President's Commentary: Predicting the Previously Unpredictable

As the world moves faster and faster, decision makers at all levels often face a precarious balancing act between being decisive and taking the time to properly analyze and think through decisions.
Predictive analytics can help with this challenge by integrating the power and techniques of modeling, simulation, statistics, cloud computing, machine-to-machine learning and other decision aids and coupling them with the appropriate underlying data to improve decision making. It holds major potential for providing improved, cost-effective and suitable outcomes. More informed decisions result in a better allocation of resources and likely more favorable results for any given task or mission.

Predictive analytics already is having a major impact on decision making in the areas of national security, industry, academia and even daily life. The worldwide explosion of disparate data has grown beyond the imaginable. To help deal with this growing data onslaught, technologies deployed by the Defense Department, intelligence community, industry and academia already are enhancing the ability to forecast with an improved level of confidence—sometimes months or even years in advance—heretofore unpredictable events that now seem to occur at a dizzying pace. Predictive analytics also is becoming more and more useful in a variety of other areas, including predicting the need to replace machinery parts, the spread of disease, the development of international and domestic crises, the potential for cyber attacks and the outcome of medical treatments, to name a few.

Time is frequently a key element in effective decision making. The ability to capture, analyze and use critical data often needs to occur at machine speed, well beyond human comprehension and reasoning. For example, in the domain of cyber operations, we need to analyze, decide and act or react to events faster than any human actions possibly can. Used in conjunction with key automated tools, predictive analysis offers a potential means to help automatically counter cyberthreats in real time.   

Properly identifying, weighing and vetting data sources is a key factor in getting predictive analysis right. Government, industry and academia have access to large volumes of dissimilar data from a wide array of sources, including websites, social media, news media, diverse databases and different types of sensors in everything from automobiles to surveillance and reconnaissance centers and hospitals. However, much of the pertinent data in its myriad forms fails to factor into the decision-making process either because it is too cumbersome and too pervasive to process or because systems that need to share data are unable to do so. In addition, the information is often not in a format conducive to rapid recovery and analysis. In other words, because we have not had the resources and tools to assimilate and assess the data, much of it falls on the floor, unseen, untouched, unused.

Additionally, the available data often lacks the proper vetting to distinguish truth from conjecture, misinformation or disinformation. We must develop the capability to verify the veracity of information that goes into the analysis so that we’re not dealing with rumor, innuendo or intentional distortions that lead to false outcomes. The ability to collect, fuse and assess the data to predict events and outcomes well in advance only becomes a decision-making force multiplier if the information is available and accurate.

Further, as information in all its forms becomes a more valuable asset, we must secure and protect it vigorously. Without data security, state actors, terrorists, criminal groups and individuals can steal or manipulate data for nefarious purposes, leading to improper decisions and undesired consequences.

There is also a need for users to be confident in the outcomes of an analysis. The sheer number of variables presents challenges, and any altering of data can significantly change a model’s outcome. A mistake in the weather pattern, for example, can change a decision and produce disastrous consequences. Disruptive event forecasts—and more importantly, the decisions resulting from those forecasts—are only as good as the data itself. Properly vetting the information and keeping it secure are critical elements to achieving the tremendous promise of predictive analytics.

A continually shifting global economic and political environment, evolving national and international security interests, and a more austere fiscal environment beg for sound decisions and appropriate actions in a timely manner. Predictive analysis is an enabler to this end. We need to continue advancing this capability quickly to further its positive impact on decisions. Our decision makers require nothing less.