Handling Missing Values In Preterm Birth Dataset Using An Optimized Multinomial Naïve Bayes Model

Abstract

Missing values lower the quantity of information that machine learning (ML) methods learn during the training phase which harms classification accuracy. This research work presents an optimized multinomial Naïve Bayes (NB) classifier as a solution to this problem by handling the missing values. To lower the quantity of incorrect classification, it also suggests a feature selection and classification procedure. Our calculation of the research approach was implemented using the Kaggle dataset for maternal healthcare risk prediction. This research work compared the performance of different methods like decision tree (DT), and Random forest (RF).  The outcomes of our simulations define that an optimized multinomial NB classifier attained the maximum accuracy rate of 94%. The DT classifier has achieved 66% accuracy and RF values attained 69%. The research method gives a reliable outcome for predicting maternal healthcare risk at PTB by identifying the problem of missing values in the dataset. The outcomes define that utilizing an optimized multinomial NB classifier improved the performance of the classification models. This work gives a reliable involvement in the domain of maternal healthcare risk prediction in Preterm Birth (PTB) research and the best part of identifying the missing values in ML uses