Decision Optimization and Machine Learning: Complementary Techniques for an AI-Driven Future

I am pretty sure you have heard about a lot of different terms when discussing about new data processing solutions. One person talk about Analytics, another one talk about Big Data.

One person says they are using Machine Learning while another calls it Data Mining. Still others may claim to be doing Decision optimization while Machine Learning is the favored terms and technology for some. You are confusing, i will try to help you in better understanding and using the right (combination of) technologies.

Regardless the technologies (and domains) you could hear (big data, Artificial Intelligence - AI, or Business Intelligence - BI), you should consider the point of view of the data and the benefits for the business (client). This is why we talk about descriptive analytics, predictive analytics and prescriptive analytics.

Descriptive Analytics: provides tools to look at data, reporting and dashboard, such as Cognos or Qliktech, describing what has happened

Predictive Analytics: helps to understand or identify what might happen next,, such as SPSS, R or all ML DL frameworks, predicting, forecasting or correlating data, anticipating what should happen, next (ML-DL, data mining)

Prescriptive Analytics: allows organizations to decide which actions to take in response to the likely effect of different decisions), business analytics optimization such as ILOG, answering to the question: what is the best action in light of information/insights i got

Now let me give you more details about Machine Learning predictive) and Decision optimization (prescriptive).

ML discovers unknown data that is required as input for DO. It can be prediction of sales, forecast of some external conditions, of classification of customers.

Then DO uses this initially unknown data in addition to other deterministic data into a defined optimization model and the outcome is a proposed set of actions. 

Typical examples we use to demonstrate this are:

 marketing campaign optimization: 

- ML predicts the expected revenue of proposing a given product to a given customer through a given channel based on historical records and customer properties

- DO select the optimal set of proposals according to budget and other limitations, and maximizing the expected total net revenue.

 another example to better understand

predictive maintenance:

- ML predicts the propensity of industrial assets to fail given their usage and ages,

- DO schedules how a limited set of technicians should execute the maintenance to minimize the expected production impact

In one word (and picture) to summarize:

descriptive analytics is about getting information on (huge volume of ) raw data

predictive analytics is about getting insights on my business based on (historical) information

prescriptive analytis is about getting the optimized outcome based on information and insights

if you would like to continue this discussion please contact us at IBM Client Center - Montpellier #ibmccmpl or please visit the following video