From search to security: the five most important algorithms in tech
We design, test, and implement algorithms for your specific situation and data. Examples of some algorithms and situations are below:
A Recommender System is an algorithm that seeks to predict a user's rating of an item based on existing data from that user and other users.
The classic example is that of recommending movies to view based on past movies that the user liked and other people's ratings of those movies.
There are many ways in which a user rating can be predicted - two examples below are Collaborative Filtering and Social Network Analysis.
Social Network Analysis is the use of network theory to analyse social networks. We also include the term to mean any network of people such as buying networks (a.g. auctions) and email networks.
Network Theory is the study of things connected to each other - it is usually applied to people rather than connections of computers such as the internet. It can determine who in a network are the leaders and followers and influencers etc and can pick out cliques for market segmentation.
Virality Prediction is the science of predicting how an item may spread through a network.
It has two apsects to it: (1) how viral or contageous the item is in its own right and (2) where is the best place in the network to introduce the item so as to maximise the rate it spreads.
Collaborative Filtering is a way of creating a recommender system.
Given a matrix of users (in rows) and ratings (in columns) it uses matrix techniques such as least squares and the singular value decomposition to predict the missing values in the matrix. Thereby we can predict the ratings of movies that users are yet to see.
Matching Algorithms are used by dating web sites to match people with each other.
It is similar to the movie rating problem but lends itself to specific algorithms such as the Gale-Shapley algorithm which help Lloyd Shapley win a Nobel Prize in 2012. It is a solution to the Stable marriage problem which is what you have when you want to match up a fixed and equal number of partners in marriages so that everone is maximally well off.
Link Analysis refers to the analysis of links between people (rather than links on a web page although it can be used in that area).
It is particularly useful for investigation of criminal activity (fraud detection, gangs, terrorist groups). It can pick up money flow and trust relationships between people.
Segmentation is a clustering technique which involves dividing a broad target market into subsets of consumers who have common attributes such as liking action movies or living in the same city.