the most effective features, first all the features are considered for training the SVM model, then the remaining features are ignored one by one and the model with the same structure is trained. The models were compared based on accuracy, recall rate, F1 criterion. Finally, the best model is used to predict the status of patients in test dataset. The data set used to evaluate the proposed method includes 410 records of patients hospitalized due to STEMI-CS complications in Shahid Madani Hospital of Tabriz University of Medical Sciences. The collected data is related to a 10-year period from 2009 to 2018. This data set includes five categories of main characteristics, which are demographic characteristics, type of myocardial infarction, risk factors, clinical symptoms, and type of treatment used. It should be noted that 80% of the records of the data set are considered as training data, and 20% of the records are considered as the test data set. The proposed method is implemented in MATLAB software.

Among M1 to M5 feature combination models, the experimental results show that the M1 model has higher performance on the training and test dataset in terms of predicting the patient's condition compared to other combination models. Model M1 includes the combination of characteristics of age, sex, type of myocardial infarction, smoking, percutaneous vascular interventions and coronary artery bypass surgery. This shows that considering the mentioned features has the greatest effect on the final condition of STEMI-CS patients. The results are in line with previous studies (2), (3) in this field, which stated that age, gender, smoking, coronary artery bypass surgery and percutaneous vascular interventions have the greatest effect on the mortality rate of patients. The M2 model ranks second in terms of efficiency in determining the status of patients, which shows that smoking also has a greater effect on the mortality of patients with STEMI-CS. Also, the M3 model indicates that the use of the balloon pump treatment strategy, along with other demographic symptoms of the patient, history of heart infarction and smoking have a great effect on the mortality rate of patients. In summary, it can be concluded that the demographic characteristics of the patient such as age and gender, smoking, history of illness and the use of coronary bypass surgery and percutaneous vascular interventions have a great impact on the mortality of STEMI-CS patients. The proposed ACVO-SVM approach is compared with several other popular approaches, which include: standard SVM model, LASSO regression, ANFIS, and XGBoost. The experimental results justify that the proposed ACVO-SVM outperformed its counterparts.

In this study, a hybrid supervised machine learning model was presented to determine the status of patients with cardiogenic shock due to ST-segment elevation myocardial infarction. The proposed ACVO-SVM model uses an ACVO optimization algorithm to estimate the optimal parameters of the SVM model, making the SVM training process more efficient. The proposed model was evaluated using a dataset of patients with cardiogenic shock and the results were compared with the LASSO, ANFIS, XGBoost, and SVM models. The results showed that the proposed method worked well compared to other proposed classification models. We also found that age, gender, type of myocardial infarction, smoking, percutaneous vascular surgery, and coronary bypass transplantation surgery are the most effective factors for survival in STEMI-CS patients. In this research, the models were evaluated on a small dataset. Therefore, one of the necessary tasks to improve this research is to evaluate the proposed method and other counterpart models on large datasets to determine their strengths and weaknesses. Another limitation of this research is the lack of examination of all factors affecting the survival of STEMI-CS patients, such as blood sugar level and duration of ischemia. For this reason, it is necessary to investigate all factors affecting the mortality of STEMI-CS patients to improve the quality of classification and prediction of the final status of patients.