Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Adaptive Recommendation System for MOOC Naveen Bansal M. Tech Project under the guidance of Prof. Deepak B. Phatak Computer Science & Engineering Indian Institute of Technology, Bombay
Oct 31, 2013
Adaptive Recommendation System for MOOC
1
Outline
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Outline
Adaptive Recommendation System for MOOC
2
Outline
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Outline • Introduction to learning in MOOC system.
Adaptive Recommendation System for MOOC
2
Outline
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Outline • Introduction to learning in MOOC system. • Components of adaptive hypermedia
Adaptive Recommendation System for MOOC
2
Outline
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Outline • Introduction to learning in MOOC system. • Components of adaptive hypermedia • Learner modeling
Adaptive Recommendation System for MOOC
2
Outline
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Outline • Introduction to learning in MOOC system. • Components of adaptive hypermedia • Learner modeling • Knowledge representation
Adaptive Recommendation System for MOOC
2
Outline
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Outline • Introduction to learning in MOOC system. • Components of adaptive hypermedia • Learner modeling • Knowledge representation • Proposed system
Adaptive Recommendation System for MOOC
2
Outline
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Outline • Introduction to learning in MOOC system. • Components of adaptive hypermedia • Learner modeling • Knowledge representation • Proposed system • Implementation status
Adaptive Recommendation System for MOOC
2
Outline
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Outline • Introduction to learning in MOOC system. • Components of adaptive hypermedia • Learner modeling • Knowledge representation • Proposed system • Implementation status • Plan for stage 2
Adaptive Recommendation System for MOOC
2
Outline
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Outline • Introduction to learning in MOOC system. • Components of adaptive hypermedia • Learner modeling • Knowledge representation • Proposed system • Implementation status • Plan for stage 2 • References
Adaptive Recommendation System for MOOC
2
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Introduction Learning work-flow in MOOC system
Introduction
Adaptive Recommendation System for MOOC
3
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Introduction “Learning workflow in Improvised MOOC system”
Introduction
Adaptive Recommendation System for MOOC
4
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Introduction “Adaptive Hypermedia”
Adaptive Hypermedia: Collection of all techniques which can be used to enable adaptation in a web based application [Bru01].
Introduction
Adaptive Recommendation System for MOOC
5
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Introduction “Adaptive Hypermedia”
Adaptive Hypermedia: Collection of all techniques which can be used to enable adaptation in a web based application [Bru01].
Figure: classification of adaptation technologies [Bru94b]
Introduction
Adaptive Recommendation System for MOOC
5
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Introduction “Adaptive Hypermedia”
Introduction
Components of Adaptive System:
Figure: components of adaptive system [Bru94a]
Adaptive Recommendation System for MOOC
6
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Introduction
Some important conferences/workshops/journals • UMAP: Conference on user modeling,Adaptation, and
Personalization (last-2012,Canada) • Hypertext: ACM conference hypertext and social media
(last-2011,Netherlands) • PING: Workshop on personalization in e-learning (last-2007,Greece) • UM: International conference on User Modeling (last-2003,USA) • International workshop on authoring of adaptive and adaptable
hypermedia (last-2011,USA) • International workshop on Information Heterogeneity and fusion in
recommender system(last-2010,Spain)
Introduction
Adaptive Recommendation System for MOOC
7
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Introduction
Some of the most active people working in this field • Peter Brusilousky (school of information science, university of
Pittsburgh, home- www.sis.pitt.edu/ peterb) • Konstantina Chrysafiadi (Dept. of informatics, university of
Piraeus,
[email protected]) • Maria virou(Dept. of informatics, university of Piraeus,
[email protected]) • Alenka kavcic (Faculty of computer and Information science,
University of Ljubljana)
Introduction
Adaptive Recommendation System for MOOC
8
Learner Modeling
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
What is Learner Modeling?
