Tuesday, August 31, 2010

Emotions Draw Close Friends: Analyzing the Social Network Structure of Facebook Fan Pages

Recently we were wondering if the social network structure of fans of a brand, a star, or a cause tells us how passionate the fans are. To be more precise, we were looking at the network structure of the friendship network of Facebook fan pages. This means that we collected – as far a publicly accessible – the friendship network of the people who clicked on the “like” button on a fan page.
For a start, look at the fan page of our own COINs2010 conference (by the way, the conference will be soon in Savannah Oct 7 to 9, at SCAD, we hope to see many of you there ☺ ).

The dark dots in the network are the fans of COINs2010, the green dots are their friends. This means that for this initial analysis we looked at how many and how well-connected friends a fan of COINs2010 has. We ignored direct links between the fans, but focused on their external friendship network.

In this first attempt we looked at a total of 15 fan groups in 5 categories, see the table below:

We (admittedly subjectively) ranked the emotionality from 1 (product brands) to 5 (medical causes). We found positive correlation of 0.33 (although non-significant) between the network density and emotionality. This means, the more connected the friends of a cause or brand are, the more emotional they are about their cause. Even more interestingly, we found significant negative correlation between the clustering coefficient of -0.57. This means that the more the friends of fans are clustered in subgroups, the less emotional they are.

The conclusions would be that the causes with the most emotional supporters have a dense, but evenly spread out network, with few clearly separated subgroups.

Based on this admittedly very preliminary analysis, what are actions you can take to further you cause? The answer is simple: Help to weave the network of your supporters.
1. broker connections between supporters
2. fight fragmentation of supporters by connecting subgroups
In short – help build one large happy familiy!

Thursday, August 05, 2010

Predicting Stock Market Indicators Through Twitter “I hope it is not as bad as I fear”

We have been working on trying to predict market indicators for quite some time by analyzing Web Buzz, predicting who will win an Oscar, or how well movies do at the box office. Among other things we have correlated posts about a stock on Yahoo Finance and Motley’s Fool with the actual stock price, predicting the closing price of the stock on the next day based on what people say today on Yahoo Finance, on the Web and Blogs about a stock title.

The rising popularity of twitter gives us a new great way of capturing the collective mind up to the last minute. In our current project we analyze the positive and negative mood of the masses on twitter, comparing it with broad stock market indices such as Dow Jones, S&P 500, and NASDAQ. We collected the twitter feeds from one whitelisted IP for six months from March 30, 2009 to Sept 4, 2009, ranging from 5680 to 42820 tweets per day. According to twitter this corresponds to a randomized subsample of about one hundredth of the full volume of all tweets, as the total volume in 2009 was about 2,5 million tweets per day. We tried to measure collective hope and fear on each day by applying the simple metric of counting all tweets containing the words “hope” – there were 54 to 467 tweets per day, and “fear” or “worry” – there were 9 to 100 tweets per day. This tells us that people prefer optimistic words (hope) to pessimist words (fear or worry).

As external benchmark of investor fear we used the Chicago Board Options Exchange Volatility Index VIX, which is strongly negatively correlated with Dow, S&P 500, and NASDAQ, which is not surprising, as the spread of stock options on a given day is used to calculate VIX. Initially we expected the number of tweets with hope to negatively correlate with VIX, and the number of tweets with fear or worry to correlate positively with VIX. Surprisingly, we found positive weak but insignificant correlation for both “hope” (0.135) and “fear” or “worry” (0.172) with VIX, and negative significant correlation with both “fear” and “worry” and “hope” with Dow NASDAQ and S&P500 (This means that people start using more emotional words such as hope, fear, worry in times of economic uncertainty. We therefore created a simple twitter-volatility index combining mentions of hope, fear and worry, normalizing it with the total amount of tweets per day as a baseline. This index displays strong significant negative correlations to Dow, NASDAQ and S&P500, and strong significant positive correlation to VIX (see table below).

The picture below visualized the negative correlation between Dow (blue) and “hope, fear, and worry” (green) in the period March 30, 2009 to Sept 4, 2009.

To put this in simple words, when the emotions on twitter fly high, that is when people express a lot of hope, fear, and worry, the Dow goes down the next day. When people have less hope, fear, and worry, the Dow goes up. It therefore seems that just checking on twitter for emotional outbursts of any kind gives a predictor of how the stock market will be doing the next day.

Just to be clear, what we have presented here are very early preliminary results, and much more work is needed to scientifically verify it.