Fisher’s discriminate analysis (FDA) and logistic regression analysis are two commonly used approaches for binary classification problems. Here even though the responses are binary, the linear regression still can be used for binary classification, because for prediction purposes, the linear regression method can still provide good prediction accuracy and in fact, it is equivalent to that of FDA and logistic regression with less assumption. One computational advantage of this linear regression approach is that the estimated regression coefficient vector is easy updated by cooperating with matrix extension and partition matrix techniques. Thus based on active learning concept, the sequential procedures are proposed due to different updating criteria. Simulations are used to demonstrate the advantages of the proposed sequential procedures. Compare with the classification results obtained by the whole training set. The simulation results show that the proposed procedures can obtain the almost the same classification rates with much fewer training points. Hands-on session will be conducted in the class.