Learner Modeling
• System’s understanding about the user
Adaptive Recommendation System for MOOC
10
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
What is Learner Modeling?
Learner Modeling
• System’s understanding about the user • Learner profile v/s Learner model
Adaptive Recommendation System for MOOC
10
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
What is Learner Modeling?
Learner Modeling
• System’s understanding about the user • Learner profile v/s Learner model • Two types of information is stored in Learner model
Adaptive Recommendation System for MOOC
10
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
What is Learner Modeling? • System’s understanding about the user • Learner profile v/s Learner model • Two types of information is stored in Learner model
Learner Modeling
Adaptive Recommendation System for MOOC
10
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Classification of Learner models
Figure: classification of learner models [Bru01]
Learner Modeling
Adaptive Recommendation System for MOOC
11
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Existing Approaches for Learner Modeling • Bayesion Belief Network • Machine Learning techniques • Neural network techniques • Fuzzy clustering techniques • Neuro Fuzzy techniques
Learner Modeling
Adaptive Recommendation System for MOOC
12
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Bayesion Belief Network Bayesian networks are directed graphs that represents a set of random variables and their conditional dependencies [Ngu08][GSMBFPK10]. nodes represents random variables, edges represents conditional probabilities. • Knowledge: set of learning objectives • Actions: depends on learning objective • Passive • Individual active • Collective active • Evidence: • Utility: U = w1 .E1 + w2 .E2 + w3 .E3 + w4 .E4 ........ + wn .En
Learner Modeling
Adaptive Recommendation System for MOOC
13
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Machine Learning approach
Learner Modeling
Adaptive Recommendation System for MOOC
14
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Machine Learning approach
Learner Modeling
• A user always leaves some patterns while interacting with
Hypermedia system.
Adaptive Recommendation System for MOOC
14
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Machine Learning approach
Learner Modeling
• A user always leaves some patterns while interacting with
Hypermedia system. • System Learns about users’ interests, habits and preferences, based
on interaction.
Adaptive Recommendation System for MOOC
14
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Machine Learning approach
Learner Modeling
• A user always leaves some patterns while interacting with
Hypermedia system. • System Learns about users’ interests, habits and preferences, based
on interaction. • Constructs a behavior oriented learner model.
Adaptive Recommendation System for MOOC
14
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Machine Learning approach
Learner Modeling
• A user always leaves some patterns while interacting with
Hypermedia system. • System Learns about users’ interests, habits and preferences, based
on interaction. • Constructs a behavior oriented learner model. • Used extensively in e-commerce to understand the behavior of users.
Adaptive Recommendation System for MOOC
14
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Machine Learning approach
Learner Modeling
• A user always leaves some patterns while interacting with
Hypermedia system. • System Learns about users’ interests, habits and preferences, based
on interaction. • Constructs a behavior oriented learner model. • Used extensively in e-commerce to understand the behavior of users. • can also be used for learner classification and plan recognition in
adaptive systems.
Adaptive Recommendation System for MOOC
14
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Limitations of Machine Learning approach • Required large amount of data e.g. click streams, search logs etc
[Bau96]. • Complexity of this approach is high (it can be managed but we can
achieve same system with low complexity )
Learner Modeling
Adaptive Recommendation System for MOOC
15
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Neural network Method • Infer a meaningful pattern from a set of large and imprecise data [Fin96]. • provides methods to model human behavior. • J. Beck and B. Woolf attempts to model the learner by a “two
phase learning algorithm” • training phase exploits data from all the users to learn all types of
stereotypes. • learning phase learns about a given learner on the basis of the data
stored in the system.
