Optimized Classification Predictions with a New Index Combining Machine Learning Algorithms

Abstract

Voting is a commonly used ensemble method aiming to optimize classification predictions by combining results from individual base classifiers. However, the selection of appropriate classifiers to participate in voting algorithm is currently an open issue. In this study we developed a novel Dissimilarity-Performance (DP) index which incorporates two important criteria for the selection of base classifiers to participate in voting: their differential response in classification (dissimilarity) when combined in triads and their individual performance. To develop this empirical index we firstly used a range of different datasets to evaluate the relationship between voting results and measures of dissimilarity among classifiers of different types (rules, trees, lazy classifiers, functions and Bayes). Secondly, we computed the combined effect on voting performance of classifiers with different individual performance and/or diverse results in the voting performance. Our DP index was able to rank the classifier combinations according to their voting performance and thus to suggest the optimal combination. The proposed index is recommended for individual machine learning users as a preliminary tool to identify which classifiers to combine in order to achieve more accurate classification predictions avoiding computer intensive and time-consuming search.

Practices and Trends of Machine Learning Application in Nanotoxicology

“…Furxhi et al [72] proposed a composite score based on a Copeland index to rank classifiers according to their performance on diverse datasets, validation stages, and performance metrics. Tamvakis et al [133] proposed a dissimilarity performance index based on their voting performance to recommend the optimal ensemble combination. A variety of different datasets were used in this scenario to evaluate the relationship between voting results and dissimilarity measurements.…”

Missing data analysis using machine learning methods to predict the performance of technical students

“…This is because, in most case, an ensemble learning model is more accurate than any single model used separately. The effectiveness of ensemble learning models has been proved in different applications (Tamvakis et al, 2018).…”

Machine learning prediction of nanoparticle in vitro toxicity: A comparative study of classifiers and ensemble-classifiers using the Copeland Index

“…The most popular classifiers in predicting the toxicity of NPs range from artificial Neural Networks (NN) [1,10], Bayesian Networks (BN) [12,[25][26][27], Quantitative Structure Activity Relationships (QSARs aka nano-QSARs) [18,28,29], Linear Regression (LR), Random Forest (RF) and Support Vector Machines (SVM) [14,30,31]. Recently, integrated approaches (ensemble classifiers) are used to merge results from individual classifiers (base) in order to optimize the predictions [32][33][34][35][36][37]. Voting is a comprehensive ensemble learning method that collect votes from multiple base classifiers to predict the outcome via a voting mechanism to obtain a better predictive performance [38,39].…”

“…Classifiers are one of the most common ML tools exploiting experimental data such as physicochemical properties, quantum-mechanical attributes and toxicological outputs for nanotoxicity prediction [13,28,41]. Despite the wide variety and selection of classifiers and modelling approaches, no optimal classifier can be identified so far [32,42]. Instead the predictability of the classifier depends on the dataset characteristics (missing values, training size, input variables) or the methods used to assess classifier performance [32].…”

“…Despite the wide variety and selection of classifiers and modelling approaches, no optimal classifier can be identified so far [32,42]. Instead the predictability of the classifier depends on the dataset characteristics (missing values, training size, input variables) or the methods used to assess classifier performance [32].…”