A SMART MEDIA based RECOMMENDATION SYSTEM
A SMART MEDIA BASED RECOMMENDATION SYSTEM
A SMART MEDIA BASED RECOMMENDATION SYSTEM ABSTRACT
Smartphones and tablets, for example, are becoming more powerful, smarter, and affordable, and hence more popular.
Recommendation systems have grown quite prevalent in e-business and e-Commerce, with Amazon, Google, eBay, Facebook, and others all using them to promote their businesses. Recommendation systems are infrequently utilised in education, yet they can be quite effective.
The following is how the suggested project works:
Send a learning query to sites, sources, and repositories all over the Web to acquire relevant knowledge using a recommendation system that filters out all worthless or irrelevant materials from the primary list of recommended items.
Filter the results based on other users' preferences, keeping the current query in mind.
TFIDF should be used for content-based filtering and ranking of shortlisted sites or articles.
Present the shortlist to the system user depending on their rank.
With the expansion of the web, the amount of content available to users all around the world has increased dramatically.
The resources supplied are a mix of valuable and useless documents for the user's benefit. This show emphasises the need for a better method of obtaining useful documents from the web, mostly through the preferences of web users.
This project intends to use a user's preferences and behaviour to obtain documents that are relevant to what the user requires. Depending on the user's selections, they may have others with similar tastes as them, and the system is to capitalise on this chance and recommend things based on interaction of the user's neighbour with other interesting materials.
The system can locate items (documents) that other users with comparable interests to the present user have read and rated highly using a collaborative filtering approach.
In the early stages of the system's existence, the system cannot rely solely on the Collaborative method, or else the system's recommendation will be insufficient, necessitating the need for another filtering approach to compensate for the collaborative approach's shortcomings.
The “Content-based filtering” approach is the most commonly used filtering approach with collaborative filtering. Adding a content-based filter to the collaborative result will produce a more definite recommendation that is near to what the user requires.
Recently In e-technologies such as e-business and e-learning, recommendation systems are becoming more widespread and crucial. Many large corporations, such as Amazon and Google, have their own recommendation system.
Customers can use recommendation systems to find the most relevant products that meet their demands. There has not been much study done in the field of e-learning and learning in general to design and build a trustworthy recommendation system that can assist learners in selecting the most relevant materials that can speed up and increase their learning process.
In this project, we intend to create and implement a learning recommendation system that accepts smart media as clients, such as smartphones and tablets.
To accomplish this, the system must be hosted on a distant server that is capable of running codes efficiently and optimally. A server side will be created in Java Servlets to allow for the use of multiple libraries. All communication will take place over a network, with data sourced from the internet.
1.2. STATEMENT OF THE problem
In today's digital age, the amount of content making its way to the internet at an alarming rate has led to a rise in the monitoring of this content by search engines and other pointer-sites. It is now clear that there is a need to provide a human touch to material recommendations to consumers.
If a type of recommendation can be added to user search results, the likelihood of them finding more interesting things increases. Consider a school that has a large number of items in its repository and allows students or the general public to access it, but only provides a generic search.
If there was a way to find articles or papers of interest while also finding materials with similar relevant content, based on the content similarities and the people who previously viewed and rated the materials, the next person would not have to waste time viewing irrelevant materials.
1.3. OBJECTIVES OF RESEARCH
The goal of this research is to increase the quality of resources given to individuals or system users by assisting common search with the use of filtering algorithms.
Because locating material for research and study on the web can be difficult, a user may search for materials and information using a search engine, and the user may or may not get the result he/she seeks from the first search engine, and he/she may attempt the second, third, fourth, and so on.
This practise of looking for information in multiple places can be time consuming and irritating. Then a system that searches multiple locations will undoubtedly be superior.
We intend to give a multi-platform search based on hybrid filtering to present the user with a result that is close to what the user wants or requires.
