Vol 2, 2014

1. AUTOMATIC ESTIMATION OF THE NUMBER OF SOIL PROFILE LAYERS USING BAYESIAN INFORMATION CRITERION
Faculty of Applied Informatics and Mathematics, Warsaw University of Life Sciences
Faculty of Engineering and Environmental Sciences, Warsaw University of Life Sciences
Jarosław Kurek jaroslaw_kurek@sggw.pl, Michał Kruk michal_kruk@sggw.pl, Piotr Bilski piotr_bilski@sggw.pl, Bartosz Świderski bartosz_swiderski@sggw.pl
Simon Rabarijoely simon_rabarijoely@sggw.pl [Download PDF]

Abstract: In this study, Bayesian Information Criterion algorithm is utilized for the estimation of number of soil profile layers. In order to collect data, several probes are performed by geotechnical specialists in Warsaw University of Life Sciences (WUoLS) campus. Then soil profiles have been manually generated by geotechnical experts. It lets us to compare the results of novel automated method presented in this paper to real soil profile manually generated by geotechnical engineers. The database has been generated based on values derive from a probe CPT applied by geotechnical experts. Examination and accuracy calculation of the proposed method is presented and compared to reference real soil profile obtained by experts group.

2. MODELLING AND APPLICATION OF LOGISTIC REGRESSION AND ARTIFICIAL NEURAL NETWORKS MODELS
School of Ocean Engineering, Universiti Malaysia Terengganu, 21030 Kuala Terengganu, Terengganu, Malaysia.
Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, 40450 Shah Alam, Selangor, Malaysia.
Norhazlina Suhaimi, Adriana Ismail, Nurul Adyani Ghazali
norhazlinasuhaimi@gmail.com, adriana@tmsk.uitm.edu.my [Download PDF]

Abstract: Data mining involves exploration of large dataset to find valuable information that can aid decision making. The uses of data mining approach to build predictive model for predicting blood glucose level of T2DM patient after receiving treatment in ward. There are two data mining predictive models; logistic regression and artificial neural networks (ANNs). The purpose of this study work was to compare the performance of logistic regression and ANNs models for identifying risk factors that contributing to blood glucose level on T2DM patients. The scope of this study only involves one public hospital in Kelantan, Malaysia. 229 patients with T2DM who had received treatment in ward between 2008 and 2012 with ten input variables were selected. The classification accuracy, sensitivity and specificity have been measured to evaluate the performance for both models. For overall dataset, the logistic regression models achieved classification accuracy of 69.9% with a sensitivity of 50% and a specificity of 70.4%. The ANNs model reached classification accuracy 77.3% with sensitivity of 78.4% and specificity of 71.8%. Meanwhile, after partitioning dataset, logistic regression achieved classification accuracy of 71.3% with a sensitivity of 58.3% and specificity of 73.5%; and the ANNs reached classification accuracy of 72.5% with sensitivity of 72.3% and specificity of 75.3%. Hence, the ANNs model for the overall dataset had the highest classification accuracy compared to logistic regression model. Five important independent variables were identified on blood glucose level including diastolic blood pressure, platelet, white blood cell, low density lipoprotein and total cholesterol in ANNs analysis. This study would be able to contribute to the improvement of strategies and planning in hospital in Malaysia.
Keywords: T2DM, Logistic Regression, Artificial Neural Networks, Accuracy, Risk Factors

3.FIRST SEMESTER COMPUTER SCIENCE STUDENTS’ ACADEMIC PERFORMANCES ANALYSIS BY USING DATA MINING CLASSIFICATION ALGORITHMS
Faculty Informatics & Computing, Universiti Sultan Zainal Abidin (UniSZA)
Azwa Abdul Aziz, Nor Hafieza Ismail and Fadhilah Ahmad
azwaaziz@unisza.edu.my [Download PDF]

Abstract: The research on educational field that involves Data Mining techniques is rapidly increasing. Applying Data Mining techniques in an educational environment are known as Educational Data Mining that aims to discover hidden knowledge and patterns about students’ behaviour. This research aims to develop Students’ Academic Performance prediction models for the first semester Bachelor of Computer Science from Universiti Sultan Zainal Abidin (UniSZA) by using three selected classification methods; Naïve Bayes, Rule Based, and Decision Tree. The comparative analysis is also conducted to discover the best classification model for prediction. From the experiment, the models develop using Rule Based and Decision Tree algorithm shows the best result compared to the model develop from the Naïve Bayes algorithm. Five independent parameters (gender, race, hometown, family income, university entry mode) have been selected to conduct this study. These parameters are chosen based on prior research studies including from social sciences domains. The result discovers the race is a most influence parameter to the students’ performance followed by family income, gender, university entry mode, and hometown location parameters. The prediction model can be used to classify the students so the lecturer can take an early action to improve students’ performance.
Keywords: Educational Data Mining, Classification, Students’ Academic Performance, Naives-Bayes

4.MACHINE LEARNING BASED TICKET CLASSIFICATION IN ISSUE TRACKING SYSTEMS
Istanbul Technical University Istanbul, Turkey
Mücahit ALTINTAŞ, A. Cüneyd TANTUĞ
maltintas@itu.edu.tr, tantug@itu.edu.tr [Download PDF]

