Penerapan Model Deep Learning Terhadap Pendidikan Seks Anak Usia Dini

Abstract

Sexual violence against young children in Indonesia remains a serious concern, highlighting the need to strengthen sex education as an early preventive measure. This study aimed to describe the level of sex education understanding among children aged 5–6 years at Cemara Liliba Kindergarten and to apply an Artificial Neural Network (ANN) based on a Multilayer Perceptron (MLP) as a deep learning approach to classify children's levels of understanding based on observational data. A quantitative descriptive design was employed involving 20 children selected through total sampling. Data were collected using an observation instrument consisting of 10 basic sex education indicators rated on a 1–4 scale. Data analysis was conducted using descriptive statistics and an ANN model validated through Leave-One-Out Cross Validation (LOOCV). The findings revealed a mean understanding score of 33.35 out of a maximum score of 40 (83.38%), with 14 children (70%) categorized as having a high level of understanding and 6 children (30%) categorized as having a low level. The ANN model accurately classified the research data, while the indicators related to protecting private body parts from strangers and identifying private body parts contributed the most to the classification results. The research instrument demonstrated excellent reliability (Cronbach's Alpha = 0.877). The novelty of this study lies in applying ANN as a data analysis method for early childhood sex education, offering a more objective and data-driven approach to evaluating children's understanding.