Learner Modeling
Adaptive Recommendation System for MOOC
16
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Fuzzy Logic Method • Fuzzy clustering • Neuro Fuzzy representation of domain
Learner Modeling
Adaptive Recommendation System for MOOC
17
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Fuzzy Logic Method • Fuzzy clustering • Neuro Fuzzy representation of domain • Clustering is the classification of a data point into different
stereotypic clusters. • Hard clustering (non-fuzzy clustering) • Soft clustering (fuzzy clustering)
Learner Modeling
Adaptive Recommendation System for MOOC
17
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Fuzzy Logic Method • Fuzzy clustering • Neuro Fuzzy representation of domain • Clustering is the classification of a data point into different
stereotypic clusters. • Hard clustering (non-fuzzy clustering) • Soft clustering (fuzzy clustering)
• A membership function is associated with every data point which
specifies the degree of belonging of a particular data point to a given cluster.
Learner Modeling
Adaptive Recommendation System for MOOC
17
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Fuzzy Logic Method • Fuzzy clustering • Neuro Fuzzy representation of domain • Clustering is the classification of a data point into different
stereotypic clusters. • Hard clustering (non-fuzzy clustering) • Soft clustering (fuzzy clustering)
• A membership function is associated with every data point which
specifies the degree of belonging of a particular data point to a given cluster. • Fuzzy classification technique fuzzifies the classification of learners
in to different classes.
Learner Modeling
Adaptive Recommendation System for MOOC
17
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Comparative Analysis Parameters for comparison
Learner Modeling
Adaptive Recommendation System for MOOC
18
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Comparative Analysis Parameters for comparison • Computational complexity
Learner Modeling
Adaptive Recommendation System for MOOC
18
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Comparative Analysis Parameters for comparison • Computational complexity • Dynamic modeling
Learner Modeling
Adaptive Recommendation System for MOOC
18
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Comparative Analysis Parameters for comparison • Computational complexity • Dynamic modeling • Labeled/Unlabeled
Learner Modeling
Adaptive Recommendation System for MOOC
18
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Comparative Analysis Parameters for comparison
Learner Modeling
• Computational complexity • Dynamic modeling • Labeled/Unlabeled • Size of training data
Adaptive Recommendation System for MOOC
18
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Comparative Analysis Parameters for comparison
Learner Modeling
• Computational complexity • Dynamic modeling • Labeled/Unlabeled • Size of training data • Uncertainty
Adaptive Recommendation System for MOOC
18
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Comparative Analysis Parameters for comparison
Learner Modeling
• Computational complexity • Dynamic modeling • Labeled/Unlabeled • Size of training data • Uncertainty • Noisy data
Adaptive Recommendation System for MOOC
18
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling Comparative Analysis Technique
Fuzzy Logic Neural Networks Fuzzy clustering NeuroFuzzy
Compl- Dynamic Labeled/ exity modelUnlaing beled Med Yes N/A High Yes Both
Size of training data N/A High
Uncertanity Yes Yes
High/ Med High
No
Both
Med/High
Yes
Yes
Labeled
Med/High
Yes
Table: Characteristics of different techniques for user modeling [KY95]
Learner Modeling
Adaptive Recommendation System for MOOC
19
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Inferences Task Prediction Recommendation Classification
Needed NeuroFuzzy NeuroFuzzy Fuzzy Logic Neuro Fuzzy
Filtering
Fuzzy Logic
Not Needed Neural Networks Neural Networks, Fuzzy clustering Neural Networks Fuzzy Clustering Neural Networks
Table: Techniques recommended for adaptation tasks [FMMCM04]
Learner Modeling
Adaptive Recommendation System for MOOC
20
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Learner Modeling
Available adaptive systems and their charaterstics Adaptive System ELM-ART AHA! Hyperadapter Netcoach Interbook
based on knowledge YES YES YES YES YES
based on preferences YES YES -
Table: Adaptive systems based on learners characterstics [GYZ11]
Learner Modeling
Adaptive Recommendation System for MOOC
21
Knowledge Representation
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Knowledge Representation • To achieve adaptation, the domain knowledge should be represented
in a manner that there should be a mapping between the adaptive system and the human expert [SKPR05].
