Online dating sites have transformed the way that people make friends and find partners. Many dating websites have recommendation algorithms that match people with potential partners based on their mutual likes and dislikes, interests, hobbies and so on.
That certainly helps to prune the field but the inescapable truth of the online dating game for most people is that you have to kiss a lot of frogs before finding your prince or princess.
So finding a better way of matching potential partners is a much needed goal, not to mention a valuable one given that the online dating business is worth some $3 billion a year.
That’s why the work of Kang Zhao at the University of Iowa in Iowa City and a few pals might attract some interest. These guys have built a recommendation engine that not only assesses your tastes but also measures your attractiveness. It then uses this information to recommend potential dates most likely to reply, should you initiate contact.
Most web users will be familiar with the recommendation engines on Amazon, Netflix and so on. These work by analysing the set of books you have bought, for example, and then finding other people who have bought a similar set. It then recommends books from their lists that you haven’t bought.
The dating equivalent is to analyse the partners you have chosen to send messages to, then to find other boys or girls with a similar taste and recommend potential dates that they’ve contacted but who you haven’t. In other words, the recommendations are of the form: “boys who liked this girl also like these girls” and “girls who liked this boy also liked these boys”.
The problem with this approach is that it takes no account of your attractiveness. If the people you contact never reply, then these recommendations are of little use.
So Zhao and co add another dimension to their recommendation engine. They also analyse the replies you receive and uses this to evaluate your attractiveness (or unattractiveness).
Obviously boys and girls who receive more replies are more attractive. When it takes this into account, it can recommend potential dates who not only match your taste but ones who are more likely to think you attractive and therefore to reply. “The model considers a user’s “taste” in picking others and “attractiveness” in being picked by others,” they say.
At least, that’s the theory. Zhao and co have tested their idea for a dating recommendation using an anonymised data set from a dating website with 47,000 users over 196 days. They use the first 98 days of data as a training set to work out each user’s tastes and attractiveness.
They then tested the recommendations of the new dating engine against the rest of the data set and compared it with other methods of recommendation, such as using taste only or by matching other variables such as each person’s likes and dislikes.
They first compared the number of potential dates recommended with each method that users actually contacted. This is a measure of how well each method captures a user’s taste. But they also measured how often these contacts were reciprocated, thereby measuring the suitability of the match.
Zhao and co say the results clearly show the superiority of their method, although they do not quantify the benefit in their analysis. “If a user approaches a partner recommended by [our engine], he/she will have a better chance of getting responses,” they say. What’s more, the engine does while still recommending partners that the user should like.
That looks to be an interesting advance. It has one weakness, however, which is the well-known “cold start” problem. How can the engine evaluate a new user?
It’s straightforward to ask a new user to rate a number of potential partners to get an idea of their tastes. But it’s much harder to get a sense of their attractiveness until the contacts they’ve initiated have been reciprocated or not. That’s a problem that many recommendation engines face.
This type of recommendation engine might have other applications too. The network of links in heterosexual dating networks is special for two reasons. The first is that a link can only exist if contact is reciprocated–that’s like the links in Facebook or LinkedIn: both parties must agree.
The second is that the network contains two types of nodes–male and female–and links can only exist between nodes of different types (in a heterosexual network, anyway).
This is known as a reciprocal bipartite network. Another example is job application networks so this kind of recommendation system ought to work here as well.
Of course, the true test of the pudding is in the eating. Zhao and co don’t say whether they’ve had interest in testing their dating engine from real dating sites or whether they plan to start their own dating site. So we may have to wait a while to find out whether it offers a measurable competitive advantage in the real world.
In the meantime, there’s only one thing for it: keep kissing those frogs.
Ref: http://arxiv.org/abs/1311.2526: User Recommendation In Reciprocal And Bipartite Social Networks–A Case Study Of Online Dating