Performance Analysis and Ranking of Data Mining Algorithms Across Multiple Datasets

Data mining is a powerful emerging tool for analysis and prediction. It involves searching large volumes of data to discover interesting relationships that can provide meaningful information for understanding hidden trends and forecasting future incidences. Advanced mathematical machine learning algorithms are used in data mining. These algorithms vary in their accuracy and effectiveness depending on various factors. In this paper, a comprehensive analysis of the performance of several machine learning algorithms was conducted. The algorithms were analyzed using multiple performance metrics across multiple datasets. Two separate rankings were developed after evaluating the performance of twenty-three algorithms across ten nominal-class datasets based on nine performance metrics and seventeen algorithms across seven numeric-class datasets based on five performance metrics. The first ranking was based on the average score of algorithms and the second ranking was based on the top scoring algorithms. The results showed that Logistic Model Trees (LMT) and Random Forest (RF) were the top performing algorithms for nominal-class datasets and M5P and Linear Regression (LIR) were the top performing algorithms for numeric-class datasets.