Knowledge Representation
Adaptive Recommendation System for MOOC
23
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Knowledge Representation • To achieve adaptation, the domain knowledge should be represented
in a manner that there should be a mapping between the adaptive system and the human expert [SKPR05]. • knowledge representation must be specific to the learner modeling
technique
Knowledge Representation
Adaptive Recommendation System for MOOC
23
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Knowledge Representation • To achieve adaptation, the domain knowledge should be represented
in a manner that there should be a mapping between the adaptive system and the human expert [SKPR05]. • knowledge representation must be specific to the learner modeling
technique • knowledge can be represented in various ways
Knowledge Representation
Adaptive Recommendation System for MOOC
23
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Knowledge Representation • To achieve adaptation, the domain knowledge should be represented
in a manner that there should be a mapping between the adaptive system and the human expert [SKPR05]. • knowledge representation must be specific to the learner modeling
technique • knowledge can be represented in various ways • Initial thinking about the system • Hierarchical knowledge representation • Goal oriented connectionist knowledge representation • Fuzzy cognitive Maps(FCM)
Knowledge Representation
Adaptive Recommendation System for MOOC
23
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Initial thinking about the system • Domain is the set of concepts
Di = {c1 , c2 , c3 ......cn }
Knowledge Representation
Adaptive Recommendation System for MOOC
24
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Initial thinking about the system • Domain is the set of concepts
Di = {c1 , c2 , c3 ......cn } • Recommendation will be based on the score/concept.
Knowledge Representation
Adaptive Recommendation System for MOOC
24
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Initial thinking about the system • Domain is the set of concepts
Di = {c1 , c2 , c3 ......cn } • Recommendation will be based on the score/concept. • Initialization System will be initialized on the basis of quiz in which
there is a one to one or many relationship among the concepts and questions
Knowledge Representation
Adaptive Recommendation System for MOOC
24
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Limitations • concepts are dependent c1 = “Find the sum in a for loop” c2 = “Find the average in a for loop”
Knowledge Representation
Adaptive Recommendation System for MOOC
25
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Limitations • concepts are dependent c1 = “Find the sum in a for loop” c2 = “Find the average in a for loop” • Recommendations should be such that : • Weak learner gets incremental suggestions based on the complexity. • System should infer/skip some concepts for smart learners.
Knowledge Representation
Adaptive Recommendation System for MOOC
25
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Hierarchical knowledge representation M. Siddappa and A. S. Manjunath tried to conduct a course on “Computer Architecture” based on hierarchical representation of domain .
Figure: hierarchy based representation of domain [SM07]
Knowledge Representation
Adaptive Recommendation System for MOOC
26
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Advantage • gives the order of presentation of concepts in increasing order of
complexity • only part − of /pre − requisite relationship can be represented
efficiently
Knowledge Representation
Adaptive Recommendation System for MOOC
27
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation
Advantage • gives the order of presentation of concepts in increasing order of
complexity • only part − of /pre − requisite relationship can be represented
efficiently Limitations • other types of relationships such as related relationships among the concepts cannot be represented. • Increasing the knowledge levels of one concept can only affect the
knowledge levels of concepts in hierarchy [Mae94].
Knowledge Representation
Adaptive Recommendation System for MOOC
27
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation “Fuzzy Cognitive Maps (FCM)”
Fuzzy Cognitive Maps • In all the previous approaches, if the knowledge level of a particular
concept reaches above a threshold value then the concept will be considered as learned. • Knowledge level is not discrete e.g. “He is intelligent”,“He knows
this concept 50%” • Fuzzy logic fuzzifies the classification of users’ knowledge levels in
classes • Each point in the Fuzzy logic is associated with a membership
function which tells the degree of membership of this point to the set [Fin96].
