Data Mining and Knowledge Engineering

P. Usha ¹, K. Nandhini ²

Affiliations
1 Department of Computer Science, Dr. N.G.P. Arts and Science, Coimbatore -48, Tamil Nadu, IN
2 Dr.N.G.P. Arts and Science, Coimbatore -48, Tamil Nadu, IN

Abstract

Predicting the performance of a student is a great concern to the higher education managements, where several factors affect the performance. The scope of this paper is to investigate the accuracy of data mining techniques in such an environment. The first step of the study is to gather student’s data and their marks and technical, analytical, communicational and problem solving abilities. We collected records of 200 under graduate students of computer science course, from a private Educational Institution conducting various Under Graduate and Post Graduate courses. The second step is to extract and select the data and choose the relevant attributes for the accuracy using genetic algorithm. Attributes were classified into two groups “Demographic Attributes” and “Performance Attributes”. In the third step, support vector machine algorithms were constructed and their performances were evaluated. This work will help the institute to accurately predict the performance of the students for their growth.

Keywords

Decision Tree, Data Mining, Feature Extraction and Selection, Genetic Algorithm, Support Vector Machines.

Full Text

     

E. Chandra ¹, K. Nandhini ²

Affiliations
1 Department of Computer Applications, D. J. Academy for Managerial Excellence, Coimbatore-32, Tamilnadu, IN
2 Computer Science Department, Dr. N. G. P. Arts and Science College, Coimbatore-48, Tamilnadu, IN

Abstract

The ability to predicting the performance of a student is very essential task of all educational institutions. This will not be decided by using only the academic excellence of a student. The behaviors such as aptitude, attitude, communications, technological, interpersonally, problem solving ability etc., should be taken into care to predict the real excellence of a student. This form a heterogeneous dataset covering cross section of categorical, integer type data types etc. This has given rise to a high dimensional dataset which will hamper classification process. Since this is the task of prediction and mining the classification algorithms of data mining is used. The decision tree algorithms of classification are one of the fine grained methods to bring the more accuracy of prediction. The first phase of the work is collecting the wide cross section of atabase of values for attributes which are quite cross functional. The second phase plays vital role for effective classification by narrowing down by selection of predictive attributes. This phase is done by Feature Extraction techniques to reduce the high dimensional dataset in to a low dimensional dataset. The third phase applying the algorithms uses the Naive Bayes and tree induction of decision tree methods for actual classification of the data. The scalability of these methods has improved by perception based learning. Also, there is a school of thought that one can take up the classification and data mining without incorporating any Dimensionality reduction techniques like Feature Extraction. This work compare results obtained by the both process and study the performance of the Prediction accuracy. It is not that only the student domain can be used for excellence prediction. It can be applied for any kind of domain.

Keywords

Data Mining, Decision Tree, Feature Extraction, Performance Prediction.

Full Text

     

K. Nandhini ¹, S. Jayanthi ¹

Affiliations
1 Department of Computer Applications, S.A. Engineering College, IN

Abstract

In this paper and analyse a behaviour-rule specification-based technique for intrusion detection of medical devices embedded in a Medical Cyber Physical System (MCPS) in which the patient's safety is of the utmost importance. A methodology to transform behaviour rules to a state machine, so that a device that is being monitored for its behaviour can easily be checked against the transformed state machine for deviation from its behaviour specification.

Using vital sign monitor medical devices as an example; In demonstrate that our intrusion detection technique can effectively trade false positives off for a high detection probability to cope with more sophisticated and hidden attackers to support ultra-safe and secure MCPS applications. Moreover, through a comparative analysis, A demonstrate that our behaviour-rule specification-based IDS technique outperforms two existing anomaly-based techniques for detecting abnormal patient behaviours in pervasive healthcare applications.

Keywords

Privacy Protection, Data Sharing, NTRU, Collaborative IDS.

Full Text