Classification of Contraceptive Use Among Undergraduate Students Using a Supervised Machine Learning Technique

The Kenyan government in partnership with other stakeholders involved in providing family planning services have put in place various strategies and policies to increase uptake of contraceptives. This results in an increase in contraceptive prevalence rate (CPR), reduction of both total fertility rate (TFR) and sexually transmitted infections (STIs). Despite the various strategies and policies, the total fertility rate still remains high, while CPR has been unattained, respectively. The aim of this study was to classify contraceptives use among undergraduate students using a supervised machine learning technique. The target population constituted students at Jomo Kenyatta University of Agriculture and Technology (JKUAT) (Eldoret Campus). The study applied simple random sampling technique to obtain data from a sample of 252 using structured questionnaires. A decision tree classifier based on CHAID and C5.0 algorithms were used for classification. Pearson Chi-Squared statistic was used as feature selection technique to rank significant factors influencing contraceptives use based on their Chi scores. The findings show that the use of Chi-Squared feature selection led to contraceptives factors that were ranked higher having higher classification performance. The fitted decision tree model based on CHAID algorithm had a higher classification accuracy of 64.68% with 195 correct classifications as compared to the C5.0 decision tree model with accuracy of 61.18% with 163 correct classifications. The study findings contribute to a better insight on the classifications of contraceptives use among undergraduate students in Kenya. Hence, the government of Kenya can implement policies to enhance contraceptives awareness.