Research Article
Fahad Ayyaz, Amna Asif, Asif Z
Abstract
Education is important for developing nations. It also needs to be improved in Pakistan. The education system of Pakistan has developed in past years but there is still some needs to improve it further. The measure of student performance is important but it is not possible with on manual manners from some specific personals. Measuring the student performance by analyzing the large amount of accurate results can be done using the machine learning algorithms. From these algorithms the sorted form of data is analyzed and the results are made for making any specific actions. For this purpose, in this research work institute data of different primary schools in Punjab province is gathered. After the collection, raw data is changed into the sorted data and by selecting the eight suitable set of attributes. As each entity has a specific set of attributes and it is depending upon the situation that which attribute is selected for which purpose. From these attributes some are originally taken and some are derived from other set of original attributes. This data is uploaded in the rapid miner analysis tool and analyzed the data in terms of prediction by selecting the retention as special attributes. The algorithms are selected which are more suitable for this data. The attributes are Naive Bayesian, random forest, support vector machine and deep learning. Each of the algorithm has unique nature of classification and produced results in different manners but in the collective sense student and teacher both are involve in the student’s performance. The degree of the teacher and age of the teacher are matter a lot while the attendance is another important factor which affects the performance of the student in terms of retention. The reason to select four algorithms is to make more accurate analysis and each algorithm has different level of accuracies. This methodology improves the quality education because through this large amount of student records can be processed and analyzed. Furthermore, this analysis can also be done in future by applying different set of parameters.