Lalit Mohan Goyal

Spatial data analysis using association rule mining in distributed environments: a privacy prospect

In this paper, we investigate the privacy issue that is remained in the spatial association rule mining (SARM). The main aim of SARM is to calculate the relationship among attributes according to their geographic locations. However, the major problem of the distributed data mining is the privacy and security issues, which require executing global results without disclosing private information by third parties. The problem of privacy preservation for SARM in distributed environments is controlled by the proposed algorithm, which is able to extract association among different numbers of attributes in geo-graphical distributed database with high privacy. The proposed algorithm is validated in term of data utility rate, efficiency and privacy preservation against the existing algorithms. It has been revealed that this algorithm decreases the execution time, memory requirements, and privacy failure rate when the size of database increases within the geographically distributed database environment.

Improved FP-Linked List Algorithm for Association Rule Mining

One of the more important techniques used in Data Mining is, Association Rule Mining and it involves the finding of frequent item sets from the database. A linked list version of one of the most widely used association mining algorithms that are the FP-Growth algorithm was proposed, called the FPBitLink Algorithm. It uses a bit matrix along with linked lists to find the desired item sets. In this paper, we propose two things, first is a variant of the FPBitLink Algorithm, in which instead of treating the items in the datasets as individual nodes, we take a single transaction to be the node in the linked list, and finally use UNION set operation to obtain the frequent pattern set. Since this transactional version is a highly efficient alternative when the number of transactions are greater than the number of items, as it saves valuable space and time, whereas the original FPBitLink algorithm is more efficient when the number of items is greater than the transactions, we further propose the installation of a checkpoint in the beginning, such that depending upon the data either of the two algorithms can be chosen. This way we arrive at a frequent pattern set finding procedure, which is both, highly efficient and extremely efficacious.

Fuzzy model generation using Subtractive and Fuzzy C-Means clustering

Clustering is a process of partitioning similar data into groups. For this, number of clustering algorithms have been proposed in literature. Some of them can also be used for the generation of Fuzzy Models. In this work, Sugeno fuzzy models being generated by Subtractive and FCM clustering have been discussed. Experiments have been performed on real datasets to compare the Subtractive and FCM clustering. Further, the effect of increase in the radius size is analyzed in Subtractive clustering. The average absolute error and root mean square error is also found out when using FCM clustering and Subtractive clustering with different values of radius.

Improved filtration step for mining association rules

In the course of association rules mining, frequent itemsets need to be identified. Well known association rule mining algorithm, Apriori, generates frequent itemsets. It generates frequent itemsets by repeating candidate generation and verification process until large itemsets are generated. Candidate generation process includes two steps-joining and pruning. An alternate for pruning step does the same job efficiently and is known as filtration. In this paper, an improved filtration step is proposed and evaluated for five standard databases. It is observed that improved filtration step is more efficient than pruning and filtration steps.

Evaluation of filtration and pruning approach for Apriori algorithm

One of the Association rule mining (ARM) algorithm, Apriori, is most popular algorithm. Pruning approach used in this algorithm differentiates between potential frequent and infrequent itemset well before verifying them in the given database. An alternate approach known as filtration does the same. In this paper, five experiments are carried out to prove that filtration approach works as efficient as Aprioris pruning approach but it requires an extra data structure.