Papers presented in SAS Forum

1) Neural Network Models Using SAS Enterprise Miner

Neural network is advocated to outperform traditional methods in time series analysis. This paper intends to investigate the dominance of neural network to methods such as exponential smoothing, regression and Box-Jenkins using the same ground rules. The models are developed using large number of historical data with dynamic patterns to predict future by dividing the data set into training and validating set. Models are then compared based on their forecast accuracy in the validation sample, and model with the least forecast error is selected. Significant emphasis in this paper is on the use of one of the powerful data-mining method in SAS module called Enterprise Miner. The modeling process incorporates the Enterprise Miner SEMMA methodology which stands for Sampling, Exploring, Modifying, Modeling, and Assessing data. Lastly, this paper will conclude with a brief summary of the advantages and disadvantages of using neural network model in time series forecasting.

Full Paper

2) Evaluation of Computer Ethics: Confirmation Factor Models Using Proc Calis

The computer is considered one of the most essential technological advances and has become an everyday tool. Many tertiary institutions have become heavily dependent on computers, Internet and informations systems for educational purpose. Computers and the Internet also represent many people, organizations, and governments. However, at the same time they raise some ethical issues such as unauthorized access and use of computer systems, software piracy and information privacy. This study proposes the employment of computer use and computer security concepts for investigating students' ethical conduct related to computer ethics. Specifically, an ethical computer awareness (ECA) construct concerning computer use and security is developed and validated. The process of evaluating a measurement instrument for reliability and investigating the factor structure are discussed using the scale of ethical computer awareness (SECA). SAS? procedures served to provide an indication of the internal consistency, that is, reliability with PROC CORR, to explore the factor structure with exploratory factor analysis using PROC FACTOR, and to verify the factor structure with confirmatory factor analysis using PROC CALIS of the measurement instrument. The ECA construct developed from the study could be useful to research a wide range of computer ethics in the future.

Full Paper

3) Using SAS Graphics to Explore Employers Preference When Hiring New Graduates

One of the fundamental goals of universities is preparing students for the workforce. Thus, understanding the job competencies required by employers is crucial in matching university's curriculum with industry's requirement. This is to ensure that new graduates are well-equipped with the necessary skills when they enter the job market. This paper presents the results of an exploratory study to evaluate employers preference based on the essential skills for various types of jobs that require tertiary qualification. Analysis is done by analyzing a broad set of online job descriptions extracted from job advertisements in online job portals. Analysis is based on six categories of employability skills: communication, analytical, interpersonal, personal, information management and technical skills. Analyses are carried out using SAS/STAT and the results are presented using graphics provided by SAS/GRAPH and SAS/IML. Among the tools used are mosaic diagrams and link graphs. The results show that there are indications of differences in employability skills requirement between different categories of jobs. The findings of this study can be used as a basis for the determination of patterns of skills required for jobs in any areas of specialization. This in turn will help in creating and sustaining workforce development programs to create a skilled and knowledgeable workforce.

Full Paper

4) Students' Learning and Academic Performance

The computer is considered one of the most essential technological advances and has become an everyday tool. Institutions of higher learning are engaged in a sustained and continuous process of maximising the quality of their graduates so as to enhance their readiness for the job market. Thus, it is important for educational institutions to focus on improving the critical aspects of teaching and learning. One area which has received increasing attention is the learning styles of students. Several studies have shown that academic performance of university students is related to their learning styles. The objectives of this study were to ascertain the dominant learning styles of the students and to discover the relationship between learning style and academic performance. The Grasha-Riechmann Student Learning Style Scales (GRSLSS) was administered to determine student learning preferences in six learning style categories. The subjects of this study were the first year students at the International Education Center (INTEC), Universiti Teknologi MARA, Shah Alam. These sponsored students were undergoing their preparatory programmes at INTEC before pursuing their degree at reputable universities in, Australia, New Zealand, the United Kingdom and the United States, among others. Cluster analysis was used to identify their dominant learning styles, while discriminant analysis was used to analyse the relationship between learning styles and the various demographic and educational variables. Academic performance based on learning style was found to be significant. Analysis was carried out using SAS/STAT, SAS/GRAPH and SAS/EM.

Full Paper

5) Ensemble Method Hit Ration For Robust Tests of Spread

The computer is considered one of the most essential technological advances and has become an everyday tool. An ensemble method was used to combine the outputs of several diverse classifiers to form a potentially stronger solution in ensemble system. The SAS system facilitates the building of program that can perform a composite method that combined logistic regression and discriminant analysis by using PROC LOGISTIC, PROC DISCRIM, DATA step, PROC FREQ and other SAS functions. We aim to improve the correct classification rate from the same data set. To achieve this, classification from logistic regression and discriminant analysis were integrated to form the ensemble. We took the averages of posterior probabilities (for the target values) from logistic regression and discriminant analysis, and classified according to the average posterior probabilities. Then, we created correct classification tables by defining predicted values based on the prior probability from the average posterior probabilities. The intended audiences for this paper are those who have working knowledge of Base SAS and have the fundamental grasp of statistics.

Full Paper