Knowledge Representation
Adaptive Recommendation System for MOOC
28
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation “Fuzzy Cognitive Maps (FCM)”
Fuzzy Cognitive Maps • A = φ : ∀xX : µA (x) = 0 • Represent the domain in the form of graph where each node
represents a domain concept [Kav04]. • The relationship between two nodes can be of two types • essential pre-requisites(E) • supportive pre-requisites(S) • R =E ∪S
E ∩S =φ • E ⊆C ×C
µE : C n × C → [0, 1] o µE (ci ,cj ) E= : (c , c )C i j (ci ,cj )
Knowledge Representation
Adaptive Recommendation System for MOOC
29
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation “Fuzzy Cognitive Maps (FCM)”
Fuzzy Cognitive Maps • Domain can be represented by a triplet :
GD = (C .E .S) C: set of concept nodes, E: set of essential prerequisites, S: set of supportive prerequisite • Each node is associated with a knowledge level ‘l’. • Membership functions are defined which value of fuzzifies the
classification of learner on the basis of ‘l’. • Learner knowledge about a particular concept c is therefore
expressed by providing the values of three fuzzy sets [SKPR05] (µU , µK , µL ) µU + µK + µL = 1 µU > 0 =⇒ µL = 0 µL > 0 =⇒ µU = 0 Knowledge Representation
Adaptive Recommendation System for MOOC
30
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Knowledge Representation “Fuzzy Cognitive Maps (FCM)”
Figure: Membership Functions for fuzzy sets of unknown (CU),known(CK), and learned (CL) [VS01] Knowledge Representation
Adaptive Recommendation System for MOOC
31
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Recommendation system in MOOC
Requirements:
Proposed system
Adaptive Recommendation System for MOOC
32
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Recommendation system in MOOC
Requirements: • Adaptive
Proposed system
Adaptive Recommendation System for MOOC
32
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Recommendation system in MOOC
Requirements: • Adaptive • Dynamic updation
Proposed system
Adaptive Recommendation System for MOOC
32
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Recommendation system in MOOC
Requirements: • Adaptive • Dynamic updation • low computational complexity
Proposed system
Adaptive Recommendation System for MOOC
32
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Recommendation system in MOOC
Requirements: • Adaptive • Dynamic updation • low computational complexity • Domain independent (e.g. should not need any expert to assign
weightage scores on the edges)
Proposed system
Adaptive Recommendation System for MOOC
32
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Recommendation system in MOOC
Input to the system
Proposed system
Adaptive Recommendation System for MOOC
33
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Recommendation system in MOOC “4-layered architecture”
Proposed system
Figure: A 4-layered architecture for Knowledge representation
Adaptive Recommendation System for MOOC
34
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Recommendation system in MOOC Relationship among the concept modules
Proposed system
• hierarchical relationship • non-hierarchical relationship
Adaptive Recommendation System for MOOC
35
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Recommendation system in MOOC
Example Module id C1 C2 C3 C4 C5 C6
Concept if-else statement switch statement for loop while loop do-while loop Using conditional statements with iterative statements Table: An example lecture
Proposed system
Adaptive Recommendation System for MOOC
36
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Example Concept id C1 C1 C1 C1
Objective id objp obj1 obj2 obj3
C1
obj4
C2 C2 C2 C2
objp obj1 obj2 obj3
C2
obj4
Proposed system
Description syntax of if-else statement check a given number is positive or negative check a given number is even or odd given two numbers, check which one is greater given three numbers, check which one is greater syntax of switch statement check a given number is positive or negative check a given number is even or odd given two numbers, check which one is greater given three numbers, check which one is greater
Adaptive Recommendation System for MOOC
37
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Example Concept id C3 C3 C3 C3 C3 C3 C4 C4 C4 C4 C4 C4
Proposed system
Objective id objp obj1 obj2 obj3 obj4 obj5 objp obj1 obj2 obj3 obj4 obj5
