Technical Seminar 2004 An Adaptive Algorithm for Detection

Technical Seminar 2004 An Adaptive Algorithm for Detection of Duplicate Records Presented By: Rama kanta Behera IT 200127207 Under the guidance of : Miss Ipsita Mishra RAMAKANTA BEHERA IT 200127207 1

An Adaptive Algorithm for Detection of Duplicate Records Technical Seminar 2004 INTRODUCTION v A “records set” is a list of prior distinct records. A new record is to be verified for a duplicate against the records set v. A database is a collection of related data. v Various Algorithms like • • • Matching learning algo, Learnable string similarity measures Adaptive Algo RAMAKANTA BEHERA IT 200127207 2

An Adaptive Algorithm for Detection of Duplicate Records Technical Seminar 2004 OBJECTIVES v Reduced cost of duplicate record detection. v Perfect scalability of one such detection procedure. v Cache prior information of distinct records and thus cause retaining of prior records redundant for furthering the search v Keep the algorithm adaptive. RAMAKANTA BEHERA IT 200127207 3

An Adaptive Algorithm for Detection of Duplicate Records Technical Seminar 2004 PREVALENT METHODS v. The Brute Force Method This method consumes complexity of the order number of records in the records set and requires all prior records to be stored. v. Method by Rail et. al The comparison of a new record against the records set is reduced from being full text match to comparing two integers RAMAKANTA BEHERA IT 200127207 4

An Adaptive Algorithm for Detection of Duplicate Records Technical Seminar 2004 OUTLINE OF THE PROPOSED SOLUTION The central idea behind the present algorithm is based on the fundamental property of primality of numbers Record set Integer number space f(x) I Fig: hashing Record set Integer number f(x) I Prime number g(x) P Fig: Extended hashing into prime space RAMAKANTA BEHERA IT 200127207 5

Technical Seminar 2004 An Adaptive Algorithm for Detection of Duplicate Records f(x) g(x) r 1 r 2 … rn I 1 I 2 … In P 1 P 2 … Pn PRODUCT(Pprior) P 1*p 2…*pn= Pprior Fig: The complete algorithm RAMAKANTA BEHERA IT 200127207 6

An Adaptive Algorithm for Detection of Duplicate Records Technical Seminar 2004 REALIZATION OF THE ALGORITHM Two functions f(x) and g(x) are to be realized for the implementation of the algorithm. v Realizing f(x) v Realizing g(x) RAMAKANTA BEHERA IT 200127207 7

An Adaptive Algorithm for Detection of Duplicate Records Technical Seminar 2004 STEPS OF THE ALGORITHM Step 1 : For each new record, hash is performed and unique hash value (Hnew) for each distinct record is obtained. Step 2 : Hnew is mapped to its corresponding unique prime (Pnew). Step 3 : Pprior is divided with Pnew. If Pnew exactly divides Pprior, then the corresponding record to Pnew is a duplicate and already exists in Pprior. Else, Pnew is a distinct record. Step 4 : If Pnew is a distinct record, Pprior is multiplied with Pnew and the result is stored back in Pprior. Thus updating Pprior renders the algorithm adaptive. RAMAKANTA BEHERA IT 200127207 8

Technical Seminar 2004 An Adaptive Algorithm for Detection of Duplicate Records Fig: Flowchart RAMAKANTA BEHERA IT 200127207 9

An Adaptive Algorithm for Detection of Duplicate Records Technical Seminar 2004 IMPLEMENTATIONS There are three important implementation details that need to be discussed v Size of Records set v Use of Logarithms v Subsets of Records set RAMAKANTA BEHERA IT 200127207 10

An Adaptive Algorithm for Detection of Duplicate Records Technical Seminar 2004 CONCLUSION v A new approach to handle duplicate records is presented v This approach combines the concepts of number theory and algorithmic to solve the oftener felt problem of “duplicate record detection”. RAMAKANTA BEHERA IT 200127207 11

Technical Seminar 2004 An Adaptive Algorithm for Detection of Duplicate Records THANK YOU !!! RAMAKANTA BEHERA IT 200127207 12