Youssouf EL ALLIOUI

Youssouf EL ALLIOUI

Research scientist - Semantic web, machine learning, educational data mining (EDM)

Talks and presentations

A new collaborative approach to solve the gray-sheep users problem in recommender systems

October 30, 2019

Talk, 2019 Third International Conference on Intelligent Computing in Data Sciences (ICDS), Marrakech, Morocco

Abstract: Recommender systems aim to help users to find items that fit their requirements and preferences. In that field, the collaborative filtering (CF) approach is considered as a widely used one. There are two main approaches for CF: memory-based and model-based. Both of the two approaches are based on the use of users’ ratings to predict the top-N recommendation for the active user. Despite its simplicity and efficiency, The CF approach stills suffer from many drawbacks including sparsity, gray sheep and scalability. The aim of this work is to deal with the gray sheep problem, by proposing a novel collaborative filtering approach. This novel approach aims to enhance the accuracy of prediction by turning the users whose preferences disagree with the target user, into new similar neighbors. For instance, if a user X is dissimilar to a user Y then the user ¬ X is similar to the user Y. To evaluate the performance of the proposed approach, we have used two datasets including MovieLens and FilmTrust. The Experimental results show that our approach outperforms many traditional recommendation techniques.

Improving Collaborative Filtering Approach by Leveraging Opposite Users

July 11, 2019

Talk, International Conference on Advanced Intelligent Systems for Sustainable Development (AI2SD’2019), Marrakech, Morocco

Abstract: Collaborative filtering is a widely used recommendation approach that aims to predict for a target user the most appropriate items. This approach uses the ratings given by users who share similar tastes and preferences to predict ratings for items that haven’t been rated yet. Despite its simplicity and justifiability, CF approach stills suffering from several drawbacks and problems, including sparsity, gray sheep and scalability. These problems affect the accuracy of the obtained results. In this work, we present a novel collaborative filtering approach based on the opposite preferences of users. We focus on enhancing the accuracy of predictions and dealing with gray sheep problem by inferring new similar neighbors based on users who have dissimilar tastes and preferences. For instance, if a user X is dissimilar to a user Y then the user ┐X is similar to the user Y. The Experimental results performed on two datasets including MovieLens and FilmTrust show that our approach outperforms several baseline recommendation techniques

A Multiple Linear Regression-Based Approach to Predict Student Performance

July 11, 2019

Talk, International Conference on Advanced Intelligent Systems for Sustainable Development (AI2SD’2019), Marrakech, Morocco

Abstract: Predicting students’ academic outcome is useful for any educational institution that aims to ameliorate students’ performance. Based on the resulted predictions, educators can provide support to students at risk of failure. Data mining and machine learning techniques were widely used to predict students’ performance. This process called Educational data mining. In this work, we have proposed a methodology to build a student’ performance prediction model using a supervised machine learning technique which is the multiple linear regression (MLR). Our methodology consists of three major steps, the first step aims to analyze and preprocess the students’ attributes/variables using a set of statistical analysis methods, and then the second step consists in selecting the most important variables using different methods. The third step aims to construct different MLR models based on the selected variables and compare their performance using the k-fold cross-validation technique. The obtained results show that the model built using the variables selected from the Multivariate Adaptive Regression Splines method (MARS), outperforms the other constructed models.

Mining Learners’ Behaviors: An Approach Based on Educational Data Mining Techniques

May 03, 2019

Talk, ESAI'19: International Conference On Embedded Systems And Artificial Intelligence, Faculty of Medicine and Pharmacy Fez, Morocco, April 30-May 3, 2019

Abstract: Educational data mining is a research field that aims to apply data mining techniques in educational environments. Many data mining techniques such as clustering, classification, and prediction can be performed on educational data in order to analyze the learner behaviors. In this work, we have used the clustering and classification techniques to predict the learners’ learning styles. The students’ behaviors while using the e-learning system have been captured from the log file and given as an input of a clustering algorithm to group them into 16 clusters. The resulted clusters were labeled with learning styles combinations based on the Felder and Silverman learning style model. Then the labeled behaviors were given as input to four classifiers: naive Bayes, Cart, Id3, and C4.5 to compare their performance in predicting students’ learning styles. The four classifiers were performed using Weka data mining tool, and the obtained results showed that Id3 yielded better results than the other classifiers.

