Dynamic Log Session Identification Using A Novel Incremental Learning Approach For Database Trace Logs
No Thumbnail Available
Date
2015
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Identification of session is very significant task in database for determining helpful patterns from database trace logs files In
recent years, several number of work have been done to dynamic log session identification among them n-gram models is recently
proposed, which produces higher log session identification results. But the major issue of the n-gram model is that it assumes the entire
database query to be static, so dynamic query type is not applicable. In this paper, a novel online incremental learning for dynamic log data
trace session identification schema based on the adaptation method to database trace logs is proposed. It is applied for dynamic log
session identification with automatic selection of threshold based on the standard deviation schema, so it is named as DS-OILSD. The
proposed novel DS-OILSD schema is varied from normal n-gram model since the proposed work consists of two major modifications. The
first one is to solve the parameters adjustment problem of the IL in online and offline manner incrementally. The second one is used to
dynamic management of query types and allocating initial probabilities to the n-grams models. The proposed DS-OILSD leaning method is
based on modified MAP estimation schema for dynamic changing adaptation of the query types and is instinctively reasonable. It directly
solves the problem of dynamic log session identification, in which three types of learning are performed such as labeled data, semi-labeled,
unlabeled data for various categories of training data Finally experimental work is conducted to proposed and existing state of art schema
for dynamic log session identification and experimentation results are evaluated based on the parameters like, F-measure, and precision
for clinic database log files.