The Application of Data Analytics to Assist Human Decision-Making

Abstract

Both commercial and military systems are rapidly evolving into complex Systems of Systems (SoS) that incorporate advanced technologies, such as Machine Learning (ML) and Artificial Intelligence (AI). Associated with these systems are emerging behaviors that require decision support beyond the capacity of human reasoning alone. The field of Data Analytics provides methodologies to assist human decision-making with respect to complex SoS.

Data analytics is the science of analyzing raw data in order to make conclusions about the information derived from it. In fact, many data analytics methods have been designed to autonomously transform raw data into information for human consumption. This information can then be used to optimize processes and increase the overall efficiency of a business or system. The data analytics process includes the following steps:

1. Determine data requirements
2. Define the process for collecting data
3. Organize that data so it can be analyzed
4. Clean the data before the analysis
5. Perform analysis and generate results to assist with human decision-making.

Data analytics is broken down into four basic types: Descriptive analytics describes what has happened over a given time period; Diagnostic analytics focuses on why something happened; Predictive analytics describes what is likely going to happen in the near term; Prescriptive analytics suggests a course of action. Data analytics assist human decision-making for quality control systems in the financial world. The retail industry also uses data analytics to meet the ever-changing demands of shoppers. The travel and hospitality industries have also adopted data analytics to assist with human decision-making where quick response times are critical. Likewise, the healthcare industry applies data analytics to assist doctors and nurses with time-critical decisions on which life and death depend 

This presentation provides an in-depth review of the concept of “Data Analytics” methods, such as data analysis, data fusion, data storage, data sources, infrastructure and technology, screening and filtering algorithms, machine learning, and complexity; and it discusses how each of these assists human decision making. The presentation then describes specific commercial and military use cases where these methods can assist human decision-making. These use cases include:

1. Information Collection Architecture (ICA):  This data analytics use case focuses on providing accurate and timely information collection as the primary enabler for effective high-speed data network management and services. Information assessment for these emerging networks is computationally intensive to the point of stressing both technology and network architecture. Commercial companies such as IBM and British Telecom have used the ICA for high-speed electrical and optical network analysis.
2. Multifactor Information Distributed Analytics Technology Aide (MiData): This use case applies data analytics to reduce the flood of sensor data to only actionable information that is directly applicable to military missions at hand. MiData focuses on target discovery and analysis, communication capacity management, and automation techniques that enable Intelligence, Surveillance, and Reconnaissance (ISR) system operators and analysts to derive the knowledge they need to meet end-user mission requirements. By doing so, MiData greatly improves the productivity of operators and analysts to enable them to meet end-user time-critical needs while using fewer resources.
3. Mission Information Autonomous Intelligent Decision Engine (MiAide): This use case integrates data analytics capabilities to create an automated system of systems (SoS) that provides end-user capacity improvement in support of end-to-end mission activities. MiAide has been demonstrated for aspects of both manned and Unmanned Air Systems (UASs) and has proven to reduce staffing while improving mission capacity (e.g., multiplying the number of missions and mission functions) across all stages of the mission life cycle.
4. MiData Application to Local / Regional / Global Joined Object Recognition (MAJOR): For this use case, MAJOR applies sensors and data analytics technology in a new way to create a novel capability to rapidly screen massive collections of sensor images (still and video) that will transform raw data into actionable information from which analysts can locate lost objects in arbitrary geographic locations in a timely manner. This system has been applied to time-critical events, such as attempting to locate the missing Malaysian Boeing jet that disappeared in a flight traveling from Kuala Lumpur to Beijing China on May 8, 2014.
5. Object Recognition and Detection Enhancement via Reinforcement Learning Yield (ORDERLY): This use case assumes a commercial environment in which ORDERLY autonomously screens massive collections of sensor data from multiple and diverse data sources in order to transform raw data into actionable information. Similar to MAJOR, ORDERLY assists human analysts with locating objects in arbitrary geographic locations in a timely manner. ORDERLY improves MAJOR by applying Reinforcement Learning (RL), as an independent, self-teaching system. An ORDERLY prototype was implemented for which preliminary results achieved the goal of reducing overall processing time on a set of test images, along with improving analyst time from image ingestion to actionable intelligence by 33%.
6. Self-Healing Course of Action Revision (SCOAR): This use case applies Markov Decision Processes (MDP) and a Stochastic Mathematical Model (SMM) to assist humans with the generation of step-by-step military mission plans, sometimes called Courses of Action (COAs). SCOAR assists with human decision-making for both deliberate (non-real-time) COA generation and with crisis (dynamic, real-time) COA generation during the execution of complex missions. A SCOAR prototype was implemented that generated the Probability of Success (Psuccess), other selected metrics, and analytic results for both deliberate and crisis-complex mission plans.

In summary, the audience will emerge from this presentation with a focused understanding of data analytics principles and, through the examples from 6 use cases, how these principles can be applied to assist humans in making decisions for complex SoS.