Advanced libraries, E-business representatives and comparative incomprehensible data arranged frameworks depend on reliable information to offer top notch administrations. In any case vicinity of copies, semi imitations, or close copy entrances (Dirty Data) in their archives maligns their capacity assets straightforwardly and conveyance issues by implication. Critical speculations in this field from invested individuals incited the requirement for best strategies for expelling imitations from information vaults. Former methodologies included utilizing SVM classifiers, or Genetic Programming (GP) methodologies to handle these grimy information. Despite the fact that execution savvy GP frameworks are superior to Svm's, both methodologies endured with handling overheads that obliges a pretraining to really execute Deduplication process. So propose to utilize Active Learning Genetic Programming Mechanism a question ward record matching strategy that obliges semi administered information set. AGP utilizes a dynamic learning approach as a part of which a council of multi characteristic capacities votes in favor of characterizing record matches as copies or not. Results demonstrates that AGP ensures the nature of record deduplication while decreasing the amount of marked samples was required.