AN APPROACH TOWARD PREDICTION OF SM-CO ALLOY’S MAXIMUM ENERGY PRODUCT USING FEATURE BAGGING TECHNIQUE

The work aims to solve the problem of predicting magnetic properties on the example of Sm-Co alloy using artificial intelligence. In particular, the authors solved the Sm-Co alloys maximum energy product prediction task using the feature bagging technique. To implement this approach, we have chosen the Random Forest algorithm, which efficiently processes short data sets by reducing variance and, as a result, reducing the impact/avoidance of overfitting. Experimental modelling was based on a short set of data (190 observations) collected by the authors with many independent attributes. As a result, it has been experimentally established that the studied machine learning method provides a high value of the coefficient of determination - 0.78 when solving Sm-Co alloy’s maximum energy product prediction task. Furthermore, by comparing with other ensemble methods from different classes, the highest accuracy of the researched process is established based on various performance indicators.