On the Prediction Accuracies of Three Most Known Regularizers : Ridge Regression, The Lasso Estimate and Elastic
Net Regularization Methods

Adel Aloraini
Computer Science department, Qassim University, Saudi Arabia.

ABSTRACT
The work in this paper shows intensive empirical experiments using 13 datasets to understand the regularization effectiveness of ridge regression, the lasso estimate, and elastic net regularization methods.the study offers a deep understanding of how the datasets affect the goodness of the prediction accuracy of each regularization method for a given problem given the diver- sity in the datasets used. the results have shown that datasets play crucial rules on the performance of the regularization method and that the
predication accuracy depends heavily on the nature of the sampled datasets.

KEYWORDS
ridge regression, regularization, the lasso estimate, elastic net.

Original Source URL :http://aircconline.com/ijaia/V8N6/8617ijaia03.pdf

http://airccse.org/journal/ijaia/current2017.html

Web Evolution – The Shift from Information Publishing to Reasoning

Abdulelah A. Algosaibi, Saleh Albahli, Samer F. Khasawneh, and Austin Melton4
1Department of Computer Science, King Faisal University, KSA.
2Department of Information Technology, Qassim University, KSA
3Department of Computer Science, Walsh University, USA.
4Department of Computer Science, Kent State University, USA

ABSTRACT
The Web, as communication channel, has had variety of development that allows information to be published and accessed in a scaleable approach. With the revolution of the information, some research studies have conducted to boost the present situation and propose advance version of the Web. Therefore, it is important to look into the new version of the Web in order to improve the way that information is expressed, to make more intelligent choices and to obtain a better meaning of the information over the Web. That is, future web would require specific architecture in order to support the extracting of better meaning or “reasoning”. With Web 1.0 and Web 2.0, the current information over the Web is not understandable for the machines. Understanding is big shift for wide open door for innovatoion and reasoning. In this work, we research the progress of the Web from Web 1.0, Web 2.0, Web 3.0, Web 4.0, to Web 5.0. We are pointing out document types and technologies employed to understand the changes from Web 1.0 to Web 3.0 and to predicate the future of the Web (Web 4.0 and Web 5.0). Also, we present the current status and concerns about the Web as an information source and communication channel.

KEYWORDS
Web Generations; Web 1.0; Web 2.0; Web 3.0; Web 4.0; Web 5.0; World Wide Web; WWW; Semantic Web; WebOS; Intelligent Agent.

Original Source URL:http://aircconline.com/ijaia/V8N6/8617ijaia02.pdf

http://airccse.org/journal/ijaia/current2017.html

An Effective Arabic Text Classification Approach Based on Kernel Naive Bayes Classifier 

Raed Al-khurayji1
and Ahmed Sameh2
1,2Faculty of Computer and Information Science, University Prince Sultan University,
Riyadh, Saudi Arabia.

ABSTRACT
With growing texts of electronic documents used in many applications, a fast and accurate text classification method is very important. Arabic text classification is one of the most challenging topics. This is probably caused by the fact that Arabic words have unlimited variation in the meaning, in addition to the problems that are specific to Arabic language only. Many studies have been proved that Naive Bayes (NB) classifier is being relatively robust, easy to implement, fast, and accurate for many different fields such as text classification. However, non-linear classification and strong violations of the independence assumptions problems can lead to very poor performance of NB classifier. In this paper, first, we preprocess the Arabic documents to tokenize only the Arabic words. Second, we convert those words into vectors using term frequency and inverse document frequency (TF-IDF) technique. Third, we propose an efficient approach based on Kernel Naive Bayes (KNB) classifier to solve the non-linearity problem of Arabic text classification. Finally, experimental results and performance evaluation on our collected dataset of Arabic topic mining corpus are presented, showing the effectiveness of the proposed KNB classifier against other baseline classifiers.

KEYWORDS
Arabic Language, Text Classification, Machine Learning, Naïve Bayes Classifier, Kernel Estimation Function.

Original source Link: http://aircconline.com/ijaia/V8N6/8617ijaia01.pdf

http://airccse.org/journal/ijaia/current2017.html

Hardware Design for Machine Learning

Pooja Jawandhiya,

School of Electrical and Electronic Engineering,
Nanyang Technological University, Singapore.

ABSTRACT

Things like growing volumes and varieties of available data, cheaper and more powerful computational processing, data storage and large-value predictions that can guide better decisions and smart actions in real time without human intervention are playing critical role in this age. All of these require models that can automatically analyse large complex data and deliver quick accurate results – even on a very large scale. Machine learning plays a significant role in developing these models. The applications of machine learning range from speech and object recognition to analysis and prediction of finance markets. Artificial Neural Network is one of the important algorithms of machine learning that is inspired by the structure and functional aspects of the biological neural networks. In this paper, we discuss the purpose, representation and classification methods for developing hardware for machine learning with the main focus on neural networks. This paper also presents the requirements, design issues and optimization techniques for building hardware architecture of neural networks.

KEYWORDS

Artificial intelligence (AI), application specific integrated circuit (ASIC), artificial neural network (ANN), central processing unit (CPU), field programmable gate array (FPGA), graphics processing unit (GPU), machine learning (ML), neurochip.

Original Source Link: http://aircconline.com/ijaia/V9N1/9118ijaia05.pdf 

http://www.airccse.org/journal/ijaia/current2018.html

 

Estimate of the Head Produced by Electrical Submersible Pumps on Gaseous Petroleum Fluids, A Radial Basis Function
Network Approach

Morteza Mohammadzaheri1, Mojataba Ghodsi1 and Abdullah AlQallaf 2
1Department of Mechanical and Industrial Engineering,
Sultan Qaboos University, Muscat, Oman
2Department of Electrical Engineering, Kuwait University
Kuwait City, Kuwait.