A Novel Collaborative Filtering Approach Based on the Opposite Neighbors’ Preferences

May 03, 2019

Talk, ESAI'19: International Conference On Embedded Systems And Artificial Intelligence, Faculty of Medicine and Pharmacy Fez, Morocco, April 30-May 3, 2019

Abstract: Collaborative filtering (CF) has become an effective way to predict useful items. It is the most widespread recommendation technique. It relies on users who share similar tastes and preferences to suggest the items that they might be interested in. Despite its simplicity and justifiability, the collaborative filtering approach experiences many problems, including sparsity, gray sheep and scalability. These problems lead to deteriorating the accuracy of the obtained results. In this work, we present a novel collaborative filtering approach based on the opposite preferences of users. We focus on enhancing the accuracy of predictions and dealing with gray sheep problem by inferring new similar neighbors based on users who have dissimilar tastes and preferences. For instance, if a user X is dissimilar to a user Y then the user ┐X is similar to the user Y. The Experimental results performed on two datasets including MovieLens and FilmTrust show that our approach outperforms several baseline recommendation techniques.

Combining supervised and unsupervised machine learning algorithms to predict the learners’ learning styles

October 05, 2018

Talk, The second International Conference On Intelligent Computing in Data Sciences (ICDS'2018), Fez, Morocco

Abstract: The implementation of an efficient adaptive e-learning system requires the construction of an effective student model that represents the student’s characteristics, among those characteristics, there is the learning style that refers to the way in which a student prefers to learn. Knowing learning styles helps adaptive E-learning systems to improve the learning process by providing customized materials to students. In this work, we have proposed an approach to identify the learning style automatically based on the existing learners’ behaviors and using web usage mining techniques and machine learning algorithms. The web usage mining techniques were used to pre-process the log file extracted from the E-learning environment and capture the learners’ sequences. The captured learners’ sequences were given as an input to the K-modes clustering algorithm to group them into 16 learning style combinations based on the Felder and Silverman learning style model. Then the naive Bayes classifier was used to predict the learning style of a student in real time. To perform our approach, we used a real dataset extracted from an e-learning system’s log file, and in order to evaluate the performance of the used classifier, the confusion matrix method was used. The obtained results demonstrate that our approach yields excellent results.

A Hybrid Machine Learning Approach to Predict Learning Styles in Adaptive E-Learning System

July 14, 2018

Talk, International Conference on Advanced Intelligent Systems for Sustainable Development (AI2SD’2018), Tanger, Morocco

Abstract: The increasing use of E-learning environments by learners makes it indispensable to implement adaptive e-learning systems (AeS). The AeS have to take into account the learners’ learning styles to provide convenient contents and enhance the learning process. Learning styles refer to the preferred way in which an individual learns best. The traditional methods detecting learning styles (using questionnaires) present many limits, as: (1) the time-consuming process of filling in the questionnaire and (2) producing inaccurate results because students aren’t always aware of their own learning preferences. Thus in this paper we have proposed an approach for detecting learning styles automatically, based on Felder and Silverman learning style model (FSLSM) and using machine learning algorithms. The proposed approach is composed of two parts: The first part aims to extract the learners’ sequences from the log file, and then using an unsupervised algorithm (K-means) in order to group them into sixteen clusters according to the FSLSM, and the second part consists in using a supervised algorithm (Naive Bayes) to predict the learning style for a new sequence or a new learner. To perform our approach, we used a real dataset extracted from an e-learning system’s log file. In order to evaluate the performance, we used the confusion matrix technique. The obtained results demonstrate that our approach yields excellent results.

Integrating web usage mining for an automatic learner profile detection: A learning styles-based approach

April 04, 2018

Talk, 2018 International Conference on Intelligent Systems and Computer Vision (ISCV), Fez, Morocco

Abstract: With the technological revolution of Internet and the information overload, adaptive E-learning has become the promising solution for educational institutions since it enhances students’ learning process according to many factors such as their learning styles. Learning styles are a criteria of great import in E-learning environment because they can help the system to effectively personalize students’ learning process. Generally, the traditional way of detecting students’ learning style is based on asking students to fill out a questionnaire. However, using this static technique presents many problems. Some of these problems include the lack of self-awareness of students of their learning preferences. In addition, almost all students are bored when they are asked to fill out a questionnaire. Thus, in this work, we present an automatic approach for detecting students’ learning style based on web usage mining. It consists in classifying students’ log files according to a specific learning style model (Felder and Silverman model) using clustering algorithms (K-means algorithm). In order to test the efficiency of our work, we use a real-world dataset gathered from an E-learning system. Experimental results show that our approach provide promising results.