Please use this identifier to cite or link to this item: https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/4247
Title: Intelligent Twitter Agent
Authors: Padmasiri, U.C.B.I
Issue Date: 27-Jul-2021
Abstract: The rapid development of information technology has impacted the human behaviour in different ways. The invention of social media services has connected people from all around the world irrespective of the geographical locations. These platforms provide different ways of connecting with people via different contents. Unique features, popularity, UI/UX, business usage etc. of each social media services compel the users to use more than one social media service. Twitter is one of popular social media service where users connect with messages called tweets which limited to 140 characters. Twitter user’s twitter-feed is constructed from the tweets from followers of a specific user. All tweets of followers are shown to the user irrespective of preferences since Twitter does not provide a feature to provide feedback on the preference of tweets. The Thesis describes a computerized system that gathers Twitter user preference for tweets, analyze the preferences of tweets using Natural language processing and machine learning techniques and generate Twitter content based on the user preference. When considering the high-level architecture of the system, It includes several modules such as API gateway module that integrate the Twitter API with the system, Core application module developed using python for the analysis purposes and the mobile application that used to gather and display user preferred tweets. The system implementation was carried out by using multiple languages, Java and Python. Python language was very effective when processing text and it includes various libraries and platforms in machine learning and classifications. The evaluation was carried out for each data pre-processing techniques. Thus the implementation could use the best and efficient data pre-processing steps without losing data. Some of the data pre-processing steps such as removing stop words were not performed because it reduces the accuracy of the classification. The basic requirement was to select the best classification algorithm to perform in a small amount of data. Thus The suitable classifier was also selected after evaluating multiple classifiers. After performing the evaluation, logistic regression classifier was selected for classification. It could gain around 70% of accuracy in the classification.
URI: http://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/4247
Appears in Collections:2018

Files in This Item:
File Description SizeFormat 
2015MCS051.pdf2.31 MBAdobe PDFView/Open


Items in UCSC Digital Library are protected by copyright, with all rights reserved, unless otherwise indicated.