Frequent Patterns

Web Usage Mining the data mining techniques to extract exciting usage patterns from SNS data, in order to understand and serve the needs of users in an efficient manner. Usage data captures the origin and the identity of the user along with the browsing behavior at the social web site so that they can serve well. It is further classified depending on the type of data to be mine. This paper proposes an algorithm which takes advantages of both apriori growth algorithm and Fp split algorithm and shows the experimental results which shows the efficiency of the proposed algorithm.

Cellular networks monitoring and management tasks are based on huge amounts of continuously collected data from network elements and devices. Log files are used to store this data, but it might need to accumulate millions of lines in one day. The standard name of this log is in GPEH format which stands for General Performance Event Handling. This log is usually recorded in a binary format (bin). Thus, efficient and fast compression technique is considered as one of the main aspects targeting the storage capabilities. On the other hand, based on our experience, we noticed that experts and network engineers are not interested in each log entry. In addition, this massive burst of entries can lose important information; especially those translated into performance abnormalities. Thus, summarizing log files would be beneficial in specifying the different problems on certain elements, the overall performance and the expected network future state. In this paper, we introduce an efficient compression algorithm based log frequent patterns. In addition, we propose a Mixed Mode Summary-based Lossless Compression Technique for Mobile Networks log files (MMSLC) as a mixed on-line and off-line compression modes based on the summary extracted from the frequent patterns. Our scheme exploits the strong correlation between the directly and consecutively recorded bin files for utilizing the online compression mode. On the other hand, it uses the famous “Apriori Algorithm” to extract the frequent patterns from the current file in offline mode. Our proposed scheme is proved to gain high compression ratios in fast speed as well as help in extracting beneficial information from the recorded data.

This research paper proposes an algorithm to find association rules for incremental databases. Most of the transaction databases are often dynamic. Suppose consider super market customers daily purchase transactions. Day to day customer’s behaviour to purchase items may change and new products replace old products. In this scenario static data mining algorithms doesn't make good significance. If an algorithm continuously learns day to day, then we can get most updated knowledge. This is very much helpful in present fast updating world. Famous and benchmarked algorithms for Association rules mining are Apriory and FP- Growth. However, the major drawback in Appriory and FP-Growth is, they must be rebuilt all over again once the original database is changed. Therefore, in this paper we introduce an efficient algorithm called Binary Decision Tree (BDT) to process incremental data. To process continuously data we need so much of processing and storage resources. In this algorithm we scan data base only once by which we construct dynamic growing binary tree to find association rules with better performance and optimum storage. We can apply for static data also, but our main intension is to give optimum solution for incremental data.

In this paper, we propose a new algorithm for mining frequent itemsets. This algorithm is named AMFI (Algorithm for Mining Frequent Itemsets). This algorithm compresses the data while maintaining the necessary semantics for the frequent itemsets mining problem and it is more efficient that traditional compression algorithms. The AMFI efficiency is based on a compressed vertical binary representation of the data and on a very fast support count. AMFI performs a breadth first search through equivalence classes. We compare our proposal with an implementation using PackBits algorithm.

Text mining is an emerging research field evolving from information retrieval area. Clustering and classification are the two approaches in data mining which may also be used to perform text classification and text clustering. The former is supervised while the later is un-supervised. In this paper, our objective is to perform text clustering by defining an improved distance metric to compute the similarity between two text files. We use incremental frequent pattern mining to find frequent items and reduce dimensionality. The improved distance metric may also be used to perform text classification. The distance metric is validated for the worst, average and best case situations [15]. The results show the proposed distance metric outperforms the existing measures.

Data mining is used to discover interesting and useful knowledge from massive data. Finding interesting patterns play an important role in knowledge discovery process and are essential for many real life applications. Recently high utility pattern mining is important for mining high utility itemsets which overcomes the limitation of frequent pattern mining. High utility pattern mining is used to identify the itemsets with highest utilities, by considering profit, quantity, cost or other user preferences. This research paper proposes an enhanced high utility pattern approach to mine the high utility itemsets with less computation time and less memory space when larger itemsets are explored for complex datasets.

Spatial Data Mining is based on correlation of spatial objects in space. Mining frequent pattern from spatial databases systems has always remained a challenge for researchers. In the light of the first law of geography “everything is related to everything else but nearby things is more related than distant things” suggests that values taken from samples of spatial data near to each other tend to be more similar than those taken farther apart. This tendency is termed spatial autocorrelation or spatial dependence. It’s natural that most spatial data are not independent, they have high autocorrelation. In this paper, we propose an enhancement of existing mining algorithm for efficiently mining frequent patterns for spatial objects occurring in space such as a city is located near a river. The frequency of each spatial object in relation to other object tends to determine multiple occurrence of the same object. We further enhance the proposed approach by using a numerical method. This method uses a tree structure based methodology for mining frequent patterns considering the frequency of each object stored at each node of the tree. Experimental results suggest significant improvement in finding valid frequent patterns over existing methods.