Complete project work PDF and DOC download titled Enhancing Prediction Accuracy Of A Multi-Criteria Recommender System Using Adaptive Genetic Algorithm
Enhancing Prediction Accuracy Of A Multi-Criteria Recommender System Using Adaptive Genetic Algorithm
TABLE OF CONTENTS
CERTIFICATION ………………………………………………………………………………………………..ii
ABSTRACT ……………………………………………………………………………………………………… v
ACKNOWLEDGEMENTS ……………………………………………………………………………………vi
DEDICATION …………………………………………………………………………………………………. viii
LIST OF FIGURES ……………………………………………………………………………………………. xii
LIST OF TABLES …………………………………………………………………………………………….. xiii
CHAPTER ONE: INTRODUCTION ……………………………………………………………………….. 1
1.1 Introduction……………………………………………………………………………………………….. 1
1.2 Background of the study ……………………………………………………………………………. 1
1.2.1 Recommender system techniques …………………………………………………………… 2
1.2.2 Multi-criteria recommender system ………………………………………………………… 3
1.2.3 Genetic algorithm ………………………………………………………………………………. 3
1.3 Statement of the problem …………………………………………………………………………… 5
1.4 Aim and objectives of the study ………………………………………………………………….. 5
1.5 Significance of the study …………………………………………………………………………… 5
1.6 Scope of the study …………………………………………………………………………………… 6
1.7 Expected results ……………………………………………………………………………………… 6
1.8 Thesis structure ………………………………………………………………………………………. 6
CHAPTER TWO: LITERATURE REVIEW ……………………………………………………………… 7
2.1 Introduction …………………………………………………………………………………………… 7
2.2 Overview of Recommender Systems and their applications ………………………………… 7
2.3 Ratings …………………………………………………………………………………………………. 9
2.4 Types of rating ……………………………………………………………………………………….. 9
2.5 Measure of accuracy ………………………………………………………………………………. 10
2.6 Overview of collaborative filtering …………………………………………………………….. 11
2.6.1 Types of collaborative filtering ……………………………………………………………. 11
2.6.2 Application of collaborative filtering …………………………………………………….. 14
2.7 Genetic Algorithm“………………………………………………………………………………… 15
2.7.1 Initial population ……………………………………………………………………………… 16
2.7.2 Fitness evaluation …………………………………………………………………………….. 16
x
2.7.3 Selection ……………………………………………………………………………………….. 17
2.7.4 Crossover ………………………………………………………………………………………. 17
2.7.5 Mutation ……………………………………………………………………………………………. 18
2.7.6 Termination ……………………………………………………………………………………. 18
2.8 Cold start problem …………………………………………………………………………………. 19
2.9 Multi-criteria recommender systems …………………………………………………………… 19
2.9.1 Similarity-based approach ………………………………………………………………….. 20
2.9.2 Aggregation function-based approach …………………………………………………… 21
2.10 Review of related studies………………………………………………………………………. 22
2.10.1 New recommendation techniques for multi-criteria rating systems ……………….. 22
2.10.2 An item-based multi-criteria collaborative filtering algorithm for personalized recommender systems ………………………………………………………………………………….. 23
2.10.3 Multi-criteria collaborative filtering with high accuracy, using higher-order singular-value decomposition and a neuro-fuzzy system ……………………………………….. 24
2.10.4 Accuracy improvement for multi-criteria recommender systems ………………….. 26
2.10.5 Evaluation of recommender systems: A multi-criteria decision-making approach 26
2.10.6 Using genetic algorithm for measuring similarity values between users in collaborative filtering recommender systems ……………………………………………………… 27
2.10.7 Improving collaborative filtering recommender system results and performance, using genetic algorithm ………………………………………………………………………………… 28
2.10.8 Our solution …………………………………………………………………………………… 29
CHAPTER THREE: RESEARCH METHODOLOGy ………………………………………………… 30
3.1 Introduction……………………………………………………………………………………………… 30
3.2 Multi-criteria recommender system ……………………………………………………………. 30
3.3 Data set description ………………………………………………………………………………… 31
3.4 Choice of programming language ………………………………………………………………. 34
3.5 Proposed system ……………………………………………………………………………………. 34
3.5.1 Predicting N multi-criteria ratings ………………………………………………………… 35
3.5.2 Asymmetric singular-value decomposition (ASVD)………………………………….. 35
3.5.3 Learning the function………………………………………………………………………… 36
3.5.4 Predicting the overall rating ……………………………………………………………….. 40
xi
CHAPTER FOUR: IMPLEMENTATION ……………………………………………………………….. 44
4.1 Introduction …………………………………………………………………………………………. 44
4.2 Performance evaluation …………………………………………………………………………… 44
4.3 Result and discussion ……………………………………………………………………………… 46
4.4 Conclusion …………………………………………………………………………………………… 49
CHAPTER FIVE: SUMMARY, CONCLUSION AND RECOMMENDATION ……………….. 50
5.1 Introduction……………………………………………………………………………………………… 50
5.2 Summary and contributions ……………………………………………………………………… 50
5.3 Conclusion …………………………………………………………………………………………… 51
5.4 Recommendation and future work ……………………………………………………………… 52
REFERENCE …………………………………………………………………………………………………… 53
CHAPTER ONE
1.1 Introduction
Intelligent systems are systems that require knowledge organisation to interpret, test and analyse acquired information. Intelligent systems are required in most of our day-to-day activities, such as e-commerce, online-booking, social media, e-shopping and other information-rich environments. Recommender systems interact with users in a personalized way, obtain information about a user’s tastes or preferences and use this knowledge to make suggestions and provide assistance in situations where users have to make a decision between a wide range of possible options. In this chapter we endeavour to explain the recommender system and its techniques, introduce multi-criteria recommender systems and also a genetic algorithm. Statement of the problem, aims and objectives, significance and scope of this study will also be introduced in this chapter.
