How Do We Select a Combined Algorithm to Determine High-Quality Aerospace Researchers by Utilizing Data Mining Techniques?

Abstract

The aerospace industry and technology are always considered one of the country’s most important and valuable industries. The research area of "Aerospace" is among the priorities of the grand science and technology development strategies, and addressing it is strategically vital. The present research aims to estimate and predict the appropriate algorithm for identifying high-quality aerospace researchers based on Advanced Ensemble Classifier Techniques (AECT) in data mining on the outputs of scientometric analyses and predicting the most essential scientometric-related metrics to identify high-quality researchers. The present study was performed using the protocols of applied research and multiple methods. The studied population includes all aerospace researchers (1945 and 2021) indexed in "The Web of Science Core Collection (WOSCC)". DataLab software and multiple programming languages have been applied in this research. All three algorithms have an accuracy of 0.96 and an F1-score of 0.97, which indicates that the models have high accuracy, validity, sensitivity, and predictive power. The "Blending" algorithm is considered a suitable and predictive model. The output of the LightGBM algorithm is that the most important and robust metric in the evaluation of prominent researchers is a metric from the researchers' effectiveness dimension, the Q parameter. According to the knowledge obtained from the ability to predict AECT in the prediction of high-quality researchers, it is possible to use the metrics mentioned in the evaluation of researchers in the field of scientometrics for more accurate and comprehensive prediction. An algorithm that can lead to the optimal and efficient classification of researchers provides the possibility of in-depth analysis of the available data about researchers and smooths the predictive power of the most high-quality researcher. The use of the proposed algorithms in this research, while suggesting the appropriate algorithm, led to reliable and valuable knowledge in classifying high-quality aerospace researchers.

Keywords

• aerospace
• scientometrics
• data mining (dm)
• advanced ensemble classifier techniques (aect)
• light gradient boosting machine (lightgbm)
• confusion matrix