In order to detect emerging trends of food consumption from social media data, I employed two simple methods:
-
In the first method, I creat word counts of each ingredient mentioned in Facebook posts from 01-2015 to 12-2015. For each month, I look at the top mentioned ingredients, then I plot the trend of these ingredients being mentioned over time to detect abrupt changes in the time series. However, with this method, I can only identify the trend of a single ingredient and cannot tell what ingredient it is being mentioned with. For example, I can only see the trend of "Cauliflower," however this trend has a lot of noise because "Cauliflower" can be used with many other ingredients in different recipe, making it hard to detect the rise of "Cauliflower Rice."
-
From the limit of single-word-count method mentioned above, in the 2nd method, I built a co-occurence matrix of ingredients for each month and observed the top 50 ingredient combination. After that, I calculated Lift and PPMI. If I identify any interesting ingredient combination emerging from this analysis, I will plot time-series data of this ingredient combination to see whether it could be an emerging food trend. After validating this method on pumpkin pie, cauliflower rice and vegetable noodle, I found this method showed a fast and accurate detection of food trends.