Trend analysis of twitter
Twitter users are familiar with top trends. It is a small list of topics which are being discussed more often by twitter users. These topics can be anything from #followfriday to #pakistan. This varied characteristics of twitter trends is due to the varied nature of twitter users. These topics basically depends upon the thought, local concern, international concern, hot discussions and much much more. These trends can have a short term effect like #District 9 or a very long term effect like #h1n1 or #NHF. Thus there is a equally proportional relation between the temporal length of a trend and the importance of the discussion.
Trend analysis is a scientific tool to analyse the pattern of change over time in a specified subject group. Thus, while doing trend analysis three main factors are considered: topic (anything whose trend we want to see), period (duration of the analysis of trends) and subject (here social network users).
The reason behind doing this analysis in twitter is to discern the level of knowledge, awareness, willingness to pay, demand, acceptability, effect etc of certain topic over time. For example a health institute can do a twitter trend analysis to understand the level of awareness regarding swine flu in a non-geographically bounded area. Whereas, coca-cola might do this study in order to find the attitude of it’s consumers about it’s products over time. Some might want to do a interrupted time series analyais. In this sort of study, a company will do a time series analysis of a product and then launches a new program about the product and do a study after this interrupt to find out if the program which was launched was actually successful. For example, yahoo might do a interrupted time series analysis before and after upgrading it’s new website. If Yahoo finds that people in twitter are quite positive about yahoo’s this new change, it will continue with it’s program and won’t worry about it any more. But if people in social networks are against it’s move, obviously yahoo would like to re-think it’s project.
When we have to do some analysis of trends in twitter, we have to be quite clear with our topic, as it’s very important and a lot of information regarding the subject is necessary pre study. The sample size of the study and the time frame of the study must be significant and choosen carefully. For example, the trend of drinking soft drinks like pepsi will always go down in winter. This time frame must not be selected for our study if we want to study trends about usage of pepsi. But if we are doing the trend analysis of pepsi in a year, we might have to keep our stretch in time to be more than a year. We must also be able to sort out the extremities of our study. If our topic of research does only have a local significance we don’t necessarily have to cover the international audience. These people will fall under our outliners.
While analysing trends, it will be better if we have a numerator and denumberator data available before we do the research. For example lui vitton might not have a significant sales in china thus having a low trend. This information about the on-going low trend of lui vitton in china would be considerably important in our study, as it may obviously show a low trend. Whereas, there are other factors that may influence the trend of the topic in the specified time frame. For example, when michael jackson died, he remained a top trend for a very long period, this may have declined the trends of some other important topics. Thus these confounding factors needs to be monitered before doing any trend anlaysis.
Twitter is a powerful social networking site that has a large information about almost anything. This new sensation has an impact on people and companies. Thus anlaysing the way they are thought of, their product is considered and much more, is very important.