International Journal of Computer Applications (0975 – 8887) Volume 77– No.15, September 2013 12 Mining Time Variant Frequent Pattern using PPM and PWM: A Comparison P. A. Shirsath M.Tech (CSE) Lord Krishna College of Technology, Indore Vijay Kumar Verma Assistant Professor M. Tech (CSE) Lord Krishna College of Technology, Indore ABSTRACT The process of exploring and analyzing data from different perspective, using automatic or semiautomatic techniques is called Data mining. Data mining extracts knowledge or useful information and discovers correlations or meaningful patterns and rules from large databases [1, 2]. Using these patterns and rules it is possible for business enterprises to identify new and unexpected trends, subtle relations in the data and use them to increase revenue and cut cost. In this paper we proposed a comparative study over Progressive Partition Miner (PPM) and Progressive Weighted miner (PWM). Keywords Progressive, Partition, Miner, Weighted, Comparative 1. INTRODUCTION The traditional data mining techniques not have ability to analyse variation of data over time and treat them as ordinary data. Temporal datasets includes stock market data, manufacturing or production data, maintenance data, web mining and point-of-sale records. Temporal data mining means mining or discovering knowledge and patterns from temporal databases. Temporal data mining is an extension of data mining with ability to include time attribute analysis. Due to the significance and complexity of the time attribute, a lot of different kinds of patterns are of interest [3, 5]. Two time aspects are included in temporal databases namely, valid time and transaction time. The time period during which a fact is true with respect to the real world is considered as valid time and the time period during which a fact is stored in the database is called transaction time. According to these two time aspects temporal databases allow the division of three different forms [4, 25]. They are a. A historical database stores data with respect to valid time. b. A rollback database stores data with respect to transaction time. c. A bitemporal database stores data with respect to both valid and transaction time, that is, they store the history of data with respect to valid time and transaction time. 2. TEMPORAL DATA MINING TASKS A main question is how to apply traditional data mining techniques on a temporal database. Temporal data mining may involve the following areas of investigation. Temporal data mining tasks includes: i. Temporal association rules ii. Temporal data classification and comparison iii. Temporal pattern analysis iv. Temporal clustering analysis v. Temporal prediction and trend analysis vi. Temporal classification [6] 3. RELATED WORK A temporal association rule is defined as the frequency of an itemset over a time period T and is the number of transactions in which it occurs divided by total number of transaction over a time period. To solve the problem on handling time-series by including time expression into association rules temporal association rule mining has been introduced. Temporal association rule mining is first introduced by Wang, Yang and Muntz in years 1999-2001.Temporal association rule mining is introduced together with the introduction of the TAR (Temporal Association Rule) algorithm. With the help of Temporal association we can finds the valuable relationship among the different item sets, in temporal database. There are several types of temporal association rules defined by various researcher e such as inter transaction rules, episode rules, trend dependencies, sequence association rules [6, 7, 8]. Roddick and Spiliopoulou (2002) have presented a comprehensive overview of techniques for the mining of temporal data using three dimensions: data type, mining operations and type of timing information (ordering). Winarko and Roddick, 2005 proposed a non Apriori-based technique that avoids multiple database scans, this methods not only avoid multiple data scan but efficiently mine arrangements and rules in a temporal database. The main drawback of this method is that it do not consider any constraints for the temporal relations and do not examine any measures for their rules other than the traditional confidence [9]. Tansel and Imberman (2007) proposed a method where association rules were extracted for consecutive time intervals with different time granularities. They proposed a simple operation that extracts portions of a temporal relation was used during mining process and was combined with the first step of discovering association rules. Using this approach, the process of knowledge discovery can observe the changes and variation in the association rules over the time period when these rules are valid [10, 11]. Gharib et al. (2010) proposed a method for generating temporal association rules to solve the problem of handling time series by including time expressions into association rules. To solve this they extended an incremental algorithm to maintain the temporal association rules in a transaction database, at the same time maintains the benefits from the results of earlier mining to derive the final mining output [7, 12, 14]. C. H. Lee et proposed progressive partition miner (PPM). In PPM the database is first partitioned the dataset by the size of time granularity. Then it applies filtering threshold mechanism on each partition of the database and prune out