Abstract: 【Purpose/significance】The assessment readability of text plays an important role in measuring the difficulty of reading ma⁃ terials. The development of Chinese text readability assessment method and the use of this method to evaluate children′s reading mate⁃ rials can screen out suitable reading materials for different readers, so as to effectively improve reading ability.【Method/process】This paper uses primary school Chinese textbooks as database. Ten classifying models are built based on four traditional machine learning algorithms (linear regression, support vector machine, decision tree classifier, K-nearest neighbor), five ensemble learning algorithms (random forest classifier, extremely random tree, AdaBoost, Bagging and XGBoost) and one artificial neural network.【Result/conclu⁃ sion】It is found that the models based on random forest, Bagging and XGBoost with better performence than other models, and their maximum accuracy and F1 value of cross-validation are more than 0.75. In particular, the text readability classifier based on random forest classification model shows excellent performance in the readability classification prediction of primary school Chinese teaching materials, with the maximum accuracy and F1 value of cross-validation exceeding 0.76【. Innovation/limitation】This study provides an effective tool for the assessment of the readability of text and material screening of Chinese children′s books. It is planned to collect more Chinese text data in the future, and combine with more advanced deep machine learning algorithms to further improve the accu⁃ racy and scope of application of the Chinese text readability classifier.