Abstract

In this work, we introduce the use of a weighted extreme learning machine (ELM) to give an automated predictive value to mass spectrometry data. In specific, the data obtained with Matrix-Assisted Laser Desorption/Ionization-Time of Flight (MALDI-TOF) technique are explored for balanced and unbalanced dataset scenarios, and compared with benchmarking machine learning algorithms (Naive Bayes, Support Vector Machine, Random Forest, and Logistic Regression). Finally, the evaluation of the performance of the proposed weighted ELM was realized in order to determine the most efficient technique in terms of predicting diseases. In the training phase, the weighted ELM reaches the 100% of accuracy, sensitivity and specificity, which are 25% and 30% higher than the rest of benchmarking machine learning algorithms. Meanwhile, in the testing phase results, the ELM observations highlight the scarce bias to predict positive and negative classes in unbalanced datasets.