A Simple Approach using Statistical-based Machine Learning to Predict the Weapon System Operational Readiness

Abstract

Weapon system operational readiness is a critical requirement to ensure the combat readiness in order to guarantee the state defense sustainability time by time. Weapon systems are only operated by the military and their readiness are programmed every year based on some factors such as the amount of the allocated budget, the weapon system strength, and its circulation. Usually, the weapon system readiness is programmed based on the planner’s experiences that are inherited from time to time. In this research, we proposed a simple approach by using statistical-based machine learning method called linear regression for helping the planner to predict the weapon system operational readiness faced to its affecting factors such as scheduled and unscheduled maintenance. We used a dataset from a randomized primary data for 5 years from year 2016 to year 2020 to predict year 2021. To ensure the performance of the model, two measurements are used namely, Mean Absolute Percentage Error (MAPE) to measure its accuracy and goodness, and R-squared (R²) to measure the ability of the independent variables, the weapon system circulation, influences the dependent variable, the weapon system readiness. From the measurement results, the models, in general, are able to achieve MAPE as much as 1.99% that has interpretation as very accurate prediction with the accuracy of 98.02%. On the other hand, the system is able to achieve R² as much as 84.15% that means the combination of the independent variables altogether have given a strong influence to the dependent variable. The higher the value of R² the better the model is. Our research conclude that linear regression is the proper machine learning model for predicting the weapon system operational readiness.