Virtual Learning Environment (VLE) such as Moodle, Blackboard, and WebCT are commonly and successfully used in E-education. While they focus on supporting educators in creating and holding online courses, they typically do not consider the individual differences of learners. However, learners have different needs and characteristics such as prior knowledge, motivation, cognitive traits, and learning styles. Recently, increasing attention is paid to characteristics such as learning styles, their impact on learning, and how these individual characteristics can be supported by learning systems. These investigations are motivated by educational theories, which argue that providing courses and contents which fit the individual characteristics of students makes learning easier for them and thus their learning progress. This research primarily focuses on providing adaptation to VLEs by inferring learning styles according to the Felder-Silverman Learning Style Model (FSLSM). An automated data-driven approach for identifying learning styles from behavior and actions of learners has been designed, implemented, and evaluated, demonstrating that the approach is suitable for identifying learning styles. Based from this approach, an Adaptive Virtual Learning Environment prototype for automatic classification of learning styles in VLEs had been implemented. This approach was experimented on five hundred seven (507) students of Computer Programming 1 Course created using Moodle. Student's behaviors have been extracted from log data and the learning style for each student was mapped according to FSLSM. Classification accuracy and kappa statistics have been observed to measure the performance of each classifier. The results show that the efficiency of classification by means of J48 decision tree technique had the highest average value of correctly classified instances at 87.42% accuracy and it could be used to infer the learning style of students in an Adaptive VLE.