User:Datastudent20

Twitter Data and Business
Big data is a valuable resource, and Twitter data has many applications. One way Twitter data is used in the medical field is to improve the accuracy of predictions of the spread of influenza. Twitter data has been used to examine the way information spread after the Fukushima disaster as well as the sources and validity of that information. Sociologists can also apply Twitter data; one study focused on how people discuss insomnia.

Twitter data also has applications in business, including the following:
 * 1) Stock Market Predictions
 * 2) Customer Service Provision
 * 3) Tourism and Marketing Research

Stock Market Predictions
Public mood data extracted from Twitter has been used to more accurately predict the stock market. This is important because accurate predictions can bring large profits to investors or help prevent stock market crashes. Researchers from the University of Indiana collected tweets through the Twitter API and Dow Jones Industrial Average (DJIA) data from Yahoo! Finance from March-December of 2008. The unstructured Twitter data had features including mood words and phrases, inclusion of links, the date, etc. The DJIA data was structured data consisting of the closing value for each date.

The tweets were filtered to remove tweets containing links and to remove stop words; then all tweets containing "I feel" type statements were extracted. These tweets were run through 2 tools that performed sentiment analysis based on tweet wording: OpinionFinder (OF) and Google Profile of Mood States (GPOMS). OF ranks the positivity of the posts, and GPOMS classified the tweets in 7 different dimensions: Positive, Calm, Happy, Alert, Vital, Kind, and Sure. Then, linear regression was used to compare OF to GPOMS and found correlations between Positive and Happy, Sure, and Vital. Granger Causality Analysis compared DJIA to each of the 7 mood dimensions and found correlations between the DJIA and the Calm and Happy dimensions. A Self-Organizing Fuzzy Neural Network was then used to remove the assumption of a linear relationship between public mood and the stock market and to test whether we could more accurately predict the stock market by incorporating public mood into existing DJIA prediction methods. They found that including Calm or a combination of Calm + Happy greatly increased the DJIA prediction method. Future research opportunities include adding news sources and economic indicators to the prediction method. Limitations include the lack of demographic and geographic filtering of tweets, and uncertainty of the causal relationship between DJIA and public mood.

Customer Service Provision
No one enjoys listening to low-quality music for hours while on hold with a company's customer service. In addition to traditional customer service pathways, social media presents a venue for easy, rapid communication between customers and companies. "An exploration of logistics-related customer service provision on Twitter" aims to analyze logistics-related customer service interactions on Twitter.

Data was collected from Twitter's API using a tool the researchers created to mine conversations from the Twitter verified accounts of the top 50 e-retailers in 10 countries. These unstructured tweets were sorted to remove duplicates, non-English tweets, and tweets focused on coupons and classifieds. The data was then turned into a structured database using features including the tweet ID, the date and time of the tweet, the tweet's content, the location of the tweet, and information about the tweeter's profile and the conversation it was part of. A data coding program was created based off 500 tweet conversations and then tested on 5000 more customer service conversations. They were then able to classify the conversations into 20 topics that fell into 3 main categories: pre-delivery conversations, post-delivery conversations, and post-delivery services. They discovered that customers often post pictures of damaged goods or shipping labels and are more likely to use hashtags when they are upset. They also studied how many conversations were resolved on Twitter or in DMs or whether consumers were being rerouted to email, phone calls, or other methods of communication. Overall, they found that Twitter is an effective tool for customer service conversations and urge companies to seek to resolve issues on Twitter instead of rerouting them and report a lack of shipping company involvement on Twitter. Future research opportunities include improving conversation collection and analysis tools to reduce researcher workload and comparing social media customer service interactions for eretailers and bricks and mortar retailers.

Tourism and Marketing Research
"Using Twitter Data for Cruise Tourism Marketing and Research" sought to examine how social media could be applied to the field of cruise tourism. There is a lot of existing research related to social media analysis and the cruise industry, but not very much that combines them. Through ScraperWiki and the Twitter API, 50,414 tweets with cruise-related hashtags were collected. After non-English tweets and tweets related to Tom Cruise, cruise control, etc were removed, 42,785 tweets remained (84% of the collected data). As unstructured data, tweets have features such as location, timestamp, hashtags, and number of retweets. The tweets went through three stages of analysis. Sentence tokenization and natural language processing techniques were applied to the dataset to give a work frequency analysis. RapidMiner, an open source data analytics tool, was used to perform descriptive analysis and look at the themes in three different user groups: commercial, blogs, and individuals. Finally, a scientific social network analysis tool named Gephi was used to visualize the network, calculate social network metrics, and perform cluster analysis. The study found that cruise-related tweets had four themes: travel, destination, ships or travel companies, and emotions. Tweeters fell into 3 categories, commercial users, news outlets/blogs, and individuals, and each user base tweeted about different themes. Additionally, the social network analysis found that celebrity endorsements were an important way of reaching large numbers of users but that professional travel bloggers shared a lot more information and had a larger number of tweets. This study opened the door for research on Twitter and cruise tourism; its limitations included a short time period of data collection and no way to test for validity and reliability of user-related content. Future research would benefit from more data analytics methods. Potential pathways include development of tools to analyze customer perceptions, attitudes, and trends related to the cruise tourism industry as well as the development of guidelines for how companies in the tourism sector can effectively use social media.