Improving the accuracy of multiclass classification in machine learning: A case study in a cell signaling dataset

Abstract

It is important to make sense of the data within its context to propose a useful model to solve a problem. This domain knowledge includes information not contained in the data, but that will help us understand the data to be fed into a machine-learning algorithm and guide us on what features might help our model. Nevertheless, domain knowledge may become insufficient as the input variables increase, forcing the need to try automated feature selection techniques. In this study, we investigate whether the joint use of 1) feature selection techniques, such as Chi-square, Tree-based Feature Selection, Pearson’s Correlation, LASSO, Low Variance, and Recursive Feature Elimination, 2) outlier detection methods such as Isolation-Forest, and 3) Cross-Validation techniques lead to improving the accuracy in multiclass classification in machine learning. Specifically, we address the classification of patterns representing the activation state of cell signaling components into classes that symbolize the different cellular processes triggered in cancer cells. The results presented in this work have shown an accuracy increase with up to 80% fewer input features by only using 3 out of the 16 original descriptors.

Addressing the class imbalance in tabular datasets from a generative adversarial network approach in supervised machine learning

“…Note that the Running the simulation, Data farming Method, Data preparation, and EDA phases will not be treated in this work since these were already addressed in a previous study. 9…”

“…The Euclidean, Chebyshev, and Manhattan distances are a few of the distances frequently employed with the k−nn algorithm. 9…”

“…Note that the Running the simulation, Data farming Method, Data preparation, and EDA phases will not be treated in this work since these were already addressed in a previous study. 9 Note that data produced by the computational simulation is structured in rows and columns, where the first m − 1 columns of each record correspond to the predictive characteristics, and the mth column denotes the class or category. Each predictive characteristic represents the active/nonactive concentration level of a particular element belonging to the signaling network.…”

“…In a previous study, 9 we dealt with the dimensionality reduction of a cell signaling dataset for multiclass classification tasks by using a group of feature selection techniques, including Low variance, Chi-Squared, Recursive feature elimination, the LASSO, Tree-based feature selection, and Pearson’s correlation. The results produced by the dimensionality reduction analysis based on this group of feature selection techniques showed the importance or weight of eight predictive attributes (signaling elements) from the whole set of 16 attributes.…”

“…In this work, we present a machine-learning-based approach to address the class imbalance in the cell signaling dataset previously referred to and presented in González-Pérez and Sánchez-Gutiérrez. 9 In this approach, we use a conditional tabular GAN (ctGAN) to generate new instances for the minority class as if using the oversampling technique. We aim to assess how much the accuracy of supervised classification can be improved by using dimensionality reduction in conjunction with class balancing.…”