Implementasi Regresi Logistik dengan Machine Learning pada Kasus Titanic Survivors Problem

Abstract

The sinking of the Titanic is one of the most famous events in history. Due to an insufficient number of lifeboats for all the people on board, this resulted in 1,502 deaths out of the total 2,224 passengers and crew members. This research aims to create a mathematical model to predict passenger survival in the event of a Titanic shipwreck. Additionally, the study analyzes factors that significantly influence the passengers' safety on the Titanic. The data used is obtained from secondary sources, taken from the official Kaggle website (https://www.kaggle.com). The relevant dataset for this study is the Titanic dataset, containing information about passengers who survived and those who didn't during the Titanic's accident on April 15, 1912. Kaggle is a platform that facilitates competitions to build the best analytical and predictive models for specific datasets. The platform is also often utilized by companies as a means to recruit professionals in the field of data science or machine learning engineering through recruitment-focused competitions. The research utilizes a sample consisting of two parts: the training data comprising 891 passenger entries and the testing data comprising 418 passenger entries. The method employed in this research is binary logistic regression. The focus factors of the study include ticket class (Pclass), gender (Sex), age (Age), number of siblings/spouses (SibSp), number of parents/children (Parch), ticket number (Ticket), fare (Fare), cabin (Cabin), and embarkation port (Embarked). Out of these nine variables, four were found to significantly impact the developed model. Through appropriate analysis, the successfully constructed model follows the function ���(���)=0,6944−0,8229���12−1,9999���13+2,7202���22+2,0986���31−1,1144���34−2,4317���35+0,6271���91. The classification accuracy rate of this model reaches 80%, with a recall rate of 60.27% and a precision rate of 83.01%.