With the rapid development of Internet technology, network information not only brings many conveniences to the majority of users but also brings about the problem of information overload. It is increasingly difficult for users to accurately obtain the information they need from the vast ocean of information. This problem has led to the research of recommendation technology, especially the simulation path of online microvideo personalized recommendations. The video recommendation model in this study adopts the idea of an ant colony algorithm. The evaluation data and browsing record data generated by users are the basis of video recommendation, and the microvideo feature model and user preference model are abstracted from the original data by improving the ant colony algorithm. The recommendation algorithm is then used to recommend online microvideos that meet users’ preferences. There are two recommendation methods, one is to find similar students through the user preference model of each student, and find the microvideo content that the students have not learned; the other is to use the user preference model and the content feature model to match the content that conforms to the user’s preference. Compared with the traditional BP neural network algorithm, GA-BP algorithm, and PSO-BP algorithm, the mean square error MSE value of the ACO algorithm for microvideo scoring classification is reduced by 11.41%, 5.93%, and 2.41%, respectively. It can be seen that the classification of video ratings by the improved ACO algorithm has lower average errors than other algorithms. The network microvideo personalized recommendation scheme designed in this study has very practical significance.