Replication data for: evaluation of concept drift detection approaches for stock market forecasting

This package contains the datasets, source codes (in Python), and complete tables with the results reported in the work entitled Evaluation of Concept Drift Detection Approaches for Stock Market Forecasting. File "database.zip" contains, for each stock, (i) the original dataset, (ii) the preprocessed dataset (incremental and non-incremental preprocessing), and (iii) the dataset without normalization.<\li> File "classic_algorithms.zip" contains the source codes of the K-Nearest Neighbors (KNN), Random Forest (RF), Adjusted Random Forest (RF-A), and Support Vector Machines (SVM) models used in the experiments. This file also contains the experimental results of each model, in text files. File "grid_search_classic_algorithms.zip" contains the source code for hyperparameter tuning of the classical machine learning algorithms. File "algorithms_with_drift_detection.zip" contains the source codes (both for hyperparameter tuning and experimental runs) of the Online Sequential Extreme Learning Machines (OS-ELM), Dynamic and Online Ensemble for Regression (DOER), DOER with component ranking (DOER-Rank), Ensemble of Online Learners With Substitution of Models (EOS), and EOS with weighted average (EOS-D) models. These models were implemented according to [1]. Finally, the file "results.zip" contains a spreadsheet with the detailed results of the performed experiments. [1] BUENO BARAJAS, Jorge Andrés. Dynamic ensemble mechanisms to improve particulate matter forecasting: Mecanismos para ensemble dinâmicos aplicados para a previsão de material particulado. 2018. 1 recurso online (91 p.) Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Tecnologia, Limeira, SP. DOI: 10.47749/T/UNICAMP.2018.1031495