Using Machine Learning to Determine United States Army Readiness at the BattalionLevel

摘要

This paper develops a model for determining the combat readiness of an ArmyBattalion and investigates the problem of automating this task. The central problem is one of classification and the domain is that of military readiness in United States Army Battalions, but it could be applied to a wide variety of problems with only minor modifications. The problem is simply defined but very difficult to solve. It is first necessary to understand the current method for evaluating unit readiness and the problems associated with it. Readiness reports are called Unit Status Reports (USRs) in the Army, and are conducted at the Battalion level. Each service has a different report with its own problems. I will address only the Army's method. Battalions are the lowest level of organization in the Army that is considered able to support its own operations and is probably the highest level where commanders are fully aware of the condition of the unit. It consists of three to eight hundred soldiers with equipment dependent on its mission, i.e., Aviation, Armor, Infantry, or Administration. Table i-i compares a few typical battalions. As Table 1-1 shows, battalions are highly unique depending on their mission. What is important for an Administration Battalion might not be important for an Armor Battalion. Commands located above Battalion level are not pure, i.e., they may contain different types of units. This makes it difficult to report a consistent readiness above battalion level. The purity of battalions and their potential for independent operations make them an obvious level for status reports. Each month the battalion's key personnel gather information from a variety of sources, analyze it in examinating detail, and determine the combat readiness of the unit on a scale of 1 to 5.