In the digital era, choosing the right educational game is a challenge for parents, so a recommendation system is needed that is able to provide relevant suggestions according to children's needs. This research aims to build an educational game recommendation system using the Item-Based Collaborative Filtering method. This method works by analyzing user rating patterns for games, then calculating the level of similarity between games using the Adjusted Cosine Similarity and Weighted Sum algorithms to produce personalized recommendations. Data is obtained explicitly through user interaction in the form of likes and comments on available games. System testing was carried out involving 22 respondents. To the question "Do the recommended educational games help increase your child's knowledge or skills?", 54.5% of respondents answered "Strongly Agree" and 45.5% "Agree", with no negative responses. This shows that all respondents considered the recommended educational games to be positively beneficial for children's development. Meanwhile, to the question "Do the game recommendations suit your child's wishes?", 50% of respondents answered "Strongly Agree", 40.9% "Agree", and 9.1% "Neutral". These results indicate that the majority of respondents considered the system to be quite appropriate in adapting recommendations to children's interests.