ICTACT Journal on Soft Computing

S. K. Jayanthi ¹, S. Sasikala ²

Affiliations
1 Department of Computer Science, Vellalar College for Women, IN
2 Department of Computer Science, Hindusthan College of Arts and Science, IN

Abstract

Spamdexing is the art of black hat SEO. Features which are more influential for high rank and visibility are manipulated for the SEO task. The motivation behind the work is utilizing the state of the art Website optimization features to enhance the performance of spamdexing detection. Features which play a focal role in current SEO strategies show a significant deviation for spam and non-spam samples. This paper proposes 44 features named as NLSDF (New Link Spamdexing Detection Features). Social media creates an impact in search engine results ranking. Features pertaining to the social media were incorporated with the NLSDF features to boost the recital of the spamdexing classification. The NLSDF features with 44 attributes along with 5 social media features boost the classification performance of the WEBSPAM-UK 2007 dataset. The one tailed paired t-test with 95% confidence, performed on the AUC values of the learning models shows significance of the NLSDF.

Keywords

Web Spam, Search Engine, SVM, Decision Table, HMM.

Full Text

     

S. K. Jayanthi ¹, S. Sasikala ²

Affiliations
1 Department of Computer Science, Vellalar College for Women, IN
2 Department of Computer Science, K.S.R College of Arts and Science, IN

Abstract

Search engines are the doorsteps for retrieving required information from the web. Web spam is a bad method for improving the ranking and visibility of the web pages in search engine results. This paper addresses the problem of the link spam classification through the features of the web sites. Link related features retrieved from the website are used to discriminate the spam and non-spam sites. AIS inspired algorithms are applied for the dataset and results are evaluated. Artificial immune systems are machine learning systems inspired by the principles of the natural immunology. It comprises of supervised learning schemes which can be adapted for the wide range of the classification problems.UK- WEBSPAM-2007 Dataset [8] is used for the experiments. WEKA [9] is used to simulate the classifiers. Artificial Immune Recognition algorithm seems to perform well than the other classes. Best classification accuracy attained is 98.89 by AIRS1 Algorithm. This seems to be good when comparing with the other classifiers accuracy available on the existing literature.

Keywords

Web Spam, Search Engine, TPP, FCA, AIRS.

Full Text

     

S. K. Jayanthi ¹, S. Prema ²

Affiliations
1 Department of Computer Science, Vellalar College for Women, IN
2 Department of Computer Science, K.S.R College of Arts and Science, IN

Abstract

In the current scenario, web page result personalization is playing a vital role. Nearly 80 % of the users expect the best results in the first page itself without having any persistence to browse longer in URL mode. This research work focuses on two main themes: Semantic web search through online and Domain based search through offline. The first part is to find an effective method which allows grouping similar results together using BookShelf Data Structure and organizing the various clusters. The second one is focused on the academic domain based search through offline. This paper focuses on finding documents which are similar and how Vector space can be used to solve it. So more weightage is given for the principles and working methodology of similarity propagation. Cosine similarity measure is used for finding the relevancy among the documents.

Keywords

Semantic Web, BookShelf Data Structure, Similarity Propagation, Cosine Similarity measure, Vector Space Model.

Full Text

     

S. K. Jayanthi ¹, S. Sasikala ²

Affiliations
1 Department of Computer Science, Vellalar College for Women, IN
2 Department of Computer Science, KSR College of Arts and Science, IN

Abstract

Search Engines are used for retrieving the information from the web. Most of the times, the importance is laid on top 10 results sometimes it may shrink as top 5, because of the time constraint and reliability on the search engines. Users believe that top 10 or 5 of total results are more relevant. Here comes the problem of spamdexing. It is a method to deceive the search result quality. Falsified metrics such as inserting enormous amount of keywords or links in website may take that website to the top 10 or 5 positions. This paper proposes a classifier based on the Reptree (Regression tree representative). As an initial step Link-based features such as neighbors, pagerank, truncated pagerank, trustrank and assortativity related attributes are inferred. Based on this features, tree is constructed. The tree uses the feature inference to differentiate spam sites from legitimate sites. WEBSPAM-UK-2007 dataset is taken as a base. It is preprocessed and converted into five datasets FEATA, FEATB, FEATC, FEATD and FEATE. Only link based features are taken for experiments. This paper focus on link spam alone. Finally a representative tree is created which will more precisely classify the web spam entries. Results are given. Regression tree classification seems to perform well as shown through experiments.

Keywords

Web Link Spam, Classification, Reptree, Decision Tree, Search Engine.

Full Text

     

S. K. Jayanthi ¹, S. Prema ²

Affiliations
1 Department of Computer Science, Vellalar College for Women, IN
2 Department of Computer Science, K.S.R College of Arts and Science, IN

Abstract

Search engines are playing a vital role in retrieving relevant information for the web user. In this research work a user profile based web search is proposed. So the web user from different domain may receive different set of results. The main challenging work is to provide relevant results at the right level of reading difficulty. Estimating user expertise and re-ranking the results are the main aspects of this paper. The retrieved results are arranged in Bookshelf Data Structure for easy access. Better presentation of search results hence increases the usability of web search engines significantly in visual mode.

Keywords

Web Search Personalization, Bookshelf Data Structure, Agglomerative Hierarchical Clustering, Similarity Measure, Visualization.

Full Text