📝 Abstract

Automatically recognizing the e-learning activities is an important task for improving the online learning process. Probabilistic graphical models such as hidden Markov models and conditional random fields have been successfully used in order to identify a Web users activity. For such models, the sequences of observation are crucial for training and inference processes. Despite the efficiency of these probabilistic graphical models in segmenting and labeling stochastic sequences, their performance is adversely affected by the imperfect quality of data used for the construction of sequences of observation. In this paper, a formalism of the possibilistic theory will be used in order to propose a new approach for observation sequences preparation. The eminent contribution of our approach is to evaluate the effect of possibilistic reasoning during the generation of observation sequences on the effectiveness of hidden Markov models and conditional random fields models. Using a dataset containing 51 real manipulations related to three types of learners tasks, the preliminary experiments demonstrate that the sequences of observation obtained based on possibilistic reasoning significantly improve the performance of hidden Marvov models and conditional random fields models in the automatic recognition of the e-learning activities.

💡 Analysis

Automatically recognizing the e-learning activities is an important task for improving the online learning process. Probabilistic graphical models such as hidden Markov models and conditional random fields have been successfully used in order to identify a Web users activity. For such models, the sequences of observation are crucial for training and inference processes. Despite the efficiency of these probabilistic graphical models in segmenting and labeling stochastic sequences, their performance is adversely affected by the imperfect quality of data used for the construction of sequences of observation. In this paper, a formalism of the possibilistic theory will be used in order to propose a new approach for observation sequences preparation. The eminent contribution of our approach is to evaluate the effect of possibilistic reasoning during the generation of observation sequences on the effectiveness of hidden Markov models and conditional random fields models. Using a dataset containing 51 real manipulations related to three types of learners tasks, the preliminary experiments demonstrate that the sequences of observation obtained based on possibilistic reasoning significantly improve the performance of hidden Marvov models and conditional random fields models in the automatic recognition of the e-learning activities.

📄 Content

Mouse Movement and Probabilistic Graphical Models
based E-learning Activity Recognition Improvement Possibilistic Model

Anis Elbahi #1, Mohamed Nazih Omri#2, Mohamed Ali Mahjoub*3 and Kamel Garrouch #4

Research Unit MARS,

Department of computer sciences Faculty of sciences of Monastir, Monastir, Tunisia. 1Elbahi.anis@gmail.com 2MohamedNazih.Omri@fsm.rnu.tn 4kamelg_2001@yahoo.fr

• Research Unit SAGE,
  National Engineering School of Sousse, Sousse, Tunisia. 3Medali.mahjoub@ipeim.rnu.tn

Abstract— Automatically recognizing the e-learning activities is an important task for improving the online learning process. Probabilistic graphical models such as Hidden Markov Models and Conditional Random Fields have been successfully used in order to identify a web user activity. For such models, the sequences of observation are crucial for training and inference processes. Despite the efficiency of these probabilistic graphical models in segmenting and labeling stochastic sequences, their performance is adversely affected by the imperfect quality of data used for the construction of sequences of observation.
In this paper, a formalism of the possibilistic theory will be used in order to propose a new approach for observation sequences preparation. The eminent contribution of our approach is to evaluate the effect of possibilistic reasoning -during the generation of observation sequences- on the effectiveness of Hidden Markov Models and Conditional Random Fields models. Using a dataset containing 51 real manipulations related to three types of learners’ tasks, the preliminary experiments demonstrate that the sequences of observation obtained based on possibilistic reasoning significantly improve the performance of Hidden Marvov Models and Conditional Random Fields models in the automatic recognition of the e-learning activities.
Keywords— Hidden Markov Models, Conditional Random Fields, Sequence of Observations, Fuzzy Logic, Possibilistic Theory, E-learning Activity Recognition, Mouse Movement Tracking.

1. INTRODUCTION Human activity recognition is an important area of computer vision research [1]. Its applications include a variety of systems that involve interactions between persons and electronic devices such as Human Computer Interfaces. Indeed, identifying web users’ activities and preferences especially in a web-based educational environment is crucial for monitoring and interpreting their navigational behavior. Today, online learning has become an important part of society. Therefore, the analysis of learners’ navigational behavior plays an important role in the improvement of the learning process. This analysis can be done using, Probabilistic Graphical Models (PGM) which have been successfully used in this field as well as for pattern classification tasks [2, 3]. Two of the most important PGM are the Hidden Markov Models (HMM) and Conditional Random Fields (CRF).
   Thanks to their flexibility and elasticity in spatio-temporal pattern recognition tasks and despite all adjustments which have focused on the structure and used algorithms for HMM [12-14] and CRF models [15-18], very few studies have considered the “imperfect” quality of data used by these probabilistic models, specifically in the preparation of sequences of observations used for training and inference. In fact, the data used during training and inference processes are the sources of uncertainty for model predictions [19]. Moreover, the probability distributions on which are based probabilistic models may be affected uncertainty and inconsistency due to the effects of data randomness. HMM and CRF have been developed to be trained (parameters adjustment) in order to label stochastic temporal sequences. In order to ensure these tasks, the sequences of observations are extremely essential. To generate an observation sequence, let M={x1, x2, …, xN} be the alphabet of N discrete symbols. Each symbol (xi) is intended to describe “perfectly” at a given time, one state of a real world process. During the generation of each sequence of observations, only one symbol of M will be chosen in order to represent the real state of the dynamic process. Commonly, the choice of this symbol is based on “imperfect” information (knowledge). To better understand this, the following is a simple example describing how to build a sequence of observation based on real observations. Using the alphabet M={‘walking’, ‘jogging’, ‘running’} we can create a sequence of observations concerning the sporting activity by observing an athlete in a race. Therefore, X1={walking1, jogging2, jogging3, …, runningt, …, walkingT} is the observing sequence describing the activity of athlete 1. Indeed, for preparing the sequence X1, the observer (human or machine) noted – at the first time step - that the athlete is in the activity walking=true, jogging=false and running=false. The same observer noted – at the second time step – that

This content is AI-processed based on ArXiv data.