|
Title:
|
DATA MINING OF SEQUENTIAL PATTERNS: AN APPLICATION TO THE NEXT BEST OFFER PROBLEM |
|
Author(s):
|
Cristina Frutuoso , Fernando Mira Da Silva , João Branco |
|
ISBN:
|
972-98947-0-1 |
|
Editors:
|
António Palma dos Reis and Pedro Isaías |
|
Year:
|
2003 |
|
Edition:
|
1 |
|
Keywords:
|
Data mining, knowledge discovery, association rules, time series analysis, linear regression, best next offer. |
|
Type:
|
Full Paper |
|
First Page:
|
444 |
|
Last Page:
|
451 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Several scientific and business domains require the discovery of hidden patterns in sequences of events. Common applications range from the classification of DNA sequences to the analysis of sequence of transactions in marketing environments. One method of mining sequential patterns is based on the derivation of association rules in sequences of events (Agrawal et al, 1995a). This approach is often based on extensions (Agrawal et al, 1995b) of the so-called apriori algorithm (Agrawal et al, 1993; Agrawal et al, 1994) for static sets. However, in its original formulation, the efficiency of the algorithm drops significantly when the number of simultaneous occurrences in the sequence is high. Moreover, conventional approaches to sequential pattern analysis based on this algorithm rely on the relative position of each event, but disregard the absolute time stamp of each event in the sequence. In this paper, we describe a variant of the apriori algorithm that is able to deal efficiently with large number of simultaneous occurrences. Moreover, we describe a method to estimate the time stamp of each candidate to the next occurrence in the sequence. We apply this algorithm to the so-called best next offer marketing problem: given the purchase history of a given customer, derive the most plausible next purchase and estimate the best moment to perform a marketing offer. |
|
|
|
|
|
|
