This video demonstrates how machine learning techniques, specifically gradient descent and logistic regression, can extract decision-making weights from historical cases in Case-Based Reasoning (CBR) systems. Using as an example rescue helicopter pilots deciding whether to continue or abort combat zone missions, the presentation analyzes 96 historical cases across 14 parameters including weather, obstacles, ground fire, fuel levels, and patient condition. The logistic regression iterative process (essentially a classification method), systematically adjusts parameter weights over thousands of cycles until convergence, discovering quantifiable weights from historical data as an alternative to subjective expert opinions. This effort becomes invaluable when no current field experts exist, or when sufficient documented past cases warrant exploration, either as a standalone solution, or parallel to experts to identify discrepancies between historical patterns and expert judgment.
