Badr Eddine El Mohajir

Exploitation of the recommendation systems in the calculation of the learning path

In this paper we present a model of an adapting learning system based on a recommendation system, operating after assessment to correct learning path of the learners who have experienced difficulties in the assessment. Our approach aims to correct the learning path of learners who have not passed the assessment successfully, by calculating the similarities in behaviors between them and those who did.

A novel hybrid model based on Hodrick–Prescott filter and support vector regression algorithm for optimizing stock market price prediction

Predicting stock market price is considered as a challenging task of financial time series analysis, which is of great interest to stock investors, stock traders and applied researchers. Many machine learning techniques have been used in this area to predict the stock market price, including regression algorithms which can be useful tools to provide good performance of financial time series prediction. Support Vector Regression is one of the most powerful algorithms in machine learning. There have been countless successes in utilizing SVR algorithm for stock market prediction. In this paper, we propose a novel hybrid approach based on machine learning and filtering techniques. Our proposed approach combines Support Vector Regression and Hodrick–Prescott filter in order to optimize the prediction of stock price. To assess the performance of this proposed approach, we have conducted several experiments using real world datasets. The principle objective of this paper is to demonstrate the improvement in predictive performance of stock market and verify the works of our proposed model in comparison with other optimized models. The experimental results confirm that the proposed algorithm constitutes a powerful model for predicting stock market prices.

Optimizing stock market price prediction using a hybrid approach based on HP filter and support vector regression

Predicting stock prices is an important task of financial time series forecasting, which is of great interest to stock investors, stock traders and applied researchers. Many machine learning techniques have been used in recent times to predict the stock price, including regression algorithms which can be useful tools to provide good accuracy of financial time series forecasting. In this paper, we propose a novel hybrid approach which combines Support Vector Regression and Hodrick-Prescott filter in order to optimize the prediction of stock price. To assess the performance of this proposed approach, we have conducted several experiments using Maroc Telecom (IAM) financial time series. It is daily data collected during the period between 2004 and 2016. The experimental results confirm that the proposed model is more powerful in term of predicting stock prices.

A Recommendation System Operating after Assessment to Correct Learning Paths in a Content Adapting System

In this paper we present a model of an adapting learning system based on a recommendation system, operating after assessment to correct learning path of the learners who have experienced difficulties in the assessment. Our approach aims to correct the learning path of the learners who have failed at the assessment, by calculating the similarities in behaviors between them and those who did, then recommend them the learning objects that can build the most relevant learning path .

Short: Graphical Specification and Automatic Verification of Business Process

This paper deals with the integration of the formal verification techniques of business process (BP) in the design phase. In order to achieve this purpose, we use the graphical notation of Business Process Modeling Notation (BPMN) for modeling BP and specifying constraint properties to be verified. A formal semantics for some response properties are given.

Arabic Text Generation Using Recurrent Neural Networks

In this paper, we applied Recurrent Neural Networks (RNNs) Language Model on Arabic Language by training and testing it on “Arab World Books” and “Hindawi” free Arabic text datasets. While the standard architecture of RNNs does not match ideally with Arabic, we adapted a RNN model to deal with Arabic features. Our proposition in this paper is a gated Long-Short Term Memory (LSTM) model responding to some Arabic language criteria. As originality of the paper, we demonstrate the power of our LSTM model in generating Arabic text comparing to the standard LSTM model. Our results, comparing to English and Chinese text generation, have been promising and gave sufficient accuracy.

Syntactic entropy for main content extraction from web pages

In this paper, we present a solution for main content identification in web pages. Our solution is language-independent; Web pages may be written in different languages. It is topic-independent; no domain knowledge or dictionary is applied. And it is unsupervised; no training phase is necessary. The solution exploits the tree structure of web pages and the frequencies of text tokens to attribute scores of content density to the areas of the page and by the way identify the most important one. We tested this solution over representative examples of web pages to show how efficient and accurate it is. The results were satisfying.

A technical reading in statistical and neural machines translation (SMT & NMT)

Automatic translation of natural languages has been an active body of research in the last decades, especially when it comes to statistical translation which uses machine learning algorithms for translation tasks. Machine translation being a key application in the field of natural language processing, it leads to develop many approaches namely, statistical machine translation and recently neural machine translation. In this paper, we present a survey of the state of the art, where we describe the context of the current research studies by reviewing both the statistical machine translation and neural machine translation, and an overview of the main strengths and limitations of the two approaches.

Development of a bilingual corpus of Arabic and Arabic Sign Language based on a signed content

This paper deals with the creation process of a bilingual corpus of Arabic and Arabic Sign Language. This work will be concerned with a challenging research approach that aims to develop a solution that meets communication need of the Arab deaf community. Many sign language corpora have been proposed, but unfortunately, none of them is adapted to Arabic Sign Language (ArSL). To tackle this problem we designed a methodology for developing a bilingual Arabic and Arabic Sign Language corpus. To achieve this, the study will rely on sign content exported from Aljazeera Mubachir Egypt [1] TV news for deaf and hearing-impaired people.

ALS_CORR[LP]: An adaptive learning system based on the learning styles of Felder-Silverman and a Bayesian network

The aim of this paper is to present the adaptive learning system called ALS_CORR[LP]1. This system belongs to a very specific class of the e-learning systems, which is the adaptive learning ones. In fact they have the ability to adapt the learning process according to each learner needs, learning styles, objectives, etc. ALS_CORR[LP] is based on the learner prerequisites and the learning styles of Felder-Silverman, to design the learner model. As for the domain model, it is designed according to the recommendations of the differentiated pedagogy, which advocates creating multiple versions of the same learning object. Finally in order to ensure the adaptation inside the system, a Bayesian network, to match the designed learning object with the specifics of the learner profile was developed. It is also necessary to emphasize, that the major feature of the system is, its ability to correct the generated learning path in case of a failure in the evaluation phase. The learning path relevance is questioned, based on a recommendation system which enables updating the initial profile, or recommending the most relevant versions of the learning object, in case where the similarity calculation in behavior, reveals that the observed behavior in the system does not fit the initial profile description.