<p>Educational Data Mining (EDM)  research has taking an important place as it helps in exposing useful knowledge from educational data sets to be employed and serve several purposes such as predicting students’ achievements. Predicting student’s achievements might be useful for building and adopting several changes in the educational environments as a re-action in the current educational systems. Most of the existing research have used machine learning to predict students’ achievements by using diverse attributes such as family income, students gender, students absence and level etc.  In this paper, the effort is made to explore the effectiveness of using the deep learning algorithm more precisely CNN to predict students’ achievements which could hlp in predicting if student will be able to finish their degree or not.  The experimental results reveal how the proposed model outperformed the existing approaches in terms of prediction accuracy.</p>

The discovery of software faults at early stages plays an important role in improving software quality; reduce the costs, time, and effort that should be spent on software development. Machine learning (ML) have been widely used in the software faults prediction (SFP), ML algorithms provide varying results in terms of predicting software fault. Deep learning achieves remarkable performance in various areas such as computer vision, natural language processing, speech recognition, and other fields. In this study, two deep learning algorithms are studied, Multi-layer perceptron's (MLPs) and Convolutional Neural Network (CNN) to address the factors that might have an influence on the performance of both algorithms. The experiment results show how modifying parameters is directly affecting the resulting improvement, these parameters are manipulated until the optimal number for each of them is reached. Moreover, the experiments show that the effect of modifying parameters had an important role in prediction performance, which reached a high rate in comparison with the traditional ML algorithm. To validate our assumptions, the experiments are conducted on four common NASA datasets. The result shows how the addressed factors might increase or decrease the fault detection rate measurement. The improvement rate was as follows up to 43.5% for PC1, 8% for KC1, 18% for KC2 and 76.5% for CM1.

Software faults prediction (SFP) processes can be used for detecting faulty constructs at early stages of the development lifecycle, in addition to its being used in several phases of the development process. Machine learning (ML) is widely used in this area. One of the most promising subsets from ML is deep learning that achieves remarkable performance in various areas. Two deep learning algorithms are used in this paper, the Multi-layer perceptrons (MLPs) and Convolutional Neural Network (CNN). In order to evaluate the studied algorithms, four commonly used datasets from NASA are used i.e. (PC1, KC1, KC2 and CM1). The experiment results show how the CNN algorithm achieves prediction superiority of the MLP algorithm. The accuracy and detection rate measurements when using CNN has reached the standard ratio respectively as follows: PC1 97.7% - 73.9%, KC1 100% - 100%, KC2 99.3% - 99.2% and CM1 97.3% - 82.3%. This study provides promising results in using the deep learning for software fault prediction research.

During software maintenance, software systems need to be modified by adding or modifying source code. These changes are required to fix errors or adopt new requirements raised by stakeholders or market place. Identifying the targeted piece of code for refactoring purposes is considered a real challenge for software developers. The whole process of refactoring mainly relies on software developers' skills and intuition. In this paper, a deep learning algorithm is used to develop a refactoring prediction model for highlighting the classes that require refactoring. More specifically, the gated recurrent unit algorithm is used with proposed pre-processing steps for refactoring prediction at the class level. The effectiveness of the proposed model is evaluated using a very common dataset of 7 open source java projects. The experiments are conducted before and after balancing the dataset to investigate the influence of data sampling on the performance of the prediction model. The experimental analysis reveals a promising result in the field of code refactoring prediction.