Abstract: Due to the rise of usage of virtual systems, support ticket systems have come into prominence. Addressing the issue tickets to appropriate person or unit in the support team has critical importance in order to provide improved end user satisfaction while ensuring better allotment of support recourses. The assignment of help ticket to appropriate group is still manually performed. Especially at large organizations, the manual assignment is not applicable sufficiently. It is time consuming and requires human efforts. There may be mistakes due to human errors. Also resource consumption is carried out ineffectively because of the misaddressing. On the other hand, manual assignment increases the response time which result in end user satisfaction deterioration. Multiple-choice systems which provide the user to choose the related categories or unit within defined categories may seem like better, but the systems are not useful because of those users, especially new users which have never used the system before, usually have no idea about the related category or department. Also users do not want to fill long ticket forms which are needed to identify the issue. In this study, an extension to ITS for auto-addressing the issue ticket to the relevant person or unit in support team is proposed.. In this system, bag of word approach, machine learning techniques and other algorithms which proven performance in text processing are used. The recommended method provides high quality user support and boosts end-user satisfaction. It reduces manual efforts and human errors while ensuring high service levels and improved end-user satisfaction.
Keywords: Issue Tracking System, Automatic Assignment, Ticket Classification

5.PREDICTING PURCHASED POLICY FOR CUSTOMERS IN ALLSTATE PURCHASE PREDICTION CHALLENGE ON KAGGLE
FAST-NU, Lahore, Pakistan.
Saba Arslan Shah and Mehreen Saeed
sabarslan@gmail.com, mehreen.saeed@nu.edu.pk [Download PDF]

Abstract: This paper gives an overview of the methodology developed for predicting the purchased policy for a customer in Allstate purchase prediction challenge held by Kaggle (Kaggle). It gives an account of challenges faced during the process and strategies used to predict the policy choices for customers. The techniques used include logistic regression, naïve Bayes (Mitchell, 1997), SVM (Lin., 2011) , random forest (Breiman, 2001), probability calculation for each policy and its change and a voting mechanism. Effect of previously presented policies is also measured. The dataset presented a challenge since it included feature set in both rows and columns for each of the customer. Furthermore, seven policy options were to be correct as a combination, for a prediction to be deemed accurate. Relationships are also explored between different policy options.
Keywords: Purchased policy prediction, Allstate purchase prediction challenge, SVM, Data mining, Random forest

6.MULTI-ROBOT COVERAGE PATH PLANNING FOR SEARCH AND SECURE PROBLEM IN A POLYGONAL ENVIRONMENT
Mechatronics Engineering Department, College of EME, National University of Sciences and Technology (NUST), Islamabad, Pakistan.
Saad Shaikh and Dr. Kunwar Faraz Ahmad Khan
ssaad.shaikh87@gmail.com
kfaraz@gmail.com [Download PDF]

Abstract: In this paper, we study the visibility-based search and secure problem, in which multiple robots move through a simply connected and known polygonal environment such that they guarantee the detection of all contaminants or intruders within, that can move arbitrarily fast. Our objective is to investigate the usefulness of the GRAPH-CLEAR methodology for analytical modeling of this problem and develop a strategy for an uncoordinated search to accomplish this task, targeting application domains such as securing buildings for surveillance and security purposes. The aim is neither to take care of the actual control system of the robots, nor to handle the actual detection of an intruder (through video or image processing); rather introduce an effective coverage path planning strategy to identify the paths the robots are required to follow. To this end, we present an algorithm to compute the strategy required, using the concept of searcher and blocker robots/agents. The algorithm models the given environment in the form of a connectivity graph and then converts it into a tree, calculating the number of robots and the trajectories these robots have to follow in order to secure the complete environment. We show how the strategy devised by this algorithm is more time and cost efficient by performing simulations and comparing the results against methods already being used, giving examples of different computed trajectories. For a particular case study, the results of the algorithm show an approximately 1/4th increase in time efficiency along with a 1/3rd reduction in the robot resources required.
Keywords: search and secure, graph-clear, coverage path planning, searchers and blockers


e-Journal of AICS (Artificial Intelligence & Computer Science), is devoted to disseminate high quality refereed articles in the field of ICT. It is an bi-anually published journal, and also will be available online for free.

We would like to invite all the researches, academicians, and ICT professionals throughout Malaysia to contribute to this journal.

The theme for this edition is “Artificial Intelligence in Education“.

Medium – Articles may only be in English.

Papers are accepted throughout the year (due to the approval of the reviewer). However for the second issue publication, the due date for the full paper is on

May 1st, 2014.

Be a Reviewer

We appreciate your contribution as a reviewer.  Please email your resume to aics.wcr@gmail.com . We will get back to you in few days.


Editorial Committee

Dr. Mokmin Basri
CW Shamsul Bahri CW Ahmad
Khirulnizam Abd Rahman
Kolej Universiti Islam Antarabangsa Selangor (KUIS), MALAYSIA.

(e-ISSN: 2289-5965 )

Published by
WorldConferences.net
Koperasi KOKISDAR
Kolej Universiti Islam Antarabangsa Selangor
Bandar Seri Putra, 43600 Kajang, Selangor, MALAYSIA

Please do not hesitate to contact us at;
Tel/WhatsApp: +6012-2206954 (Dr Mokmin Basri)
Email : aics.wcr@gmail.com

Copyright © 2014 WorldConferences.net
All rights reserved. No part of this publication may be reproduced in any form or any means without prior permission from the copyright holder.
All articles in this proceeding are openly accessible here

Advertisements

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s