الفهرس 
Chapter 1Only 14 pages are availabe for public view
 |  | from | 95 |  |  |
| | | from | 95 | | |
Abstract
We are sure that most of us have spent innumerable hours in the hospital emergency room, waiting to be admitted or meet with a doctor. Thus, it led to a rise in sickness patients, which are widely altitude, through rising arrival rates in emergencies. So, Emergency Departments (EDs) in public hospitals still suffer from congestion worldwide. Anyone may unexpectedly require medical care at any time. Forecasting mortality rates for critically ill patients is necessary for emergency patronage. This will impact the division of emergency supplies. So expecting patient rates will reflect on the speed of patient recovery and minimize morbidity and mortality rates. So, emergency medical care must therefore be available 24 hour a day as an essential component of a healthcare delivery system.Emergency physicians (EPs( are first-contact providers who care for all patients. It is a major defy for them to quickly discover the patient’s deterioration and prohibit sudden dying via massive amounts of clinical data.Current work is aiming to analyze the patient data profiles at the EDs based on machine learning (ML) techniques and predict mortality rates. The rapid foretell of the patients dying through danger guarantees to reduce death through warranty effective and efficiency resources allotment and planning the therapy. This work highlights the importance of measuring EDs mortality rates in hospitals through ML techniques, which can help healthcare providers get a better planning system to test the effect of each parameter on the most susceptible group of patients and point out which DIAGNOSIS is a reason for additional deaths, reducing high-risk patients’ deaths while in EDs. The proposed model consists of six data mining (DM) phases: data extraction, data preparation, preprocessing, feature extraction, prediction, and test and evaluate. Features include the date of birth (DOB), ADMISSION_TYPE, AGE, GENDER, ADMITTIME, ED visits, hospitalizations, INSURANCE, and the patient’s DIAGNOSIS. We used the medical information mart for intensive care (MIMIC-III) dataset, which is a hospital clinical dataset that included 43809 patients for 56318 visits. We employed seven algorithms to establish the predictive models: fast large margin (FLM), random forest (RF), generalized linear model (GLM), deep learning (DL), support vector machine (SVM), naïve Bayes (NB), and decision tree (DT). DL gave us the best accuracy (ACC) of 86.3%. The run times varied for all techniques, which showed that the best technique was DL in 2 minutes and 4 seconds due to a higher ACC, area under the curve (AUC), and F-measure. As a result, ML performance models are evaluated to determine patient riskiness and severity. Results revealed 64% survival and 36% death in the hospital. Total number of patients DIAGNOSIS were 13245. 19612 patints fallen at the top 10 DIAGNOSIS that represented 44.7% of the total patient number. Total death were 7319 in the top 10 DIAGNOSIS, which forms 45.2% from the whole death. Death rates affected in: M child patients (1-19 years), 44% in DIAGNOSIS (NEWBORN) and 322 M NEWBORN patients, 33.8% of CHILD patient’s in DIAGNOSIS (SEPSIS), 34% in DIAGNOSIS (INTRACRANIAL HEMORRHAGE) and 32% in DIAGNOSIS (GASTROINTESTINAL BLEED). The proposed model affirmed that DL techniques and prediction models are accurate tools to help healthcare providers plan better and proper healthcare.