1.2 Background of the study
The recommender system was identified as a free research area in the mid-1990s, when researchers started concentrating on recommendation glitches that obviously depend on rating structure (Adomavicius & Tuzhilin, 2005). Recommender systems (RSs) are techniques and software tools for interacting with large and complex information spaces in order to prioritize and make suggestions on items, offers and objects likely to be of interest to a specific user (Ricci, Rokach & Shapira, 2015). These suggestions relay to several decision-making procedures, for example what item or object to buy, which movie to watch, what news to read online, what music to listen to, which airline to fly with or hotel to book (Ricci et al., 2015). Thus, the diversity in the feature of homogeneous products or services, related information and the choices available in the market place or in diverse application domains such as e-commerce, e-learning, e-government and e-tourism has made the recommender system broadly utilized (Shambour, Hourani & Fraihat, 2016).
Accuracy in the recommender system is a valuable factor in determining how it can effectively acquire and process information. This has made the evaluation of the recommender system a critical and challenging task. One major way of performing an evaluation of the recommender
2
system is through accuracy (Sohrabi, Toloo, Moeini & Nalchigar, 2015). This project is aimed at developing an adaptive genetic algorithm to enhance prediction accuracy and obtain high correlation between predicted and actual values of a multi-criteria recommender system.
1.2.1 Recommender system techniques
The recommender system can vary in terms of the knowledge base, addressed domain, algorithm or technique used during development. According to Burke (2002), the recommender can be classified into six different approaches:
Content-based: In this approach the system learns from the user’s previous likes and interests, then recommends matching items to the user based on that knowledge. The content-based approach places reliance on the item features and thus a learning method is employed to determine the type of user profile that will be derived by the content-based recommender. The similarity of the items is dependent on the features associated with the items (Ricci et al., 2015).
Collaborative filtering: This approach is the most prominent, developed and widely implemented technique (Burke, 2002). It generates recommendation of items to the active user based on previous items liked by other users with similar preferences. Collaborative filtering is called people-to-people correlation because the similarity in preference of two users is dependent on the similarity in the rating history of the users (Ricci et al., 2015).
Demographic: The main aim of this approach is to categorize the user based on personal attributes and to recommend items based on a user’s demographic profile (Ricci et al., 2015). It may not require a history of user ratings.
Knowledge-based: This system recommends items based on a specific field of knowledge of how useful an item is to the user and how certain item features meet the needs and preferences of the user. The similarity measure can be interpreted as the utility of the recommendation.
Community-based: This approach provides recommendations on items to the user based on the preferences of the user’s friends. It is suggested that people tend to rely more on the
3
recommendations of friends than those of anonymous individuals with similar preferences (Ricci et al., 2015).
Hybrid recommender systems: This type of recommender system is a combination of two or more recommendation techniques mentioned above (Adomavicius & Tuzhilin, 2005). A hybrid system combining two techniques tries to use the advantages of one to solve the drawbacks of the other.
1.2.2 Multi-criteria recommender system
Traditionally, most RSs obtain an overall or general preference of a particular item by the user. In other words, it recommends items based on a single criterion rating by the users as the input information to be used by the RS algorithm to evaluate user preference opinions. In most cases, a single criterion rating could produce recommendations that do not meet the needs of the user because users can express their opinions based on some specific features of the item.
In contrast, multi-criteria RSs give users the opportunity to specify their preferences for an item based on multiple attributes (Ricci et al., 2015). Multi-criteria ratings provide additional information about preferences of the user regarding several important aspects or components of an item (Adomavicius & Kwon, 2007). The additional information on each user’s preferences will lead to more accurate recommendations and will improve the quality of recommendations.
