Predicting Assignment Submissions in a Multiclass Classification Problem
Journal Title: TEM JOURNAL - Year 2015, Vol 4, Issue 3
Abstract
Predicting student failure is an important task that can empower educators to counteract the factors that affect student performance. In this paper, a part of the bigger problem of predicting student failure is addressed: predicting the students that do not complete their assignment tasks. For solving this problem, real data collected by our university’s educational platform was used. Because the problem consisted of predicting one of three possible classes (multi-class classification), the appropriate algorithms and methods were selected. Several experiments were carried out to find the best approach for this prediction problem and the used data set. An approach of time segmentation is proposed in order to facilitate the prediction from early on. Methods that address the problems of high dimensionality and imbalanced data were also evaluated. The outcome of each approach is shown and compared in order to select the best performing classification algorithm for the problem at hand.
Authors and Affiliations
Bogdan Drăgulescu, Marian Bucos, Radu Vasiu
Keywords
Related Articles
- EP ID EP163823
- DOI -
- Views 175
- Downloads 0
Comparison of Feature Selection Techniques in Knowledge Discovery Process
The process of knowledge discovery in data consists of five steps. Data preparation, which includes data cleaning and feature selection, takes away from 60% to 95% total time of the whole process. Thus, it is cruci...
Modeling of Three Dimensional Magnetic Field in Three-Phase Induction Motor with Double Squirrel Cage
From three-phase induction machines, asynchronous double-cage rotor motors stand out as more specific, as in the construction and after the performance. Such a type of motor is selected as the objective of a study, and i...
Cardiometabolic Risk Calculation in the Assessment of Cardiometabolic Risk Profiles
Numerous factors, together labelled as cardiometabolic risk (CMR), increase chances for developing cardiometabolic diseases. CMR estimation has a crucial role in the early prevention and enables the prompt treatmen...
Investigation of Partially Crystalline Zr77Ni23 Metallic Glass
This paper presents the results of an extensive research of partially crystalline Zr77Ni23 metallic glass (indicated numbers refer to atomic percentages). The partially crystalline Zr77Ni23 samples were prepared by...
Impact of Turquality Model on Branding and International Marketing
A brand consists of name, symbol, concept, design, standard, quality or several of them which are used to distinguish a product from its competitors. Without a brand, all products have the same qualities in the eye...