Frequent itemset generation using enhanced Fuzzy Apriori Algorithm for web log session data
D. Kavitha, B. Kalpana · 2017
Due to the enormous usage of web data from the web servers, there is a need to have important information about website visitors and their web usage behaviors. The collected information is used to enhance the effectiveness of websites. Some major tasks are used in the information collection process, which is identification of a session and frequent itemset generation. In order to identify dynamic web log Sessions, an Online Incremental Learning (OIL) method is used and it is named as (DS-OILSD), because of its automatic selection of threshold based on standard deviation. DS-OILSD provides better results for session identification when compared to timeout methods. Once the web log session identification is done, searching for frequent patterns helps to optimize the web site structure and enhances the performance of web servers. If a web log session has many frequent itemset, then the time for generating association rules is much longer than the time for mining frequent itemset. The challenging task lies in the selection of min support and min confidence value. These problems are reduced by Enhanced Fuzzy Apriori Algorithm (EFAA) which is a combination of itemset generation and Rule Pruning technique. The proposed EFAA method comprise three major steps: (1) Generating frequent itemset from web log session, (2) Pruning itemset and (3) Generating association rules from itemset. Fuzzy Intersection Pruning (FIP) is proposed to prune frequent itemset. In FIP, intersection operation is performed between two frequent itemset. Fuzzy Automated Support (FAS) count value is used to get the optimal minimum support and generates frequent itemset. This Fuzzy Automated Support Confidence Pruning (FASCP) technique arranges the rules in ascending order with support and confidence thresholds. FASCP technique assumes that all rules having high support and confidence values will have quality information. The experimental results shows that, the EFAA method is effective for frequent itemset generation when compared to apriori and modified apriori algorithm in terms of precision, recall, accuracy and time.