Normal search (e.g., Google) similarly filters results, but we seek to combine results from several sites and sources, providing visitors with a more scrutinised list than the result offered by a single website or service. C o m p u t e r S c i e n c e A U S T 2 0 1 3 / 2 0 1 4
Our system will be able to deliver a better result-set for the system's user with the help of a user's preference and the characteristics of users with the same preference as the current user.
Consider a university library with a large collection of materials. If there is a way to search for materials, the system will be able to deliver a good search based on the title of the document or even some element of the content.
This project focuses not only on the technique that will be used in implementation, but it also relies on the creativity of future developers to improve the programme and use it for more advanced purposes.
We want to demonstrate how useful content-based and collaborative filtering algorithms are in recommending resources based on a user's current query.
1.4. RESEARCH METHODOLOGY
The following methodologies will be used in this project:
1.4.1. Gathering Information
This is where we search the internet for information, usually utilising search engines and a tool to search the DOM for links. (The Document Object model (DOM) is an HTML and XML application programming interface (API).
It specifies the logical structure of documents as well as how they are accessed and altered .
When the lists from various sources are acquired, they will be integrated to generate a single list free of duplicates.
This list will be routed through the collaborative filter before being routed through the content-based filter.
The preceding list is a database table having the fields ID, TITLE, URL, BODY, and SCORE.
When the list is complete, the system will use JSoup to retrieve all of the URLs, the title of the page, the content of the pages (i.e. the contents within the body of the page), and save the URL in the URL field. To avoid inconsistencies in the recommendation, this table is only used by the present user.
Collaborative Filtering 1.4.2
It is a filtering strategy that exploits the relationship of peer users with comparable rating histories to create recommendations utilising the neighbourhood. .
C o m p u t e r S c i e n c e A U S T 2 0 1 3 / 2 0 1 4
The system makes recommendations based on whether the URLs in the list fetched from the internet have been rated before. For every URL in the list that has been rated, the system makes a recommendation (finding URLs/items similar to the current one) for extra URLs, and they are added to the list of URLs.
1.4.3. Filtering based on content
The system makes recommendations based on two factors: product attributes and user ratings. Content-based recommenders approach suggestion as a user-specific classification problem, learning a classifier based on product features for the user's likes and dislikes .
I aim to use this approach to rank documents based on their weight for a set of terms in a query. A document has a weight for each term, and depending on the amount of terms in a query, a document has a total weight. This combine-weight will be used to rank the temporary-corpus documents.
1.4.4. Combining Both Filtering Methods
Combining two different filtering algorithms to produce a new system that outperforms any individual filter. The use of a hybrid will aid in overcoming the shortcomings of a single filter-system.
In this case, after the collaborative filter has finished recommending and adding its recommendation to the list, the list is sent to the content-based filter for further filtering, resulting in a rank that places important documents at the top and less-important ones at the bottom.
1.4.5. Connecting the Android app to the server via the HTTP protocol
The project client is mainly an Android programme that queries the system in order to obtain certain system-proposed responses.
As a result, the mobile app must interface with the server in some way, and Android may communicate with remote web servers via the Hypertext transfer Protocol (HTTP) through a network.
To take advantage of Java's power, I opted to use the Java-Servlet to execute the request made by the Android application.
Different packages have been produced to support the system's varied methodologies, including the use of Java external libraries such as Apache Mahout , JSoup, JDBC-MySQL connector, and so on.
S c i e n c e S c i e n c e S c i e n c e S c i e n c e S c i e n A U S T 2 0 1 3 / 2 0 1 4
Figure 1 depicts the structure of the recommendation process.
1.5. ORGANIZATION OF DOCUMENT
This document is structured as follows: Chapter 2 provides a thorough examination of the approaches and theories employed in this thesis.
Chapter 3 depicts the processes of the software used in our case study. Chapter 4 is the conclusion that demonstrates the work done in the project's cause, as well as the limitations of the application and proposed techniques in the use-case.