Data Analytics & Prediction for Patients to visit clinic after booking appointments through Healthcare Data using MLOps

Abstract

Patient no-shows continue to drive up healthcare expenses, wreaking havoc on the healthcare system's day-to-day operations, lowering healthcare delivery efficacy, and limiting all patients' access to high-quality care. This study is used to identify the prediction of missed appointments by the patient after confirming the appointment with the clinic.

Patient Demographics such as Patient age, Patient gender, appointment length, and patients' socioeconomic status have the greatest impact on patient attendance at medical appointments. A patient's previous history of attending the appointment after booking, financial background, and patient appointment information are key factors in patient attendance. The 'Medical Appointment No-show' dataset was used to develop nine machine learning prediction models: Logistic Regression, XGBClassifier, KNN, DecisionTreeClassifier, Random Forest Classifier, Gradient Boosting Classifier, Ridge Classifier, Baggin Ridge Classifier, ExtraTreesClassifier.

The XGBClassifier was picked as the best performing model with a score of 80\%. When XGBClassifier has a Receiver Operating Characteristics score of 0.94, it performs better. It's in good condition. Other forms of research may be included in the future. critical aspects affecting patient attendance performance to improve the model

Roles and demands are also getting increasingly complicated as a result of the diverse technologies available for various operational phases. This research examines and comprehensively specifies ML Operations (MLOps) including different technologies and tools at various stages of the project pipeline and related responsibilities. In this study different model performance, Input data, different methods, and the model quality metrics are discussed in detail, with an emphasis on tool interoperability and comparison using carefully chosen MLOps criteria. This project also includes the integration of Jenkins and a version control system from the start of development to the deployment of the production server. Jenkins tasks are used to handle each phase, and a one-click solution module has been constructed for the entire project.