Discovering Conditional Functional Dependencies(2011)

Note: Please Scroll Down to See the Download Link.


This paper investigates the discovery of conditional functional dependencies (CFDs). CFDs are a recent extension of functional dependencies (FDs) by supporting patterns of semantically related constants, and can be used as rules for cleaning relational data. However, finding quality CFDs is an expensive process that involves intensive manual effort. To effectively identify data cleaning rules, we develop techniques for discovering CFDs from relations. Already hard for traditional FDs, the discovery problem is more difficult for CFDs. Indeed, mining patterns in CFDs introduces new challenges. We provide three methods for CFD discovery. The first, referred to as CFD Miner, is based on techniques for mining closed item sets, and is used to discover constant CFDs, namely, CFDs with constant patterns only. Constant CFDs are particularly important for object identification, which is essential to data cleaning and data integration. The other two algorithms are developed for discovering general CFDs. One algorithm, referred to as CTANE, is a level wise algorithm that extends TANE, a well-known algorithm for mining FDs. The other, referred to as FastCFD, is based on the depth-first approach used in FastFD, a method for discovering FDs. It leverages closed-item set mining to reduce the search space. As verified by our experimental study, CFDMiner efficiently discovers constant CFDs. For general CFDs, CTANE works well when a given relation is large, but it does not scale well with the arity of the relation. FastCFD is far more efficient than CTANE when the arity of the relation is large; better still, leveraging optimization based on closed-itemset mining, FastCFD also scales well with the size of the relation. These algorithms provide a set of cleaning-rule discovery tools for users to choose for different applications.

Existing System:

As remarked earlier, constant CFDs are particularly important for object identification, and thus deserve a separate treatment. One wants efficient methods to discover constant CFDs alone, without paying the price of discovering all CFDs. Indeed, as will be seen later, constant CFD discovery is often several orders of magnitude faster than general CFD discovery.

Level wise algorithms may not perform well on sample relations of large arity, given their inherent exponential complexity. More effective methods have to be in place to deal with datasets with a large arity. A host of techniques have been developed for (non-redundant) association rule mining, and it is only natural to capitalize on these for CFD discovery. As we shall see, these techniques can not only be readily used in constant CFD discovery, but also significantly speed up general CFD discovery. To our knowledge, no previous work has considered these issues for CFD discovery.

Proposed System:

In light of these considerations we provide three algorithms for CFD discovery: one for discovering constant CFDs, and the other two for general CFDs.

(Module: 1) We propose a notion of minimal CFDs based on both the minimality of attributes and the minimality of patterns. Intuitively, minimal CFDs contain neither redundant attributes nor redundant patterns. Furthermore, we consider frequent CFDs that hold on a sample dataset r, namely, CFDs in which the pattern tuples have a support in r above a certain threshold. Frequent CFDs allow us to accommodate unreliable data with errors and noise. Our algorithms find minimal and frequent CFDs to help users identify quality cleaning rules from a possibly large set of CFDs that hold on the samples.

(Module: 2) Our first algorithm, referred to as CFDMiner, is for constant CFD discovery. We explore the connection between minimal constant CFDs and closed and free patterns. Based on this, CFDMiner finds constant CFDs by leveraging a latest mining technique proposed in [24], which mines closed itemsets and free item sets in parallel following a depth-first search scheme.

(Module: 3) Our second algorithm, referred to as CTANE, extends TANE to discover general CFDs. It is based on an attribute-set/pattern tuple lattice, and mines CFDs at level k + 1 of the lattice (i.e., when each set at the level consists of k+1 attributes) with pruning based on those at level k. CTANE discovers minimal CFDs only.

(Module: 4) Our third algorithm, referred to as FastCFD, discovers general CFDs by employing a depth-first search strategy instead of the levelwise approach. It is a nontrivial extension of FastFD mentioned above, by mining pattern tuples. A novel pruning technique is introduced by FastCFD, by leveraging constant CFDs found by CFDMiner. As opposed to CTANE, FastCFD does not take exponential time in the arity of sample data when a canonical cover of CFDs is not exponentially large.

(Module: 5) Our fifth and final contribution is an experimental study of the effectiveness and efficiency of our algorithms, based on real-life data (Wisconsin breast cancer and chess datasets from UCI) and synthetic datasets generated from data scraped from the Web. We evaluate the scalability of these methods by varying the sample size, the arity of relation schema, the active domains of attributes, and the support threshold for frequent CFDs. We find that constant CFD discovery (using CFDMiner) is often 3 orders of magnitude faster than general CFD discovery (using CTANE or FastCFD). We also find that FastCFD scales well with the arity: it is up to 3 orders of magnitude faster than CTANE when the arity is between 10 and 15, and it performs well when the arity is greater than 30; in contrast, CTANE cannot run to completion when the arity is above 17. On the other hand, CTANE is more sensitive to support threshold and outperforms FastCFD when the threshold is large and the arity is of a moderate size. We also find that our pruning techniques via itemset mining are effective: it improves the performance of FastCFD by 5-10  Folds and makes FastCFD scale well with the sample size. These results provide a guideline for when to use CFDMiner, CTANE or FastCFD in different applications.

These algorithms provide a set of promising tools to help reduce manual effort in the design of data-quality rules, for users to choose for different applications. They help make CFD-based cleaning a practical data quality tool.

System Configuration:-

H/W System Configuration:-

Processor                                 -    Pentium –III

Speed                             -    1.1 Ghz

RAM                              -    256 MB(min)

Hard Disk                      -   20 GB

Floppy Drive                 -    1.44 MB

Key Board                     -    Standard Windows Keyboard

Mouse                                -    Two or Three Button Mouse

Monitor                              -    SVGA

S/W System Configuration:-

v   Operating System          :Windows95/98/2000/XP

v   Application  Server       :   Tomcat5.0/6.X                                                

v   Front End                                :   HTML, Java, Jsp

v    Scripts                                    :   JavaScript.

v   Server side Script          :   Java Server Pages.

v   Database                       :   Ms-Access

v   Database Connectivity   :   JDBC.

Click here to download Discovering Conditional Functional Dependencies(2011) source code