Topic Modeling Using Latent Dirichlet Allocation Method Based On Child Anecdotal Record Data

Kindergarten education has a very important role in shaping the foundation of child development. For parents, see reports on the achievement of children’s development at school through the results of each semester’s assessment. The process of preparing teacher assessments uses 4 types of assessments, one of which is anecdotal record assessment. However, teachers experience problems in analyzing the topics in children’s anecdotal notes so that the results become less clear. In addition, it is still done manually so it takes a long time to analyze. Assessment of children’s anecdotal notes is an assessment in the form of a record of events that occur to children while at school. Anecdotal notes are in the form of text data, therefore this research aims to find out the topics in children’s anecdotal notes data by utilizing topic modeling using the Latent Dirichlet Allocation (LDA) method. The children’s anecdotal notes data is used as one dataset with a total of 276 collected directly from YBPK Sidorejo Kindergarten, after which data preprocessing is carried out in the form of case folding, remove punctuation, tokenizing, stopwords, and stemming. In the LDA topic modeling research using the Gensim library, 5 experiments were carried out by determining the number of different iterations, and displaying a range of 2-10 topics for each experiment. Analysis and evaluation using coherence score to determine the optimal number of topics by looking at the highest coherence score value in the experiment conducted. The analysis results in 3 topics with a coherence score of 0.4704 and 40 iterations or passes.