Description syntax of the for statement print the first n integers in a for loop counting in a for loop calculating sum in a for loop calculating average in a for loop calculating max/min in a for loop syntax of the while statement print the first n integers in a while loop counting in a while loop calculating sum in a while loop calculating average in a while loop calculating max/min in a while loop
Table: An example of concept to objective table Adaptive Recommendation System for MOOC
38
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Example
Proposed system
C5 C5 C5 C5 C5 C5 C6
objp obj1 obj2 obj3 obj4 obj5 obj1
C6
obj2
C6
obj3
C6
obj4
syntax of the do-while statement print the first n integers in a do-while loop counting in a do-while loop calculating sum in a do-while loop calculating average in a do-while loop calculating max/min in a do-while loop print all the even numbers between 1 to 100 using print all the odd numbers divisible by 7, between 1 to 100 print all the numbers divisible by 3 or 5, between 1 to 100 print all the odd numbers divisible 3 or 5 but not 3 and 5, between 1 to 100
Table: An example of concept to objective table Adaptive Recommendation System for MOOC
39
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC
Proposed system
Adaptive Recommendation System for MOOC
40
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC
Proposed system
Adaptive Recommendation System for MOOC
41
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC
Proposed system
Adaptive Recommendation System for MOOC
42
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC
Proposed system
Adaptive Recommendation System for MOOC
43
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC
Proposed system
Adaptive Recommendation System for MOOC
44
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC
State diagram of an objective
Proposed system
Adaptive Recommendation System for MOOC
45
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Modeling learner • Each learner’s knowledge for a particular concept can be represented as four tuples (Un, Uk, K , L) .The membership functions for these fuzzy sets depends on knowledge level ‘x’. Membership functions for Un and Uk 1, µUn (x) = 1 − (x − 55)/5, 0, (x − 55)/5, 1, µUK (x) = 1 − (x − 70)/5, 0, 0,
Proposed system
x ≤ 55 55 < x < 60 x ≥ 60 55 < x < 60 60 ≤ x ≤ 70 70 < x < 75 x < 55 x ≥ 75
Adaptive Recommendation System for MOOC
46
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Modeling learner Membership functions for K and L (x − 70)/5, 70 < x < 75 75 ≤ x ≤ 85 1, µK (x) = 1 − (x − 85)/5, 85 < x < 90 0, x ≤ 70 0, x ≥ 90 (x − 85)/5, 85 < x < 90 µL (x) = 1, 90 ≤ x ≤ 100 0, x ≤ 85
Proposed system
Adaptive Recommendation System for MOOC
47
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC
Three steps to achieve adaptation • Initialization • Update the systems assumptions about the learner. • Prioritize the concept modules that needs to be recommended.
Proposed system
Adaptive Recommendation System for MOOC
48
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Initialization • Initialization of the system refers to capturing initial data about the user e.g. user preferences, learning style, knowledge level about all the concept taught in the week.
Proposed system
Adaptive Recommendation System for MOOC
49
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Initialization • Initialization of the system refers to capturing initial data about the user e.g. user preferences, learning style, knowledge level about all the concept taught in the week. Number of objectives achieved for Ci • Knowledge level(Ci ) = Total number of objectives in Ci
Proposed system
Adaptive Recommendation System for MOOC
49
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Initialization • Initialization of the system refers to capturing initial data about the user e.g. user preferences, learning style, knowledge level about all the concept taught in the week. Number of objectives achieved for Ci • Knowledge level(Ci ) = Total number of objectives in Ci • Example
Figure: Initialization of concept nodes Proposed system
Adaptive Recommendation System for MOOC
49
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Prioritizing concept module- Approach 1 Select the concept module According to Topological sort.
Figure: Initialization of concept nodes
Proposed system
Adaptive Recommendation System for MOOC
50
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Prioritizing concept module- Approach 1 Select the concept module According to Topological sort.
Figure: Initialization of concept nodes • Advantage- Select those concept modules which are affecting other
concept modules i.e. making them ready.
