Technique for Optimization of Association Rule Mining by Utilizing Genetic Algorithm

Abstract

Background: Due to advancement in usage of Internet and pattern discovery from huge amount of data flowing through internet, personal information of an individual or organization can be traced. Hence, to protect the private information is becoming extremely crucial, which can be achieved through privacy preserving data mining. Objective: The main objective to preserve the privacy of data and to maintain the balance between privacy and accuracy by applying privacy preserving technique and optimization respectively. Methodology: The generation of class association rule is done by utilizing associative classification technique namely class based association due to its simplicity which serves the purpose of classifying the data. Furthermore, privacy of the data should be maintained and hence privacy preserved class association rules are produced by applying privacy preserved technique namely anonymization. Hence, optimization technique specifically genetic algorithm as well as neural network has been applied to maximize the accuracy. Results: (Four various real datasets has been utilized for different experimentation). : Implemented Classification Based on Association (CBA) algorithm of Associative Classification technique and it provides virtuous accuracy as compared to other techniques by setting the support as 0.2 and confidence at 0.6. : Privacy preserving techniques namely k-anonymization was implemented to preserve the privacy but it has been observed that as privacy (k-level) increases, accuracy (percentage) decreases due to data transformation. Conclusion: (Hence, optimization technique namely Genetic Algorithm (GA) and Neural Network (NN) has been implemented to increase the accuracy (probably 7-8%). Furthermore, on comparison of GA and NN considering the time parameter, GA outperforms well.