Analytics
Marketing, Communication & Analytic Services for the Sensors Industry
Copyright (c) 2009
Sensorlytics LLC.
All rights reserved.
NFL Football Game Predictive Scoring Model
An example of a predictive analytic model is one that predicts the outcomes of NFL football games. We chose this topic because it is one that is easy to relate to, and showcases our capabilities in a tangible way. This model uses a combination of fixed-form equations and neural-network techniques to estimate future game scores based on its estimates of the relative offensive and defensive strengths of each team. Despite its simplicity - historical game final scores are its only data inputs - it can correctly predict wins and losses about 65%, or 2/3, of the time. The link below opens a text file showing our most recent predictions and a review of past predictions.
Our philosophy is that analytic models should provide actionable results and insights that provide tangible value to your business - and not just generate reams of statistics or pretty charts. For this reason we focus on simulation and predictive analytics - models whose outputs can be used to help make decisions, either by providing the ability to ask 'what if', or by providing some insight into what the future may hold.
A few of the tools we can apply in creating models include:
- Regression & other traditional statistical techniques
- Neural networks
- System Dynamics
Some simple examples.....
Interactive Modified Bass Diffusion Model
This model is based on the system dynamics paradigm and is a 'what-if' tool that can be helpful in understanding how consumer word-of-mouth feedback can influence sales of a productr over its life-cycle. This particular implementation is embedded as code in a PHP webpage, and interactively calculates results which it shows in tabular form.
* * *