<p>This study aims to create a model using machine learning algorithms to determine teacher candidates’ misconceptions about electricity, to determine the algorithm with the highest success rate, and to realize students’ conceptual changes about electricity with activities prepared based on differentiated teaching in a laboratory environment. A design-based research method was used in the study. The sample consists of thirty science teacher candidates continuing their education in the third grade. Activities prepared for differentiated teaching applications were developed using the ADDIE design cycle. The obtained artificial intelligence data were created using the Learning Analytics cycle and analyzed by the thematic analysis method. Data were analyzed using frequency distribution and dependent <i>t</i>-test. It was determined from the data that the number of sentences containing teacher candidates’ misconceptions about electricity was 249. It can be said that the detected misconceptions are related to the literature. Multilayer Perceptron, Support Vector Machine, Gradient Boosting, Decision Tree, Random Forest, Logistic Regression, k-Nearest Neighbor, and Naive Bayes Machine Learning algorithms were used in the research. As a result of the thematic analysis, it was determined that the algorithm with the highest success rate was the Ensemble model developed by the researchers. Finally, it was observed that there was a statistically significant difference in the success score of the teacher candidates in favor of the post-test in the effect of differentiated instruction. It was concluded that differentiated instruction was efficient in the conceptual change levels of the teacher candidates. In addition, the fact that the success level of misconception detection with artificial intelligence in the study was notably high suggests that it will contribute to the relevant literature.</p>