In the past few weeks, I have worked on … something different from what I normally do. Although social media data mining is use widely in almost every field, I thought I would try a different approach.
My idea was to see if I can collect real-time data and process it, while also forecast, based on keywords, hashtags and expressions, the next food scarcity hot spot (at local level). Two of the main things that can lead to food scarcity are food prices spikes (people can`t afford food anymore) and low or no availability of food (people have the money, but nothing to buy OR food is not produced in sufficient quantities). If we go on and speak about undernourishment, malnutrition or hunger, we can add to these the quality of food.
While my goal is to develop an application through which the development of food scarcity hotspots could be prevented to limit and/or cut undernourishment and malnutrition, I have only focused so far on the detection part.
Because I am not a coder/programmer/developer (but I am a nerd), I took several free social media data mining applications, out of which some, for advertising purposes, also do sentiment analysis (detection of trends based on keywords) and took them apart, code line by code line to get something that works the way I want.
I haven`t had much success in creating something new, but I actually added a feature to one application. But, instead of real-time mining, I had to settle with a keywords AND expressions processor that works with a list. From about 1000 random tweets that I added (tweet by tweet), I got some nice results. An initial analysis showed me that people in Benin are complaining about the weather conditions: high temperatures combined with heavy precipitations. I checked Weather Underground and it seems that it was (actually) right.
Now I have to figure a way to keep working on this and develop real-time forecasting. So far, this has been an idea on which I worked as a hobby.