AI: Helping Simplify Optimal Decisions
AI: Helping Simplify Optimal Decisions
Decision optimization is the prescriptive element of the data science lifecycle and is key to delivering artificial intelligence and machine learning.
Join the DZone community and get the full member experience.Join For Free
Bias comes in a variety of forms, all of them potentially damaging to the efficacy of your ML algorithm. Read how Alegion's Chief Data Scientist discusses the source of most headlines about AI failures here.
In the business world, there are many factors to consider when making the optimal decision. Rarely is it binary. There are so many data points to consider that it becomes a combinatorial problem. For example, consider when and how to raise room rates across a hotel chain based on locations and current events or how best to optimize airline ticket prices given fluctuating fuel costs, factoring in seasonal conditions and local and/or global events. This flows over into our social and personal lives, as we rightly expect to find the nearest coffee shops located to the nearest public libraries or where to buy the cheapest gas closest to the supermarket that stocks the groceries we need.
Decision optimization (DO) is the prescriptive element of the data science lifecycle and is key to delivering artificial intelligence, as machine learning (ML) and DO have somewhat of a symbiotic relationship. Consider the use of an ML churn model in the telecom industry to identify if a client is going to leave. While we can predict with a high probability of success that a client will leave, we need to add DO on top of that to decide if we want to retain that client and if so, what special offers to make. Maybe the client has not paid their bill in months, maybe the client has been abusing the service, or maybe retention funds have been spent and it is better to let the client leave. The point is that DO can help make optimal decisions on whether to retain clients that were predicted to leave. In this way, DO may help customers save money and reduce business risk.
A DO Definition
I define decision optimization as a means to get an optimal solution to a complex combinatorial problem such as complex planning, scheduling, resource management.
It is based on an optimization solver engine that executes an optimization problem and provides results — best described by Figure 1 below.
The science behind DO can be considered complex. It involves an optimization model which is a mathematical formulation of a business problem, such that an optimization engine can interpret it and find a solution that achieves the objectives while respecting constraints described in the model. And it usually requires operational research (OR) expertise.
Data science is a multi-persona discipline that requires collaboration between business users, data scientists, and developers.
A wide span of IT skills is required involving a plethora of algorithms and modeling frameworks. Each needs installation, upgrading, and integration. Business analysts are required for their domain- and/or organization-specific knowledge. Data engineers need to prepare access to relevant but often disparate data sources. Raw data and stats may be not meaningful to all stakeholders. Hypothesis and what-if analysis resonate far better. I think you get the picture.
As part of the continued mission to democratize AI and make it more consumable, many Decision Optimization (DO) capabilities have been integrated into the IBM Data Science Experience (DSX) Local as shown in Figure 2.
The benefits of integrating DO within DSX are :
- Ease of use
- Lightweight web interface
- Pre-installed essential software for performing analysis and visualization
- Support for Python notebooks for general models
- Natural language interface for scheduling problems
- Project-based collaboration for data scientists and other team members
- Project member management: adding, removing, levels of access (editor, contributor, admin)
- Sharing of project assets, like data, models, and results in current and future projects
- Ability to work with different data and model variants through multiple scenarios for exploration and what-if analysis
- Support for data preparation
DSX can help data scientists focus on DO problems from a data perspective. Optimization modeling is achieved through a wizard to help define tasks and resources. Appropriate constraints and objectives can be selected and edited — even allowing the user to elicit additional constraints and objectives using natural language — and solve and visualize the solution on a Gantt Chart.
DSX also helps data scientists and business analyst collaborate by setting up dashboards where they can mix tables, graphics, and Gantt charts, sharing the dashboard with stakeholders for them to validate their findings and conclusions as depicted in Figure 3.
For developers that simply want to consume a DO service as part of their application without having to be a data scientist, simply integrate the generated service in your application through a REST API. Simple as that.
The next step is an easy decision. Try Decision Optimization (DO) now by clicking on this Data Science Experience link. My advice: just “DO” it.
Published at DZone with permission of Steven Astorino , DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.