Abstract: Taking classification learning as an optimization problem of seeking for optimal classification rules， an adaptive ant colony algorithm， adaptive L_AMP， is proposed to solve classification problem and get the comprehensible classification rules. In the rulebased classification methods， the way of choosing of a ruleevaluating function is important. The adaptive L_AMP proposed can be used to automatically select an appropriate ruleevaluating function according to the data set， thus improving the classification correctness. Moreover， a local search technique is introduced into the algorithm proposed. The algorithm is run on the real data set and compared with other relevant algorithms. The results show the superiority of the algorithm proposed.

Key words: ant colony algorithm, adaptive ant colony algorithm, classification, rule evaluation function