Enhancing Prediction Accuracy Of A Multi-Criteria Recommender System Using Adaptive Genetic Algorithm
Need help with a related project topic or New topic? Send Us Your Topic
DOWNLOAD THE COMPLETE PROJECT MATERIAL
Enhancing Prediction Accuracy Of A Multi-Criteria Recommender System Using Adaptive Genetic Algorithm
Chapter one
1.1 Introduction.
Intelligent systems require knowledge organisation to interpret, evaluate, and analyse acquired data. Intelligent systems are required in the majority of our daily activities, including e-commerce, online booking, social media, e-shopping, and other information-rich environments. Recommender systems engage with users in a personalised manner, gathering information about a user’s tastes or preferences and using that knowledge to make suggestions and provide aid in instances when users must choose from a variety of possibilities.
In this chapter, we will discuss the recommender system and its approaches, as well as introduce multi-criteria recommender systems and a genetic algorithm. This chapter will also introduce the problem statement, aims and objectives, significance, and scope of the investigation.
1.2 Background of the Study
The recommender system was recognised as a potential study field in the mid-1990s, when researchers began focussing on recommendation flaws that were clearly dependent on rating structure (Adomavicius & Tuzhilin, 2005).
Recommender systems (RSs) are methodologies and software tools for engaging with huge and complicated information spaces in order to prioritise and give suggestions on products, offers, and objects that may be of interest to a specific user (Ricci, Rokach, & Shapira, 2015).
These ideas inform a variety of decision-making processes, such as which thing or object to purchase, which movie to watch, which news to read online, which music to listen to, which airline to fly with, or which hotel to book (Ricci et al., 2015).
Thus, the variety of homogeneous products or services, related information, and 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 widely used (Shambour, Hourani, & Fraihat, 2016).
The recommender system’s accuracy is an important component in determining how efficiently it can receive and analyse information. This has made evaluation of the recommender system an important and difficult undertaking.
According to Sohrabi, Toloo, Moeini, and Nalchigar (2015), accuracy is a key metric for evaluating recommender systems. This project aims to create an adaptive evolutionary algorithm to improve prediction accuracy and achieve a high correlation between anticipated and actual values for a multi-criteria recommender system.
1.2.1 Recommender System Techniques
The knowledge base, addressed domain, method, or technique employed during development can all influence the recommender system. Burke (2002) classifies the recommender into six approaches:
Content-based: In this strategy, the system learns from the user’s previous likes and interests and then offers appropriate goods to the user based on that knowledge.
The content-based approach relies on item features, hence a learning mechanism is used to establish the sort of user profile that will be generated by the content-based recommender. The similarities between items are determined by the attributes linked with them (Ricci et al., 2015).
Collaborative filtering: This is the most popular, developed, and widely used technique (Burke, 2002). It makes recommendations to the active user based on prior items enjoyed by other users who share similar likes.
Collaborative filtering is known as people-to-people correlation since the similarity in preferences between two users is determined by their rating histories (Ricci et al., 2015).
Demographic: The primary goal of this method is to categorise users based on personal characteristics and recommend things based on their demographic profile (Ricci et al., 2015). It is possible that no prior user ratings are required.
Knowledge-based: This method recommends items based on a certain field of knowledge, such as how beneficial an item is to the user and how certain item attributes suit the user’s requirements and preferences. The similarity metric can be viewed as the utility of the recommendation.
Community-based: This strategy recommends goods to the user based on the preferences of their peers. According to Ricci et al. (2015), consumers prioritise referrals from friends above anonymous persons with similar tastes.
Hybrid recommendation systems: This sort of recommender system combines two or more of the strategies discussed above (Adomavicius & Tuzhilin, 2005). A hybrid system that combines two methodologies attempts to leverage the benefits of one to mitigate the disadvantages of the other.
1.2.2 Multi-criteria recommendation system
Traditionally, most RSs get the user’s overall or general preference for a certain item. In other words, it recommends things based on a single criterion rating by users, which serves as input material for the RS algorithm to evaluate user preference opinions.
In most circumstances, a single criterion rating may result in suggestions that do not fulfil the user’s demands because people can voice their thoughts based on specific qualities of an item.
In contrast, multi-criteria RSs allow users to declare their preferences for an item based on a variety of features (Ricci et al., 2015). Multi-criteria ratings provide additional information about the user’s preferences for numerous significant elements or components of an item (Adomavicius and Kwon, 2007). The additional information about each user’s preferences will result in more accurate and higher-quality recommendations.
