Inferring Demographic Attributes of Anonymous Internet Users

Today it is quite common for web page content to include an advertisement. Since advertisers often want to target their message to people with certain demographic attributes, the anonymity of Internet users poses a special problem for them. The purpose of the present research is to find an effective way to infer demographic information (e.g. gender, age or income) about people who use the Internet but for whom demographic information is not otherwise available. Our hope is to build a high-quality database of demographic profiles covering a large segment of the Internet population without having to survey each individual Internet user. Though Internet users are largely anonymous, they nonetheless provide a certain amount of usage information. Usage information includes, but is not limited to, (a) search terms entered by the Internet user and (b) web pages accessed by the Internet user. In this paper, we describe an application of the Latent Semantic Analysis (LSA) [1] information retrieval technique to construct a vector space in which we can represent the usage data associated with each Internet user of interest. Subsequently, we show how the LSA vector space enables us to produce demographic inferences by supplying the input to a three-layer neural model trained using the scaled conjugate gradient (SCG) [9] method. 1: Introduction 1.1: The Problem The Internet attracts a large number of users and thus holds great potential as an advertising medium. Today it is quite common for web page content to include an advertisement. Since advertisers often want to target their message to people with certain demographic attributes, the anonymity of Internet users poses a special problem for them. The purpose of the present research is to find an effective way to infer demographic information about people who use the Internet but for whom demographic information is not otherwise available. Our hope is to build a highquality database of demographic profiles covering a large segment of the Internet population without having to survey each individual Internet user. The specific scientific and technological objectives of the research described in this paper are as follows: • To establish the possibility of inferring up to six demographic facts (sex, age, income, marital status, level of education and presence of children in the home) with a minimum 60% statistical confidence for a subset of the Internet population. • To establish the possibility of making at least one demographic inference about a subset of web users who account for more than 50% of the Internet traffic observed at a major website (e.g. a popular web search engine). • To establish the possibility of making demographic inferences in real-time, rather than building profiles of web users a priori. This would allow us to generate inferences on an as-needed basis using the most recent usage data. The research effort was successful on the first two points while work on the final objective is ongoing.