Traditional descriptive analytics looks at current and historical organizational performance to answer questions such as “How have the monthly sales been for the past twelve months?” and “Who are the most valuable customers?” On the other hand, predictive analytics forecasts trends, behavior and events for decision support to tackle questions such as “What are the projected sales for the next six months?” and “Who are the customers most likely to leave?” An obvious next step to descriptive and predictive analytics is prescriptive analytics (a.k.a. decision support), which is a process to determine alternative courses of actions or decision options given the situation along with a set of objectives, requirements, and constraints. Prescriptive analytics therefore answers questions such as “What actions could be taken to increase sales?” and “What incentives can be offered to encourage customers to stay or to prevent them from leaving?” Machine Analytics is pioneering the field of predictive and prescriptive analytics.
Traditional quantitative statistical techniques are incapable of solving complex predictive and prescriptive analytics problems that require advanced functionalities such as dealing with subjective knowledge, perform speculative analyses, and extract useful patterns from big data. The need is therefore to devise advanced analytical techniques to perform complex analytical processes.
Machine Analytics’ PPA relies on cutting‐edge machine learning (ML) (or, loosely, data mining) and artificial intelligence (AI) techniques that are grounded in rigorous mathematics and statistics and go beyond traditional clustering and regression analyses. We model hard business problem using generative (e.g. Bayesian networks), discriminative (e.g. neural networks), and symbolic (e.g. rules) modeling techniques to reason with time, actions and utility to perform decision making under uncertainty. The models are built using a combination of subjective knowledge from domain experts and automated learning from multisource data wherever available. Once we generate models, we search for relevant data sources using intelligent agents built in DAS [Distributed Analytical Search] engines to drive those models.
Machine Analytics believe that modern complex business decision problems can never be solved using one single technique given the diverse nature of information sources. For example, an organization can have and make use of structured, transactional, unstructured data, customer survey, experts observations, and opinions. Our innovation is the hybridization of statistics, machine learning, and artificial techniques. For example, we use Bayesian networks as a modeling paradigm for predictive analytics in conjunction with deep linguistics processing to search and extract evidence from text.
Machine Analytics has an array generative and discriminative modeling and inference tools. Demonstration is available upon request.