In recent years, multi-criteria ratings have been adopted by several recommender systems, instead of the traditional single criterion ratings (Ricci et al., 2015). The aim of multi-criteria recommender systems is to take a step towards analysing and understanding users’ interests and choices in a more efficient and exquisite manner and providing the users with optimal solutions.
1.2.3 Genetic algorithm
In the 1950s and the 1960s, several computer scientists independently studied evolutionary systems with the idea that evolution could be used as an optimization tool for engineering problems. The idea was to evolve a population of candidate solutions to a given problem using operators inspired by natural genetic variation and natural selection (Mitchell, 2004).
4
A genetic algorithm is an evolutionary stochastic search method applied to optimization and learning. It is also a search algorithm based on the hypothesis for moving from one population of “chromosomes” to a new population by using a kind of natural selection (survival of the fittest) and natural genetics, which can be used to solve an optimization problem. Additionally, a genetic algorithm is an evolutionary approach to solving optimization problems such as sequencing, travelling, salesman problems and scheduling (Schmitt, 2001). In a genetic algorithm there are some important components that should be kept in consideration, such as:
Representation: The way individuals are defined, either in bit string, binary or real numbers.
Fitness function: Concerned with the measure of performance, which can be either minimized or maximized?
Population: This holds the representation of possible solutions.
Parent selection mechanism: Helps to distinguish individuals based on their quality, i.e., allowing the better individual to become parent of the next generation.
Variation operators: These operators create new individuals from old ones. They are divided into crossover (single point or two points), which is done on selected individuals to mix the generic information to get new individuals or offspring and mutation (flipping).
Selection mechanism: Is often called replacement and it is based on survival of the fittest.
Since genetic algorithm is stochastic and at most times guarantees no optimum solution, a suitable termination condition is required, which could be when the fitness evaluation reaches a given limit.
5
1.3 Statement of the problem
The majority of prevailing RSs uses an overall estimation of a user rating of an item or single criterion rating techniques to evaluate users’ opinions on experienced items. Since the suitability of the recommended item for a particular user may depend on several important aspects or attributes in the decision making of the user, the efficiency of the traditional single criterion rating can be deliberated to be limited and inaccurate, because it cannot justify for the various items’ attributes.
For this purpose, a multi-criteria recommendation which implements users’ ratings on multiple or various attributes of an item using aggregate function-based approach is proposed. The proposed technique acquires an appropriate learning relationship using an adaptive genetic algorithm to achieve a more accurate and efficient prediction.
1.4 Aim and objectives of the study
The aim of this project is to use an adaptive genetic algorithm to model a multi-criteria recommendation problem using an aggregation function-based approach to achieve a more accurate and efficient prediction.
The specific objectives were:
To formulate an adaptive genetic algorithm model.
To use an adaptive genetic algorithm to model multi-criteria recommendation problems.
To develop a system that will be proficient enough to recommend the most appropriate item to a user.
To compare the predictive performance of the multi-criteria recommender technique using an adaptive genetic algorithm with the traditional recommender approach.
1.5 Significance of the study
Web users, application domains such as e-commerce, e-learning, e-government, social networks and e-tourism are the main benefactors of this study due to the fact that the system further creates an easier, faster and more efficient decision-making strategy.
A user rating of an item with multiple attributes based on his or her personal interest can efficiently improve the prediction accuracy of the recommendation to other users.
6
1.6 Scope of the study
The scope of the study is on RSs, with emphasis on multi-criteria RSs that will generate a proficient, accurate prediction for users. This study entails the development of a sophisticated system capable enough to recommend the most appropriate suggestion or item to users based on their preferences.
1.7 Expected results
The project aims to provide the predictive performance of the proposed technique and compare it with that of existing methods. These performances include a decrease in prediction errors, increase in ranking accuracy and high correlation between predicted and actual values.
1.8 Thesis structure
The rest of this thesis is organised as follows: Chapter 2 presents an overview of the RS and multi-criteria RS, discusses the adaptive genetic algorithm and its component, and reviews related studies. Chapter 3 describes the methodologies and architecture of the study. Chapter 4 presents the detailed implementation of the system. It also discusses the results obtained. Chapter 5 wraps up by discussing the summary, conclusion and recommendation.
Do you the complete work from chapters one to five titled Enhancing Prediction Accuracy Of A Multi-Criteria Recommender System Using Adaptive Genetic Algorithm, Click below.
Do you need help? Talk to us right now: (+234) 08060082010, 08107932631 (Call/WhatsApp). Email: [email protected].IF YOU CAN'T FIND YOUR TOPIC, CLICK HERE TO HIRE A WRITER»
