Supporting Smart Interactions with Predictive Analytics

Predictive analytics, which has grown out of Business Intelligence (BI), employs knowledge discovery on existing data to predict or recommend future actions. It differs from BI in that BI focuses on past performance and the discovery of trends while predictive analytics forecasts behaviour and results in order to guide specific decisions. Predictive analytics is used to analyze large amounts of data with a large number of variables. It can use a variety of techniques including clustering, classification, decision trees, neural nets, regression modeling and hypothesis testing. The core element of predictive analytics is the predictor, which is a variable that can be measured for an entity to predict future behaviour. For example, a credit card company could consider age, income, credit history and other demographics as predictors for a card applicant’s risk factor. Multiple predictors are combined into a predictive model that, when subjected to analysis, can be used to forecast future probabilities with an acceptable level of reliability. Copyright © 2009 Patrick Martin. Permission to copy is hereby granted provided the original copyright notice is reproduced in copies made. Predictive analytics can be used to support a user carrying out a smart interaction. It can be embodied as a component of a matter of concern that analyzes historical and contextual data relevant to the task and predicts probable outcomes or suggests possible actions for the user. Consider, for example, a hospital that wishes to reduce the occurrence of infections in patients undergoing operations at that hospital. When a patient arrives for an operation hospital staff could build a profile of the patient and the procedure. A predictive analytics tool could compare the profile with historical data describing previous operations and predict outcomes such as the likelihood of contracting specific types of infections. The tool could also suggest actions to be taken to limit the possibility of infection for the specific patient. We propose to study the use of predictive analytics to support smart interactions. Initially, we plan to develop a predictive analytics tool for a specific use case such as the hospital one above. We will use data mining techniques to identify predictors for a set of outcomes, build a predictive model, evaluate the model and implement the tool based on the model.