Proposed system
Adaptive Recommendation System for MOOC
50
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Prioritizing concept module- Approach 1 Select the concept module According to Topological sort.
Figure: Initialization of concept nodes • Advantage- Select those concept modules which are affecting other
concept modules i.e. making them ready. • Limitation- Knowledge level of the concept as well as distance is
not considered i.e. we may select a concept which is already in learnt state but influencing other concept modules. Proposed system
Adaptive Recommendation System for MOOC
50
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Prioritizing concept module- Approach 2 Select the concept module having minimum distance.
Proposed system
Figure: Initialization of concept nodes
Adaptive Recommendation System for MOOC
51
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Prioritizing concept module- Approach 2 Select the concept module having minimum distance.
Proposed system
Figure: Initialization of concept nodes • Advantage- Minimum number of concepts to be covered in order to
learn the concept.
Adaptive Recommendation System for MOOC
51
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Prioritizing concept module- Approach 2 Select the concept module having minimum distance.
Proposed system
Figure: Initialization of concept nodes • Advantage- Minimum number of concepts to be covered in order to
learn the concept. • Limitation- Concept module having minimum distance may have
knowledge level as learned Adaptive Recommendation System for MOOC
51
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Proposed System for Adaptive Recommendations in MOOC Prioritizing concept module- Approach 3 unknown → unsatisfactoryKnown → known → learnt
Prioritizing concept module- Approach 4 unsatisfactoryKnown → known → learnt and then unknown
Proposed system
Adaptive Recommendation System for MOOC
52
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Uniqueness of Proposed System How this approach is different then existing approaches
Proposed system
Adaptive Recommendation System for MOOC
53
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Uniqueness of Proposed System How this approach is different then existing approaches
Proposed system
• Domain independent (Experts are not required for calculating the
dependency of one concept on other)
Adaptive Recommendation System for MOOC
53
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Uniqueness of Proposed System How this approach is different then existing approaches
Proposed system
• Domain independent (Experts are not required for calculating the
dependency of one concept on other) • Data set is not required for learning
Adaptive Recommendation System for MOOC
53
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Uniqueness of Proposed System How this approach is different then existing approaches
Proposed system
• Domain independent (Experts are not required for calculating the
dependency of one concept on other) • Data set is not required for learning • Low complexity and run time updation is possible
Adaptive Recommendation System for MOOC
53
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Evaluation
Evaluation
All the possible approaches needs to be compared after implementation • In ITS community, the common practice of evaluation is Empirical
approaches. • The criteria for the evaluation of knowledge representation technique
is the mean number of times that a leaner is advised to read a domain concept, until is considered as learned. • There are statistical approaches e.g. “Independent sample T-test • Levene’s test is used according to which if the value of “Sig”
variable is higher than 0.05, then two variance are approximately equal otherwise the difference between the means are statistically significant.
Adaptive Recommendation System for MOOC
54
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
User Interfaces
Implementation
Adaptive Recommendation System for MOOC
55
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
User Interfaces
Implementation
Adaptive Recommendation System for MOOC
56
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Use-case diagram
Implementation
Adaptive Recommendation System for MOOC
57
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Plan for stage 2
Looking ahead
Goals: • Literature survey. • Find out the problems in the existing approaches.
