Abstract: Readability has been a hotly debated topic of study for years. A realistic future route for readability categorization and rating has been made possible by the current explosion in data-driven learning algorithms. In today's comprehensive environment, the investigation of textual readability is a well-established topic that has increased in importance. The challenge of assessing readability for the English language is covered in this essay. The goal is to forecast a sentence's readability based on the provided phrases, which conforms to the intended audience's expected comprehension ability. This readability factor is essential to the writing and comprehension phases of learning English. Current studies aim to improve the accuracy of classifiers by integrating ensemble learning with a range of machine learning (ML) models. This work introduces a stacked ensemble for assessing English readability, employing four classifiers as fundamental classifiers, including k-nearest neighbor, support vector machine, stochastic gradient descent, logistic regression, and linear discriminant analysis (LDA) as a meta classifier. In the present investigation, we conducted tests with twenty-five thousand phrases in English. They were also granted Flesch-Kincaid labeled into 7 distinct comprehension categories. The suggested model's successful operation was examined using several prognosis assessment indicators, comprising precision, accuracy, recall, F1 score, and area under the curve. The findings showed that the stacked model outperformed traditional ML models in terms of performance with a 98.66% accuracy rate.

Key words: artificial Intelligence, hybrid ensemble, machine learning, readability, stacking model