The development of e-learning platforms is fast growing and has greatly increased the availability of education in many parts of the world. Nevertheless, the barriers to access and diversity of learning difficulties persist for learners with disabilities, the low economic status and cultural differences. The purpose of this research is twofold: to use machine learning to create a classification model that estimates the learner’s accessibility needs and inclusiveness in e-learning to enhance adaptability. consequently, six programs/balanced algorithms of classification; namely logistic regression, decision tree, random forest, SVM, k-NN, and gradient boosting were tested for the classification of the learner profiles by making use of a dataset of 568 records which incorporates different features regarding the learners including their demographics, their interaction with the learning tools, and their learning engagement. The models were evaluated using precision recall F1 score and accuracy measurements. Decision Tree classifier had the best average with an F1-score of 0.093 and an accuracy of 9.94%, and the Random Forest had the highest precision score of 13.6% which defined that it offered less number of false positive results. Even with these general performance figures, key findings point to the need to fine-tune the dataset and apply more complex pre-processing algorithms to enhance predictive ability. The proposed system can also serve to enhance the modification of e-learning platforms hence improving the delivery of the sessions to meet the needs of every learner. They enhance the progress of digital learning in the sense that the proposed solution offers a basic reference model for the consideration of accessibility issues in e-learning systems.