In recent years, numerous recommender systems have implemented multi-criteria ratings in place of traditional single-criteria ratings (Ricci et al., 2015). The goal of multi-criteria recommender systems is to take a step forward in evaluating and interpreting users’ interests and choices in a more efficient and exquisite manner, as well as delivering ideal answers.
1.2.3 Genetic algorithm
Several computer scientists explored evolving systems independently in the 1950s and 1960s, with the goal of using evolution as an optimisation technique for engineering challenges. The goal was to generate a population of potential solutions to a given problem utilising operators inspired by natural genetic diversity and selection (Mitchell, 2004).
4
A genetic algorithm is an evolutionary stochastic search approach used for optimisation and learning. It is also a search method based on the notion of shifting from one population of “chromosomes” to another utilising natural selection (survival of the fittest) and natural genetics to solve an optimisation problem.
Furthermore, a genetic algorithm is an evolutionary method for addressing optimisation issues such as sequencing, travel, salesman problems, and scheduling (Schmitt 2001). Some significant components to consider in a genetic algorithm include:
Individuals are represented in a variety of ways, including bit strings, binary codes, and actual numbers.
Fitness function: Is concerned with performance measures that can be minimised or maximised?
Population: This contains a representation of various solutions.
Parent selection mechanism: Assists in distinguishing people based on their quality, allowing the superior individual to become the parent of the following generation.
Variation operators: These operators generate new individuals from existing ones. They are classified as crossover (single or two points), which is performed on selected individuals to combine generic information to create new individuals or children, and mutation (flipping).
The selection mechanism, also known as replacement, is predicated on survival of the fittest.
Because evolutionary algorithms are stochastic and rarely guarantee an optimum solution, a suitable termination condition, such as when the fitness evaluation hits a certain limit, is required.
5
1.3 Statement of the Problem
The bulk of current RSs use an overall estimation of a user rating of an item or single criterion rating methodologies to assess users’ thoughts on previously encountered objects.
Because the suitability of a recommended item for a specific user may be determined by several important aspects or attributes in the user’s decision-making process, the traditional single criterion rating can be considered limited and inaccurate because it cannot justify the various items’ attributes.
For this reason, a multi-criteria recommendation is developed that incorporates users’ ratings on many or diverse aspects of an item using an aggregate function-based method. The suggested technique uses an adaptive evolutionary algorithm to gain an appropriate learning relationship, resulting in a more accurate and efficient prediction.
1.4 Goals and objectives of the study
The goal of this project is to apply an adaptive evolutionary algorithm to model a multi-criteria recommendation problem with an aggregation function-based method in order to achieve more accurate and efficient predictions.
The aims included developing an adaptable evolutionary algorithm and applying it to multi-criteria recommendation situations.
Create a system capable of recommending acceptable items to users.
This study compares the prediction performance of a multi-criteria recommender technique employing an adaptive evolutionary algorithm to a classic recommender approach.
1.5 Significance of the Study
This study benefits web users and application domains like e-commerce, e-learning, e-government, social networks, and e-tourism by making decision-making easier, faster, and more efficient. Additionally, user ratings of items with multiple attributes can improve prediction accuracy for recommendations to other users.
6
1.6 Scope of the Study
The study focusses on RSs, specifically multi-criteria RSs that offer proficient, accurate predictions for consumers. This research involves the creation of a complex system capable of recommending the best idea or item to consumers depending on their preferences.
1.7 Expected Results
The project’s goal is to offer predictive performance data for the proposed technology and compare it to existing methodologies. These results include a reduction in prediction errors, an increase in ranking accuracy, and a strong connection between expected and actual values.
1.8 Thesis Structure The thesis is arranged as follows: Chapter 2 provides an overview of the RS and multi-criteria RS, delves into the adaptive genetic algorithm and its component, and reviews relevant research.
The third chapter explains the study’s methodology and architecture. Chapter 4 describes in detail how the system was implemented. It also discusses the results. Chapter 5 concludes by discussing the summary, conclusion, and recommendations.
Do you have the whole work from chapters one to five titled “Enhancing Prediction Accuracy Of A Multi-Criteria Recommender System Using Adaptive Genetic Algorithm“? Click below.
Need help with a related project topic or New topic? Send Us Your Topic