BUKU ABSTRAK - Universiti Putra Malaysia

Science, Technology & Engineering
Robust Multicollinearity Diagnostic Measures based on Robust Coefficient
Determination
Assoc. Prof. Dr. Habshah Midi
Arezoo Bagheri
Institute of Mathematical Research, University Putra Malaysia,
43400 UPM Serdang, Selangor, Malaysia.
+603-8946 6876; habshahmidi@gmail.com
In this study, we propose the Robust Variance Inflation Factors (RVIFs) in the detection of multicollinearity
due to high leverage points or extreme outliers in the X-direction. The computation of RVIFs is based on robust
coefficient determinations which we called RR2 (MM) and RR2 (GM (DRGP)). RR2 (MM) is coefficient
determination of high breakdown point and efficient MM-estimators whereas RR2 (GM (DRGP)) has been defined
through an improved GM-estimators. GM (DRGP) is a GM-estimator with the main aim as downweighting
high leverage points with large residuals. It has been introduced by employing S-estimators as initial values,
Diagnostic Robust Generalized Potential based on MVE [DRGP (MVE)] as initial weight function and an
Iteratively Reweighted Least Squares (IRLS) has been utilised as a convergence method. The numerical results
and Monte Carlo simulation study indicate that the proposed RVIFs are very resistant to the high leverage points
and unable to detect the multicollinearity in the data especially RR2 [GM (DRGP)]. Hence, this indicates that the
high leverage points are the source of multicollinearity.
Keywords: Coefficient determination, generalised M-estimators, high leverage points, variance inflation factor
Extraction of information from the web pages becomes very important because the massive and increasing
amount of diverse semi-structured information sources in the Internet that are available to users, and the variety
of web pages that make the process of information extraction from the web a challenging problem and the
ever-growing large area of research. Many researchers work on extraction of information from web pages in
different domains such as business intelligence and products. Most of the previous works are limited due to the
facts that their approaches are not able to handle (i) web pages with the genuine and non genuine web tables and
(ii) the attributes that appear under different names but refer to the same entity (i.e., synonym). In this project,
we proposed a strategy for extracting and analysing information from semi-structured web data source which
consists of two approaches. An approach for extracting and classifying information from various web pages
and an approach for analysing and simplifying the extracted and classified information. Two analyses have been
conducted on four different domains. From the first analysis, the following can be concluded: during the process
of extracting information from various web pages, it is important to handle the genuine and non genuine web
tables as well as the synonyms as ignoring them might cause one to miss the information that is relevant to the
user. While in the second analysis, for the Nokia products, our proposed approaches achieved increment in F and
P as well as decrement in R as compared to that of Ashraf et al (2008). These analyses show that the proposed
strategy which includes two approaches is able to extract and analyse information from various web pages with
the genuine and non genuine web tables as well as handling the issue of synonyms.
Keywords: HTML web pages, information extraction
Extracting Information from Semi-structured Web Pages
Assoc. Prof. Dr. Hamidah Ibrahim
Mahmoud Sh. Al-Hassan, Ali Amer Alwan, Lili Nurliyana Abdullah and Aida Mustapha
Faculty of Computer Science and Information Technology, University Putra Malaysia,
43400 UPM Serdang, Selangor, Malaysia.
+603-8943 6510; hamidah@fsktm.upm.edu.my
194