Machine learning algorithms for improving the health care of seafarers through medical text classification and predicting the onsite occurrence of diseases

An employee of a seafaring company is a person who is employed to work aboard a marine vessel. The term usually refers to active seafarers, but it can also describe a person who has served in the profession for a long period of time. It is not uncommon for seafarers to lead a far from idyllic life. For up to nine months at a time, many international seafarers are away from their families, friends, and loved ones. Others choose this career so that their families can live a better life at home. Sacrificing time away from them is well worth it. It is common for seafarers to work in an environment that is physically and psychologically dangerous. In contrast to physical health issues, the mental health status of seafarers has not received much attention. There is a lot of danger and hardship on board ships, particularly during the cold winter months when seafarers are more exposed to extreme weather conditions. Despite improvements in living conditions, cabins are still functional rather than spacious and vary in quality and comfort. Although seafarers have their own national food, sometimes they are not provided with it due to long working hours and long contracts. Seafarers sometimes face significant dangers in addition to a challenging environment. In key laces around the world, these occur because of extreme weather conditions and privacy concerns. It is commonly believed that piracy is a long-forgotten life path. Seafarers can face extreme dangers because of piracy at sea, despite this being a far cry from the truth. Work environments for seafarers are often physically and psychologically hazardous. They often suffer mental breakdowns because of this frightening experience. Heat, cold, noise, friction, multiculturalism, loneliness, piracy, and criminalization are some risk factors for health conditions. Seafaring is characterized by means of a unique set of functions which units it aside from other occupations. This includes demanding bodily working situations, probably unsafe tasks, long hours of labor and high levels of stress and fatigue. Seafaring is likewise described as a ‘lonely lifestyles’. Not only are seafarers away from circle of relatives and friends for terribly lengthy periods of time, many seafarers stay remoted lives whilst onboard. An increasing diploma of automation onboard ships has brought about smaller crews and team contributors can also have very distinctive cultural and ethnic backgrounds. In this context, the personnel may be seemed as in particular susceptible to intellectual sick fitness. Generally, seafarers face a higher risk of illnesses and accidents than land workers. When this occurs, onshore doctors may be able to provide medical advice through telemedicine by receiving better symptomatic and clinical details in the health abstracts of seafarers. The adoption of text mining techniques can assist in extracting diagnostic information from clinical texts. The motive behind the works presented in this thesis is to propose sophisticated machine-learning models for identifying major health problems among seafarers. For that, we have adopted different text-mining experiments combined with machine learning models. This thesis work is composed of three research works that majorly highlight the incorporation of different text mining concepts in medical text documents of seafarers. This was due to the lack of experimental evaluations using computational techniques. To classify diseases and their associated symptoms, regression and supervised algorithms were applied to analyze these text documents. Medical text data of seafarers were examined from 2006 to 2021 and data of seafarers who got telemedical assistance through the International Radio Medical Centre (C.I.R.M.). The Centre establishes digital medical files for each case after it makes contact with the ship and updates them and this study analysed these files. In the first experiment, we adopted both lexicon and Naïve Bayes’ algorithms was done to perform sentimental analysis, and experiments were conducted over R statistical tool. Visualization of symptomatic information was done through word clouds and 96% of the correlation between medical problems and diagnosis outcome has been achieved. We validate the sentiment analysis with more than 80% accuracy and precision. The major drawback is the low sample size of medical texts. To overcome this, to classify diseases and their associated symptoms, the LASSO regression algorithm is applied to analyze the 15-year medical text documents. We applied lexicon sentimental analysis to explore the automatic labeling of positive and negative healthcare terms to seafarers’ text healthcare documents. This was due to the lack of experimental evaluations using computational techniques. In order to classify diseases and their associated symptoms, the LASSO regression algorithm is applied to analyze these text documents. A visualization of symptomatic data frequency for each disease can be achieved by analyzing TF-IDF values. The proposed approach allows for the classification of text documents with 93.8% accuracy by using a machine learning model called LASSO regression. It is possible to classify text documents effectively with tidy text mining libraries. In addition to delivering health assistance, this method can be used to classify diseases and establish health observatories. Knowledge developed in the present work will be applied to establish an Epidemiological Observatory of Seafarers’ Pathologies and Injuries. This Observatory will be a collaborative initiative of the Italian Ministry of Health, University of Camerino, and International Radio Medical Centre (C.I.R.M.), the Italian TMAS. Finally, another experiment involved a dataset of Enhanced Pre-Employment Medical Examination (EPEM) developed by the international container company CMA CGM. The N-gram feature technique was applied to these reports for a sequence of diseases and to predict the upcoming possibilities. Four ML models have been used in coupling with N-gram features. Results mentioned that, among the three trained disease classes, the overall classification accuracy was 78%. For arterial hypertension, we achieved the maximum precision value of 1, followed by 0.8 for metabolic imbalance and 0.76 for type 2 diabetes. Similarly, the recall value of type 2 diabetes achieved is 0.82 and is followed by metabolic imbalance (0.76), and arterial hypertension (0.5). The works presented in this report can able to significantly assist medical researchers in evaluating their experience with advanced ML modeling techniques among special worker types like seafarers. With tidy text mining packages it is proven that they can effectively classify text documents in healthcare analysis projects. As well as delivering medical assistance, this approach may be used to develop health observatories and to classify diseases.