The idea starts with implementing capturing of user interactions with the system, thereby recording all their interactions such as entering data into a web form. Further, this data can be pivoted to build views and subsequently other tables that aid in building predictions for any form element considering all past inputs in that session. Predictions are built using a machine learning method Naive Bayesian Inference. This talk discusses challenges and issues and favorable outcomes in implementing a succesful predictions building system, while the user is still interacting with the system.
P(H|E) – probability of hypothesis given the evidence is computed on the fly where P(E|H) – probability of evidence given the hypothesis (possible values of the output variable) is first computed. The analysis is dynamic and additive in a way that if more fields are entered by the user on the web form, all of them are considered as evidence to determine the probability of a value in the next (unaltered) field.
Sumit Amar is Senior Research Developer in Microsoft Corp. He is responsible for incubation in Microsoft Bing user interface development. He has 10 years of experience in software development, and he has been with Microsoft since 2004. He started programming for fun in 1993 in BASIC language. In the past, Sumit worked in C, Perl, and Java, in addition to .NET development. His interest areas are Web UI development, semantic web representations, UI automation, and analysis of Web users’ expectations. Sumit has a Bachelors degree with dual major in Computer Science and Business, and an MBA in information technology and systems. He is a PhD student specializing in Computer Information Systems.
Comments on this page are now closed.