Feature Selection with Naïve Bayes Classifier

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 FeatureExtraction 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