Bayes Analytic Overview

Bayes Analytic specializes designing custom machine learning engines for use in complex applications such as stock price prediction.    We have provided a open source engine that incorporates many of the most critical machine learning techniques.   It is available at  We also work on other interesting technlogy problems most often related to high performance distributed computing.

The Bayes Analytic  machine learning engine  has  uses classification based inference techniques, statistical selection and genetic algorithms to produce output that is optimized through the day to deliver the best results.  It uses what we think is a novel adaptation the bridges concepts from KNN and Baesaian engines to deliver good classification that for some problems delivers comparable accuracy to Deep learning CNN and RNN engines but learns and runs much faster.

The engine works by analyzing statistical similarities for many measurement features to determine the similarity of current offers as compared to past offers. It develops a ranking based on the known success of past offers. It then uses this knowledge to develop a rank ordered list of the best offers for action by human buyers. This engine includes features which allow it to handle very large rates of inbound data to present near real time recommended set of purchases. It also uses genetic algorithms to adjust the statistical weight of various features to dynamically adjust its output to better optimize for maximized success rates.

Bayes Analytic Engine for Sales & Marketing

The same core engine could be applied to predict other things such as selecting customers who are most likely to respond favorably to a given offer or to determine which products will sell best in specific cities. The benefit for companies who adopt the technology is reduced marketing spend while improving customer satisfaction and revenue per contact. The same engine could also help direct sales people to maximize the sales per contact hour.

In a marketing context the system would use correlated data similarities between users to identify the probability of success if a similar offer was made to other users. For example: After buying a brand new cell phone it is less likely that I would respond favorably to another offer for a different new phone than other offers such as allowing me to try the data tether function.

The system can analyze past purchases and other actions of many customers and deduce based on similarities with other customers which have made similar purchases to find the subset of customers most likely to respond favorably to a given offer.

If a company made an offer to 300,000 customers with a 8% success rate and the system was able to statistically select a set a subset of the 300,000 customers where there was a 25% success rate then they can reduce the number of customers they presented the original offer to. This allows them to optimize offers presented to each customer so each customer ends up seeing fewer offers but each offer they look at is more likely to be something they want to purchase.

(C) 2013 Joseph Ellsworth All rights reserved
All statements and expressions are opinions of the author and are not intended to be either investment advice or a solicitation or recommendation to buy, sell, or hold securities. Many of these statements are based on sound economic reasoning, however actual response of the economy is heavily influenced by politics and large business and so the outcome could end up substantially different.

Contact Us

Leave a Reply