ABSTRACT
This paper reports successful development of an exact and an efficient radial basis function network (RBFN) model to estimate the head of gaseous petroleum fluids (GPFs) in electrical submersible pumps (ESPs). Head of GPFs in ESPs is now often estimated using empirical models. Overfitting and its consequent lack of model generality data is a potentially serious issue. In addition, available data series is fairly small, including the results of 110 experiments. All these limits were considered in RBFN design process, and highly accurate RBFNs were developed and cross validated.

KEYWORDS
Electrical Submersible Pump(ESP), Radial Basis Function Network (RBFN), Model, Petroleum, Gaseous, Head Estimation.

Original Source Link: http://aircconline.com/ijaia/V9N1/9118ijaia04.pdf

http://www.airccse.org/journal/ijaia/current2018.html

Comparison of Artificial Neural Networks and Fuzzy Logic Approaches for Crack Detection in a Beam Like Structure 

B Prakruthi Gowd1, K Jayasree2, and Manjunath N. Hegde3

1Assistant Professor, Department of Civil Engineering,
Vidya Jyothi Institute of Technology, Hyderabad, India.
2Assistant Professor, Department of Civil Engineering,
Vasavi College of Engineering, Hyderabad, India.
3Professor and Dean (Academics), Department of Civil Engineering,
Dr. Ambedkar Institute of Technology, Bengaluru, India.

ABSTRACT

This paper proposes two algorithms of crack detection one using fuzzy logic (FL) and the other artificial neural networks (ANN). Since modal parameters are very sensitive to damages, the first three relative natural frequencies are used as three inputs and the corresponding relative crack location, relative crack depth are used as the two outputs in the algorithms. The three natural frequencies for an undamaged beam and different cases of damaged beam (Single crack at various locations with varying depths) were obtained by modelling and simulating the beams using a finite element based (FEM) software. Results concluded that both the approaches can be successfully employed in crack detection in a beam like structure but FL approach performed better in determining relative crack depth whereas ANN approach performed better in  determining relative crack location. All the comparisons made in the study are based on the R2 values.

KEYWORDS

Damage detection, Modal Properties, Artificial Neural Networks, Fuzzy Logic, R2
values.

Original Source URL:http://aircconline.com/ijaia/V9N1/9118ijaia03.pdf

http://www.airccse.org/journal/ijaia/current2018.html

A New Multi-Criteria Decision Making Method: Approach of Logarithmic Concept (APLOCO)

Tevfik Bulut
The Ministry of Science, Industry, and Technology, Ankara 6000, Turkey

ABSTRACT
The primary aim of the study is to introduce APLOCO method which is developed for the solution of multicriteria decision making problems both theoretically and practically. In this context, application subject of APLACO constitutes evaluation of investment potential of different cities in metropolitan status in Turkey.The secondary purpose of the study is to identify the independent variables affecting the factories in the operating phase and to estimate the effect levels of independent variables on the dependent variable in the organized industrial zones (OIZs), whose mission is to reduce regional development disparities and to mobilize local production dynamics. For this purpose, the effect levels of independent variables on dependent variables have been determined using the multilayer perceptron (MLP) method, which has a wide use in artificial neural networks (ANNs). The effect levels derived from MLP have been then used as the weight levels of the decision criteria in APLOCO. The independent variables included in MLP are also used as the decision criteria in APLOCO. According to the results obtained from APLOCO, Istanbul city is the best alternative in term of the investment potential and other alternatives are Manisa, Denizli, Izmir, Kocaeli, Bursa, Ankara, Adana, and Antalya, respectively.Although APLOCO is used to solve the ranking problem in order to show application process in the paper, it can be employed easily in the solution of classification and selection problems. On the other hand, the study also shows a rare example of the nested usage of APLOCO which is one of the methods of operation research as well as MLP used in a determination of weights.

KEYWORDS
approach of logarithmic concept (APLOCO), multi-criteria decision analysis (MCDA), multilayer perceptron (MLP), organized industrial zones (OIZs), artificial intelligence, artificial neural networks (ANNs).

Original Source URL :http://aircconline.com/ijaia/V9N1/9118ijaia02.pdf

http://www.airccse.org/journal/ijaia/current2018.html

A Quantitative Approach in Heuristic Evaluation of E-Commerce Websites

Xiaosong Li, Ye Liu, Zizhou Fan and Will Li
Computer Science Practice Pathway,
Unitec Institute of Technology, Auckland, New Zealand

ABSTRACT

This paper presents a pilot study on developing an instrument to predict the quality of e-commerce websites. The 8C model was adopted as the reference model of the heuristic evaluation. Each dimension of the 8C was mapped into a set of quantitative website elements, selected websites were scraped to get the quantitative website elements, and the score of each dimension was calculated. A software was developed in PHP for the experiments. In the training process, 10 experiments were conducted and quantitative analyses were regressively conducted between the experiments. The conversion rate was used to verify the heuristic evaluation of an e-commerce website after each experiment. The results showed that the mapping revisions between the experiments improved the performance of the evaluation instrument, therefore the experiment process and the quantitative mapping revision guideline proposed was on the right track. The software resulted from the experiment 10 can serve as the aimed e-commerce website evaluation
instrument. The experiment results and the future work have been discussed.

KEYWORDS

E-commerce Website, Heuristic Evaluation, Regression Experiments, 8C framework, Quantitative Analysis.

Original Source URL: http://aircconline.com/ijaia/V9N1/9118ijaia01.pdf 

http://www.airccse.org/journal/ijaia/current2018.html