Plan for Stage 2
Adaptive Recommendation System for MOOC
58
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Plan for stage 2
Looking ahead
Goals: • Capture requirements. • SRS document. • Implementation plan
Plan for Stage 2
Adaptive Recommendation System for MOOC
59
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Plan for stage 2
Looking ahead
Goals: • Implementation
Plan for Stage 2
Adaptive Recommendation System for MOOC
60
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Plan for stage 2
Looking ahead
Goals: • Implementation
Plan for Stage 2
Adaptive Recommendation System for MOOC
61
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Plan for stage 2
Looking ahead
Goals: • Make test plan. • Testing
Plan for Stage 2
Adaptive Recommendation System for MOOC
62
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
Plan for stage 2
Looking ahead
Goals: • Resolve Issues. • Deployment • Write paper for UMAP 2014
Plan for Stage 2
Adaptive Recommendation System for MOOC
63
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
References I
Plan for Stage 2
Mathias Bauer. Machine learning for user modeling and plan recognition. In Proc. ICML’96 Workshop "Machine Learning meets Human Computer Interaction" 5–16, pages 5–16, 1996. Peter Brusilovsky. The construction and application of student models in intelligent tutoring systems. Computer and System Sciences International, 32(1):70–89, July 1994. Peter Brusilovsky. Student modelling and adaptivity in web based learning systems. Computer and System Sciences International, 32(1):7089, July 1994. Peter Brusilovsky. Adaptive hypermedia. User Modeling and User-Adapted Interaction, 11(1-2):87–110, March 2001. Janet Finlay. Machine learning: a tool to support improved usability? In Proc. ICML’96 Workshop "Machine Learning meets Human-Computer Interaction, pages 17–28, 1996. Enrique Fras-Martnez, George Magoulas, Sherry Chen, and Robert Macredie. Recent soft computing approaches to user modeling in adaptive hypermedia. In PaulM.E. Bra and Wolfgang Nejdl, editors, Adaptive Hypermedia and Adaptive Web-Based Systems, volume 3137 of Lecture Notes in Computer Science, pages 104–114. Springer Berlin Heidelberg, 2004. Sergio Gutierrez-Santos, Jaime Mayor-Berzal, Carmen Fernandez-Panadero, and Carlos Delgado Kloos. Authoring of probabilistic sequencing in adaptive hypermedia with bayesian networks. 16(19):2801–2820, oct 2010.
Adaptive Recommendation System for MOOC
64
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
References II
Plan for Stage 2
M.A. Ghazal, M.M. Yusof, and N.A.M. Zin. Adaptive educational hypermedia system using cognitive style approach: Challenges and opportunities. In Electrical Engineering and Informatics (ICEEI), 2011 International Conference on, pages 1–6, 2011. A. Kavcic. Fuzzy user modeling for adaptation in educational hypermedia. Systems, Man, and Cybernetics, Part C: Applications and Reviews, IEEE Transactions on, 34(4):439–449, 2004. George J. Klir and Bo Yuan. Fuzzy sets and fuzzy logic: theory and applications. Prentice-Hall, Inc., Upper Saddle River, NJ, USA, 1995. Pattie Maes. Agents that reduce work and information overload. Commun. ACM, 37(7):30–40, July 1994. Viet Anh Nguyen. Constructing a bayesian belief network to generate learning path in adaptive hypermedia system. Computer Science and Cybernetics, 24(1):12–19, 2008. Wojciech Stach, Lukasz Kurgan, Witold Pedrycz, and Marek Reformat. Genetic learning of fuzzy cognitive maps. Fuzzy Sets and Systems, 153(3):371 – 401, 2005. M. Siddappa and A. S. Manjunath. Knowledge representation using multilevel hierarchical model in intelligent tutoring system. In Proceedings of the third conference on IASTED International Conference: Advances in Computer Science and Technology, ACST’07, pages 323–329, Anaheim, CA, USA, 2007. ACTA Press.
Adaptive Recommendation System for MOOC
65
Outline Introduction Learner Modeling Knowledge Representation Proposed system Evaluation Implementation Plan for Stage 2
References III
Plan for Stage 2
S. Vrettos and A. Stafylopatis. A fuzzy rule-based agent for web retrieval-filtering. In Ning Zhong, Yiju Yao, Jiming Liu, and Setsuo Ohsuga, editors, Web Intelligence: Research and Development, volume 2198 of Lecture Notes in Computer Science, pages 448–453. Springer Berlin Heidelberg, 2001.
Adaptive Recommendation System for MOOC
66