The new start-up company “triangulate” from Palo Alto, CA, wants to “empower people to use their own data to make the world better”. To make this honourable aim possible they are distributing their first software “Wings“, a Facebook application. This tool creates user-profiles with the help of data mining methods, and matches them to find fitting partners.
The software is not working with a form, in which the users have to put their interests, but sources the traces users left in the web. It gains information not only from Facebook, it also uses others sources like for example Last.fm or netflix, to find out the users music or video taste. Twitter is used to determine the topics a user may be interested in.
Sunil Nagaraj, CEO of Triangulate, reports on significant conclusions they made while developing the software. For instance, the density of a users’ social network determines his or her preference of a partner. The more the network is closed, the less a random encounter of a new person will lead to a partnership. In this case it is more likely that the next partner will be from the closed network.
However the high volatility of the users’ characteristics are difficult to handle. Some people change their music taste every day – that may a relevant characteristic as well, but surely less important then a special music genre that is preferred and shared with the dream-partner.
The young start-up got 750.000$ (562.000 €) from investors to enter the very important dating-market. It has to compete against big players like match.com or eHarmony, which are using mighty dating data mining solutions for many years by now.
An important measure for a dating-recommendation engine is the visual nature of the potential partner. This aspect is very important for many partner seekers, and the data is mostly reduced to some body-measures that are more or less honestly given. To analyse and model a data mining solution that is capable to find out whether a persons photo creates sympathy or not is not trivial at all. Another dating-data-mining project called okcupid.com uses straight forwarded data mining algorithms, too, and focuses the visual nature in detail.