Event Detection and Prediction - Research Paper Example

Anything going on in the world is shared and communicated through the internet, especially on social media. Social media offers people the chance to interact, comment on events, and send instant messages all over the globe without geographical barriers. Some of them include Facebook, YouTube, and Twitter amongst others. Social media platforms have opened many research opportunities because of the amount of information they possess. This information can be used for many purposes including things such as prediction and detection of events and even as warning systems.

            An event is yet another issue in this research that requires defining. Events are one of the most important indications of people’s memories. They are a natural way through which people refer to any observable occurrences that bring people together in the same places and time to undertake similar activities (Mingers, 2003). They are quite useful in making sense of the world around us, as well as in helping people recall experiences they go through.

They can help in explaining the world around us by describing certain phenomena, which can in turn aid in predicting the future. Social events on the other hand refer to those events that are attended by people and presented in the multimedia content that is shared through online websites. Such events can include disasters, concerts, sporting events, public celebrations, and protests amongst others.             This research is about detecting and predicting events based on information collected from social media.

Specifically, the research shall use Twitter, which is one of the social media that have fast emerged over the last few years. Many people use Twitter for reporting events as they happen in the real world (Guthrie, 2010). This social media currently has over 500 million registered accounts all over the world that generate about 340 million messages daily. Most of these messages contain personal updates, opinions concerning current issues, moods general life observations and events amongst others.

The readily available wide range of data from Twitter offers an ideal source for mining information for this research. There have been proposals concerning event mining that make use of the tweet texts. 

However, none of them has proposed used of images in the data mining, which makes the analysis solely textual. In this research, we seek to use both textual and visual mining to detect events in order to improve the performance of the entire research considering the retrieved results will depend on the amount of information colleted (Walliman, 2011). The research aims to develop an accurate and effective detection method to detect events from the Twitter stream. We shall be monitoring the Twitter stream in order to pick up texts that have photos, which shall then be stored in a database.

This shall be followed by an extraction of feature in both text and photos to be applied in the mining tool. With extraction of feature in the text, the research shall use Bag of Words from using the Term Frequency-Inverse Document Frequency method (TF-IDF). Scale-Invariant Feature Transform (SIFT) and color histogram shall be used. These shall be followed by a comparison of our mining results with other methods in text mining by other researches. Problem Definition Considering that social media has become mainstream in interaction of people, the main issue for this research is how to identify events that are discussed.

Twitter messages come in millions everyday. Each of the messages posted contains meaning, with some related to events while others relate to personal issues and status. The number of messages makes it quite difficult to identify which ones are event related and those that are not considering one cannot go through them (Blaxter, Hughes & Tight, 2010). The main problem is to differentiate the event and non-event messages. Additionally, the number of events addressed by Twitter messages can be varied, making it hard to determine those that are related.

The problem posed here is finding the right method and techniques of detecting events from social media. One of the difficulties posed by the issue is the fact that events present a real world experience, meaning they cannot be controlled in a laboratory. This means that the research has to be undertaken within its context, which is not easy and can produce varying results compared to laboratory or controlled research (Galliers & Land, 1987). Research Questions How can one develop an accurate yet effective event detection method for detecting and predicting occurrence of events using Twitter stream by picking up tweets with texts and media photos?

Hypothesis Social media such as Twitter can be used in detecting events as they happen in real-time because the users post messages at the same time it is occurring. H1: Information systems engineering approach to research is the most suited methodology for detecting events using information across social media context H2: By clustering the information collected from Twitter, using incremental algorithm researchers can be able to separate non-event and event tweets Review of Literature Several literatures have sought to introduce various research methods within the Information Systems.

The field has seen an increase in the number of scholars that are dedicated to finding more solutions to the issue. This section offers views and perspectives of some of the top authors in the Information Systems research. One of the top authors in this field is Clarke, who argues that Information Systems research has three major traditions. They include engineering/design research, conventional scientific research and interpretivist research (Clarke, 2000). He further states that while the first one is from computer science and engineering discipline, the rest are from a business context.

The conventional scientific research purports that, a real world that compromises objects and processes exist, which cannot be understood through human artifact except through observations. It seeks to extract new hypotheses out of established theoretical frameworks, test them and finally publish the acquired knowledge. When such a research is stopped at the theoretical stage, it becomes a pure research while it becomes an instrumental one upon proceeding beyond this theory.