Evolutionary computation (EC) has been used as a powerful approach to feature selection. However, many existing EC-based feature selection methods tend to solve one problem at a time. Ignoring the fact that many feature selection tasks are related. The knowledge learned from solving one feature selection task can be used for solving its related problems. Evolutionary transfer learning, which aims to improve learning performance by transferring knowledge between related problems, has gained more and more attention. However, not much work has been done for solving related feature selection tasks by evolutionary transfer learning. This work proposes to fill the gap by developing new evolutionary-transfer-learning-based methods for feature selection in classification, which aims at improving the performance of feature selection. This will be achieved by developing novel methods for transferring useful solutions between related problems. This work also tends to develop new mechanisms for transferring knowledge of the generated classifiers between related problems, which are expected to make classification more effective and efficient.